Nature-positive goals for an organization’s food consumption

Abstract

Organizations are increasingly committing to biodiversity protection targets with focus on ‘nature-positive’ outcomes, yet examples of how to feasibly achieve these targets are needed. Here we propose an approach to achieve nature-positive targets with respect to the embodied biodiversity impacts of an organization’s food consumption. We quantify these impacts using a comprehensive database of life-cycle environmental impacts from food, and map exploratory strategies to meet defined targets structured according to a mitigation and conservation hierarchy. By considering the varying needs and values across the organization’s internal community, we identify a range of targeted approaches towards mitigating impacts, which balance top-down and bottom-up actions to different degrees. Delivering ambitious nature-positive targets within current constraints will be challenging, particularly given the need to mitigate cumulative impacts. Our results evidence that however committed an organization is to being nature positive in its food provision, this is unachievable in the absence of systems change. This Analysis illustrates how nature-positive targets aimed at protecting biodiversity can be achieved at the scale of organizations. A canteen at one UK university college is used as a case study for the application of a four-step participatory approach comprising an estimation of food-related biodiversity impacts; definition of biodiversity targets; assessment of possible interventions; and exploration of different strategies.

Full text

Nature Food | Volume 4 | January 2023 | 96–108 96
nature food
Analysis https://doi.org/10.1038/s43016-022-00660-2
Nature-positive goals for an organization’s
food consumption
I. Taylor 1 , J. W. Bull1,2, B. Ashton3, E. Biggs4, M. Clark4,5,6, N. Gray4,
H. M. J. Grub4, C. Stewart7 & E. J. Milner-Gulland 4
Organizations are increasingly committing to biodiversity protection
targets with focus on ‘nature-positive’ outcomes, yet examples of how to
feasibly achieve these targets are needed. Here we propose an approach to
achieve nature-positive targets with respect to the embodied biodiversity
impacts of an organization’s food consumption. We quantify these impacts
using a comprehensive database of life-cycle environmental impacts from
food, and map exploratory strategies to meet dened targets structured
according to a mitigation and conservation hierarchy. By considering the
varying needs and values across the organization’s internal community,
we identify a range of targeted approaches towards mitigating impacts,
which balance top-down and bottom-up actions to dierent degrees.
Delivering ambitious nature-positive targets within current constraints will
be challenging, particularly given the need to mitigate cumulative impacts.
Our results evidence that however committed an organization is to being
nature positive in its food provision, this is unachievable in the absence of
systems change.
Transformative actions are needed to address the triple challenge of
global biodiversity loss, climate change and improving human wellbe-
ing
1–3
. Bold targets are being proposed internationally (for example,
‘nature-positive’ targets that aim to achieve net-positive impacts on
biodiversity by 2030 relative to 2020 (refs. 4–6)) and nationally (for
example, UK Environment Act
7
and Biodiversity Net Gain policies
8
).
These targets are being translated to subnational levels (for example,
circular cities
9
). Organizations are committing
10,11
to strategic biodi-
versity targets
12
aimed at mitigating negative biodiversity impacts,
and increasingly to nature-positive outcomes in line with global
policy directions4,6,13.
The first step towards achieving these targets is to measure
biodiversity impacts. This enables organizations to design effective
impact-reduction strategies, assess progress towards targets and make
explicit contributions to wider environmental goals10,14–16. Targets and
strategies must also be designed in consultation with affected groups,
to ensure equitable and sustainable outcomes17.
One approach is to use the mitigation hierarchy framework, a
structured approach for impact mitigation towards a specified tar-
get (for example, net gain in biodiversity). It prioritizes prevention
before compensation, beginning with avoiding and reducing impacts
before restoration or offsetting of any unavoidable impacts
18
. In the
past, this framework has been primarily applied to impacts from the
infrastructure and extractive sectors, although it has been expanded
to agriculture and fisheries
19–23
and could be extended to all impacts
from human activities24,25.
However, reactive compensation is not enough to achieve
transformative change. Recently, the ‘mitigation and conservation
hierarchy’ (MCH) has been proposed26, which integrates the reac-
tive mitigation hierarchy with a ‘conservation hierarchy’ for actions
that proactively address historical, systemic and non-attributable
impacts
26
. It provides a framework to support individuals, commu-
nities, businesses and governments to meet ambitious biodiversity
targets. However, this framework has yet to be applied to the full range
Received: 16 February 2022
Accepted: 4 November 2022
Published online: 12 January 2023
Check for updates
1Wild Business Ltd., Kershen Fairfax, London, UK. 2Durrell Institute for Conservation and Ecology, University of Kent, Canterbury, UK. 3Lady Margaret Hall,
University of Oxford, Oxford, UK. 4Department of Biology, University of Oxford, Oxford, UK. 5Nufield Department of Population Health, University of
Oxford, Oxford, UK. 6Oxford Martin School, University of Oxford, Oxford, UK. 7Nufield Department of Primary Care Health Sciences, University of Oxford,
Radcliffe Primary Care Building, Radcliffe Observatory Quarter, Oxford, UK. e-mail: isobel.taylor93@outlook.com; ej.milner-gulland@biology.ox.ac.uk

Nature Food | Volume 4 | January 2023 | 96–108 97
Analysis https://doi.org/10.1038/s43016-022-00660-2
sourcing, which is often not available (including here), and their units
(for example, ‘species extinctions equivalents’) can be complex to
communicate. However, they do provide a data-driven approach for
estimating relative biodiversity impacts, which can be compared across
a broad scope of activities.
We estimate the college purchased ~156 tonnes of food over the
2018/2019 academic year, which required ~541,904 m2 of land, result-
ing in a negative biodiversity impact of ~7.9 × 10
−7
potential global
species extinctions equivalents (interpretable as a relative measure
of species extinction risk;
40
Fig. 1a). In line with prior research
27,34,37
,
the highest biodiversity impacts were driven by foods with dispro-
portionately large land footprints, including red meat, poultry and
fish (linked to 16.5%, 15.7% and 12.6% of total impacts, compared with
3.7%, 3.9% and 1.7% of consumption by mass, respectively; Fig. 1a).
Considerable impacts were also derived from products containing
ingredients sourced from highly biodiverse regions, including des-
serts, chocolate and confectionary (15.1% of impacts, compared with
5.8% by mass), as well as coffee and tea (9.3% of impacts, compared with
0.3% by mass) as has also been highlighted elsewhere
42,43
. Students
consumed the greatest quantity of food and had the highest overall
impact (36.8%), although the lowest impact intensity (4.6 × 10−12 spe-
cies eq. per kilogram food; Fig. 1b). Conference attendees contributed
a large proportion of overall impacts (33.0%) and had the highest
impact intensity (5.4 × 10
−12
species eq. per kilogram food). Support
of direct and indirect impacts that organizations need to tackle to meet
nature-positive targets.
For many organizations, a key consideration is the embodied
impacts from food consumption by its members
27
(individuals over
whom organizations can exert influence, for example, through food
options or information in canteens). Food systems are a major driver
of global biodiversity loss, with over one-third of land currently used
for agricultural purposes28 and ~88% of terrestrial birds, mammals and
amphibians predicted to lose habitat to further agricultural expansion
by 2050 (ref.
29
). Changes to food systems (for example, sustainable
production and trade, reduced food waste and shifting to healthy,
sustainable diets) will be essential for halting global biodiversity
loss
30
while simultaneously addressing issues of climate, food security
and health31–36.
Recently, life cycle analysis (LCA) has enabled development of
large-scale databases on the environmental impacts of foods
34
. In this
Analysis, we use these datasets to quantify impacts from food con-
sumption of a case study organization, applying the MCH to explore
feasible routes towards achieving biodiversity targets, considering
differences in risk and preference across various consumer groups. We
consider how bold but necessary nature-positive targets (for example,
cumulative biodiversity net gain) could be achieved in this context.
While others have measured biodiversity impacts associated with food
consumption27,37, we provide a robust, quantitative application of the
MCH to this crucial element of environmental impacts, generating
evidence on how organisations can contribute towards a science-based
global goal for nature4,12.
Approach overview
We focus on a higher education college (Lady Margaret Hall, Oxford
University, UK; herein, ‘the college’) and the community of individuals
that work, study and visit as part of its operations. The college canteen
provides food and beverages (herein, ‘food’) regularly for ~500 uni-
versity students, ~130 support and facilities staff, and ~75 academic
staff. The college also prepares food for commercial conference and
event attendees, and for external students attending summer school
programmes. All food is ordered and prepared through the college’s
central kitchen.
The college provides a useful case study owing to its detailed
records of food consumption for multiple groups, its keenness to
reduce its environmental impact and its controllable food system.
However, our approach is generalizable to any food-providing organi-
zation, particularly as we make use of consumption datasets that are
often readily available (sales and purchase data). Furthermore, the MCH
framework
26
is generalizable to other scales (for example, individu-
als tracking their food consumption), other forms of environmental
pressure (for example, greenhouse gas emissions) and other types of
activity beyond food consumption38.
The approach comprises four stages (Table 1): (1) estimating cur-
rent biodiversity impacts from food; (2) defining biodiversity targets;
(3) assessing possible interventions; and (4) exploring different inter-
vention combinations that achieve these targets. Each stage involved
consultation with end users (Methods). For stages 3 and 4, we predomi-
nantly focus on reactive impact mitigation, although we do discuss
possible proactive actions.
Stage 1: estimating biodiversity impacts
Biodiversity impacts from food served at the college were estimated
by pairing 2018/2019 kitchen purchasing data (by mass or volume per
product) with environmental LCA databases34,39. We applied a United
Nations Environment Programme-recommended biodiversity met-
ric
40
, which estimates the number of species destined for extinction
on the basis of land transformation and occupation in food production
locations. LCA approaches to biodiversity accounting have several
limitations41 (Methods). They ideally require information on food
Table 1 | Approach Overview
Stage Description Method
1: Baseline Estimate biodiversity
impacts from food
currently served,
identifying focal areas
of high impact (in terms
of both products and
consumer groups).
Combine consumption
data with LCA data and a
biodiversity metric40 to estimate
impacts per product type and
consumer group.
2: Targets Establish a set of possible
Speciic, Measurable,
Achievable, Realistic and
Time-bound (SMART)
targets for future levels of
biodiversity impact from
food consumption.
Identify target options based
on stakeholder consultations
and science-based best
practice (such as aligning with a
nature-positive target4)
Model annual and cumulative
impacts under each target
scenario over the target period to
gauge the level of effort required.
3: Actions Identify possible
interventions that would
reduce impacts under
each of the ‘four steps’:
Refrain, Reduce, Restore,
Renew.
Scan of relevant academic and
grey literature.
Consult with stakeholders to
identify existing interventions in
place at the college.
Assess each intervention
in terms of potential
effectiveness (‘technical
potential’).
Approximate changes to baseline
biodiversity impacts expected
under each intervention.
Assess each intervention in
terms of socio-economic
feasibility.
Conduct stakeholder
interviews with key individuals
at the college.
4: Strategy Explore strategies
(combinations of
interventions) for reaching
chosen targets using
different combinations of
interventions that balance
different risks to varying
degrees.
Model combinations of
interventions and assess
predicted progress towards
targets.
A summary of the approach used, applying the MCH framework proposed by
Milner-Gulland et al.26.

Nature Food | Volume 4 | January 2023 | 96–108 98
Analysis https://doi.org/10.1038/s43016-022-00660-2
staff, academic staff and summer school pupils accounted for 11.8%,
11.2% and 7.2% of total impacts, respectively. Further details, including
estimates for greenhouse gas emissions, are provided in Supplemen-
tary Information.
Stage 2: defining targets for impact reduction
We considered five ‘Specific, Measurable, Achievable, Realistic and
Timebound’ (SMART) targets to reduce biodiversity impacts from
food. Possible targets were explored with key stakeholders at the col-
lege (Methods), including an ambitious target broadly aligned with
nature-positive (cumulative biodiversity net gain) and intermediate
steps towards this goal. We modelled annual changes to the college’s
impacts that would be required under each target over a 15 year period
to 2035, given the pragmatic assumption of a slow start to allow for ini-
tial capacity building. A ‘business as usual’ (BAU) scenario was also mod-
elled, assuming that annual consumption remained broadly similar,
0
2
4
6
8
10
12
14
0
Fruit
Vegetables
Soft drinks and juices
Crustaceans and molluscs
Chocolate
Coee and tea
Rice
Egg and egg products
Sandwiches and wraps
Ready-made (veg)
Fish and fish products
Cooking oil
Wheat products (e.g., bread, pasta)
Nuts, pulses, seeds
Red meat
Poultry
Alcohol
Cake, biscuits and desserts
Other
Dairy and dairy alternatives
Potatoes and potato products
5
10
15
20
25
30
Total biodiversity impact (10–9 species extinctions eq.)
Total quantity of food (t or kL)
a
0
5
10
15
20
25
30
35
0
1
2
3
4
5
6
Students Conference
attendees
Support sta Academic sta Summer school
pupils
Total biodiversity impact (10–9 species extinctions eq.)
Biodiversity impact per kilogram or litre of food (10–13 species extinctions eq.)
b
Overall impacts per kilogram or litre
Other
Dairy and dairy alternatives
Vegetables
Fruit
Coee and tea
Chocolate
Cake, biscuits and desserts
Fish and fish products
Poultry
Red meat
Fig. 1 | Food consumption and biodiversity impacts at the college. a, Total
food quantities (in terms of mass or volume) consumed at the college by food
product category (shaded bars) and estimated total biodiversity impacts (in
terms of the additional impacts caused by agricultural land occupation and
transformation) by food product group (coloured bars). b, Biodiversity impacts
per consumer group: shaded bars show impacts per kilogram or litre of total
food consumption (a measure of biodiversity impact intensity) and coloured
bars show the total biodiversity impact per consumer group, separated by
food product. Further details, including impacts in terms of greenhouse gas
emissions, are provided in Supplementary Information.

Nature Food | Volume 4 | January 2023 | 96–108 99
Analysis https://doi.org/10.1038/s43016-022-00660-2
with a slight overall decline in impacts (approximately −10% relative to
annual impacts in the baseline year) driven by national dietary trends44,45
(Fig. 2 and Supplementary Table 1).
First, a process-based target was defined for switching to healthy
and sustainable diets in line with science-based best practice, as set
out by the EAT-Lancet Commission
33
(‘EL2035’, Fig. 2). It involves linear
change towards 100% adoption of a flexitarian planetary health diet by
2035. Achieving this target would reduce annual impacts by ~33.6% by
2035 against the 2018/2019 baseline. However, this would still result in
substantial cumulative biodiversity loss over the 15 year period, only
~12% less impact than under BAU. This demonstrates the importance
of understanding cumulative outcomes when target setting, and indi-
cates that gradual dietary shift alone is inadequate to tackle the full
extent of impacts.
The four remaining outcome-based targets (‘managed net loss
(MNL) (50%)’, ‘MNL (75%)’, ‘no net loss’ (NNL) and ‘net gain (10%)’ (NG10))
aim to reduce this 15 year cumulative biodiversity impact by a set amount
(respectively, by 50%, 75%, 100% or 110%) by 2035. Therefore, if the target
percentage reduction is missed in any year, it must be compensated for in
subsequent years. For example, under MNL (75%) we assume the college
begins working towards this target in 2021/2022, initially reducing annual
impacts by one-third each year until 2024. Annual biodiversity impacts
are then reduced to net zero by 2031, after which an annual net gain in
biodiversity is needed until 2035 to compensate for impacts incurred
during initial years, thereby achieving an overall cumulative impact
reduction of 75% relative to BAU (Supplementary Table 1).
This demonstrates the ambitious nature even of relative targets
such as MNL75, which would still result in a cumulative net loss of
biodiversity. Additional uncertainty is introduced by MNL targets
being calculated relative to a dynamic counterfactual (BAU), which
may change depending on factors that have not been modelled here
(for example, changes in student numbers, unanticipated effects of
coronavirus disease 2019, or changes in food production practices or
efficiency). Achieving more stringent nature-positive targets for cumu-
lative biodiversity net gain (for example, NG10) would thus require
urgent and substantial action38.
Stage 3: assessment of interventions
We collated a set of 44 interventions (actions that mitigate biodiversity
impacts from food; Supplementary Table 2) from the academic and
grey literature, categorized according to the four steps of the MCH
26
(Refrain, Reduce, Restore, Renew). Interventions included top-down
and bottom-up approaches, as well as environmentally sustainable
sourcing and options for compensation. We considered each inter-
vention’s technical potential (effectiveness for reducing biodiversity
impacts) and socio-economic feasibility (or ‘initiative feasibility’
46
),
following previous research21,22,46,47; Table 2).
Technical potential was quantified on the basis of biodiversity
impacts estimated in stage 1, further informed by relevant academic
literature. Socio-economic feasibility was qualitatively assessed for
each consumer group through semi-structured interviews with key
stakeholders (the head chef, catering manager and domestic bursar;
Supplementary Table 2). It was not within scope to conduct an in-depth
review of behaviour-change interventions, and we acknowledge the
limitations of our assessments (for example, limited accounting for
behavioural plasticity
46
; Methods). A more comprehensive analysis of
behaviour-change interventions—utilizing specific behaviour-change
frameworks (for example, refs. 48–50)—may have led to different conclu-
sions. However, the expertize of our stakeholders provides a useful
basis for decision making and strategy prioritization.
Stakeholder consultations revealed key considerations and con-
straints at the college, including: (1) ensuring consumers’ wellbeing
by providing healthy and nutritionally balanced food, (2) ensuring
freedom of choice and a variety of meal options, (3) ensuring ethical
and sustainable sourcing, and (4) budgetary constraints.
Interventions covered a broad spectrum of socio-economic fea-
sibility and biodiversity risks (Table 2 and Supplementary Table 2).
Top-down interventions to restrict the highest-impact foods (‘Refrain’)
0
2018 2020 2022 2024 2026 2028 2030 2032 2034
Cumulative net biodiversity impact
Academic year end
a
BAU EL2035 MNL50
MNL75 NNL NG10
2018 2020 2022 2024 2026 2028 2030 2032 2034
Annual net biodiversity impact
Academic year end
b
Biodiversity loss
Biodiversity gain
Biodiversity loss
Biodiversity gain
0
Fig. 2 | Net changes in cumulative and annual biodiversity impacts from
food, modelled for five illustrative target scenarios and a BAU scenario. a,b,
Values below ‘0’ represent a cumulative (a) or annual (b) net positive impact on
biodiversity (that is, biodiversity net gain). The BAU scenario assumes similar
impact each year with a slight declining trend based on data from the UK National
Diet and Nutrition Survey44. EL2035 represents a process-based target for the
college to switch to serving healthy and sustainable diets by 2035, based on
EAT-Lancet recommendations33. MNL targets aim to reduce cumulative impacts
relative to BAU by 50% (MNL50) or 75% (MNL75). NNL and NG10 aim to mitigate
100% of absolute cumulative impacts, with additional compensation to achieve
10% biodiversity net gain under NG10. For more detail, see Supplementary Table 1.

Nature Food | Volume 4 | January 2023 | 96–108 100
Analysis https://doi.org/10.1038/s43016-022-00660-2
had higher technical potential, but lower levels of feasibility due to the
social risk of restricting consumer choice and the potential for leakage
(consumers deciding to eat high-impact food products elsewhere).
The reverse was true for bottom-up interventions aimed at shifting
consumer choice (that is, behaviour-change interventions; ‘Reduce’).
Environmentally sustainable sourcing (‘Reduce’) had strong but highly
uncertain technical potential, with effectiveness contingent on supply
chain transparency10,51–53, product affordability and extent of leakage
(displacement of sourcing impacts to other organizations given supply
of sustainable products may be limited). Furthermore, the low biodi-
versity impact of these products may trade off with other aspects of
sustainability important to the college (for example, greenhouse gas
emissions or animal welfare34).
Compensatory actions under the mitigation hierarchy could
include restoring biodiversity directly affected by the college’s food
consumption (‘Restore’, for example, on or near farms where ingre-
dients are sourced) and restoring equivalent biodiversity (offset-
ting) elsewhere (‘Renew’)18. Our chosen biodiversity metric has many
assumptions (Methods), and there is limited information on food
origin or production practices, limiting the accurate calculation of
ecologically equivalent54 areas for restoration that would compensate
for negative impacts. Other uncertainties include ensuring additional-
ity, long-term monitoring, compliance and cost55.
Once all impacts have been mitigated, an aspirational bio-
diversity net gain target can be achieved through proactive
biodiversity-enhancing actions26. Examples include contributing
to research, education and innovation in sustainable food systems
(Refrain/Reduce), supporting local restoration/re-wilding projects
(Restore) or creating community food gardens (Renew). Although these
actions are challenging to quantify and cannot be counted towards a
biodiversity net gain target until direct impacts have been mitigated,
they may help avert future biodiversity losses from food systems and
can build wider support for biodiversity and its social benefits.
Stage 4: exploring strategies for impact mitigation
Our final step was to investigate the feasibility of achieving each of the
outcome-based targets described in stage 2 by combining interven-
tions. A set of five exploratory strategies was constructed, balancing
risks and uncertainties identified at stage 3 to varying degrees (Fig.
3 and Supplementary Table 3). The biodiversity impact mitigation
potential of each strategy was quantified on the basis of stage 1 and 3
results, and provides an approximation intended to inform organiza-
tional decision making.
Strategy A aims to prevent (‘Refrain’ and ‘Reduce’) as much bio-
diversity impact as possible while providing a healthy diet. It involves
the college switching to nutritionally balanced vegan food, with all
ingredients sourced from best-practice suppliers for biodiversity
(assuming the same overall mass of food is served). Given the techni-
cal and feasibility assessments, this strategy is unlikely to be socially,
financially or logistically achievable in the near future. Risk of leakage
(Table 2) means the potential benefits of this strategy are unlikely to be
realized until system-wide changes beyond the college’s direct control
are implemented. If such changes were to occur, the scenario indicates
that ~83% of annual biodiversity impacts calculated in stage 1 could
theoretically be preventable; ~42% from serving vegan food and ~41%
from biodiversity-friendly sourcing (Fig. 3). If the college also halved
its consumption of coffee, chocolate and palm oil, preventable impacts
could be up to ~88%. However, a ‘flexitarian’ diet (allowing for small
amounts of meat, fish, dairy and eggs)33 combined with best-practice
sourcing could still prevent ~79% of impacts, indicating that consider-
able progress could be made without requiring controversial measures
such as banning animal products.
Strategies B–E show various combinations of interventions
considered potentially feasible for the college, based on stage 3
results. They range from a top-down ‘avoidance-focused’ approach
(lower risk for biodiversity, but high choice infringement) to a
bottom-up strategy focusing on behavioural interventions and
best-practice sourcing (high biodiversity uncertainty, but less risk
for consumers). ‘Refrain/Reduce’ actions could enable the college
to make good progress towards each target (~37–42% impact mitiga-
tion for strategies B–E). However, achieving more ambitious targets
would require a considerable level of restoration/offsetting (68–73%
of impacts for NG10), particularly when compared with strategy A.
Given the need for a slow start to build capacity, this issue would be
exacerbated by the need to over compensate in later years to reach the
specified cumulative target (see stage 2).
Furthermore, while targets could in theory be achievable through a
more bottom-up strategy (for example, strategy E), large uncertainties
for behavioural and sourcing interventions make mitigation difficult
to predict, achieve or measure. While such interventions are valuable,
enacting top-down measures will be key to ensuring that biodiversity
outcomes are achieved. Assessments of financial feasibility for each
strategy would also need to be carried out, accounting for product
purchase and offset costs, commercial viability, fair pricing for con-
sumers, and potential funding streams.
These results highlight the challenge for organizations in
achieving nature-positive targets (for example, NG10), since even
avoidance-weighted strategies (for example, strategy B) would incur
substantial residual impacts, requiring considerable levels of offset-
ting to meet targets within current socio-economic constraints. The
challenge is enhanced by the difficulties of calculating, delivering and
monitoring offsets11,23,54. particularly when limited sourcing informa-
tion is available. Crucially, issues around leakage suggest that achieving
true nature-positive outcomes would require urgent systemic action
beyond the direct influence of the college56.
Recovering biodiversity one organization at a time
Here we have applied the MCH to address the globalized impacts of an
organization’s food consumption, quantifying biodiversity impacts
and framing potential targets and strategies within context-specific
constraints. By considering varying levels of impact and feasibility for
different consumer groups, we highlighted variation in opportunities
and constraints, and suggested targeted group-specific strategies.
Accounting for behavioural plasticity (through trialling and monitor-
ing interventions) and gathering further data on food sourcing would
inform better-targeted approaches for impact mitigation and com-
pensation in future.
Our approach is generalizable across scales, environmental pres-
sures and sectors
26,38
. Our case study organization is both an educa-
tional institution and an events catering business, so results may be
applicable across a broad range of food-providing organizations.
However, it is hard to know how scalable our results are until similar
analyses are carried out, and we recognize that consumers at the col-
lege may not be representative of the wider population.
Transparency regarding the scale of the challenge is essential
to achieving ethical and sustainable paths towards nature-positive
goals17,57. Here we show for one organization that mitigating the embod-
ied biodiversity impacts of their food may not currently be feasible. A
recent example applying this framework to the operational biodiversity
impacts of Oxford University38 similarly found that strategies consid-
ered feasible by the focal organization left substantial residual impacts
needing to be offset to achieve a net gain target.
However, reversing global biodiversity loss remains urgent and
necessary
1,4
, and will require organizations to take rapid and ambi-
tious action on food and other elements of their operations. Actions
will in some cases need to be top down, may result in socio-economic
risk and will require engagement with affected communities. Delay-
ing action will lead to extensive negative impacts requiring compen-
sation later. However, if more organizations commit to ambitious
nature-positive pathways, issues of leakage are likely to reduce

Nature Food | Volume 4 | January 2023 | 96–108 101
Analysis https://doi.org/10.1038/s43016-022-00660-2
(for example, through shifting dietary norms and creating greater
demand for biodiversity-friendly produce), and strategies that appear
highly ambitious for a single organization today may become feasible.
We have evidenced the urgent need for wider transformative
change to minimize the impacts of food systems
30,33,56
, which will allow
consumers and food providers to make more sustainable choices
Table 2 | Technical and feasibility assessments
Intervention
category: REFRAIN,
REDUCE, RESTORE
or RENEW
Technical
mitigation
potential
(average % impact
reduction)
Biodiversity risks and
beneits Socio-economic
feasibility: inancial
and logistical risks and
beneits
Socio-economic
feasibility: consumer
risks and beneits
Recommendations
REFRAIN from
serving the most
impactful foods
(for example,
‘top-down’
restrictions on
meat, ish, coffee,
chocolate and so
on)
6.2% (2.5–14.8%)
(N=11) Generally lower risk due
to focus on prevention
(not compensation). Risk
of leakage if consumers
seek restricted foods
elsewhere (particularly if
highly valued/available).
May be time- and
resource-consuming
for kitchens to produce
a good variety of
low-impact choices.
Ingredient costs may
increase or decrease,
depending on the
product/replacement.
High choice
infringement risk for
those who eat regularly
at the college. Some
(for example, academic
staff) have previously
opposed restrictions
on meat. Others (for
example, student
groups) have advocated
for restrictions.
(1) Apply to conference/ summer
school attendees who eat
infrequently at the college; (2)
avoid refraining from serving highly
valued/easily accessible foods (for
example, coffee and chocolate) to
minimize leakage; (3) for students/
staff, focus on sustainable sourcing,
bottom-up interventions and
reducing use where possible.
REDUCE
consumption of
impactful foods
through ‘bottom-up’
interventions
(for example,
behaviour change
interventions)
4.0% (1.0–9.4%)
(N=4) Uncertain effectiveness
as this would be
context dependent
(for example, owing to
behavioural plasticity);
reliant on long-term,
bottom-up changes in
dietary choices, rather
than instant changes
to organizational food
purchasing.
Most interventions
would be low cost and
simple to carry out (for
example, rearranging
menus), others may
require more resources
and collaboration (for
example, eco-labelling).
No restriction on
consumer choice,
although could be
perceived as choice
manipulation.
Awareness raising
interventions could
provide educational
beneits.
(1) Implement simple interventions
soon (for example, increasing ratio
of low-impact options on conference
menus); (2) collaborate to implement
more complex interventions
(for example, eco-labelling); (3)
monitor implemented behavioural
interventions to improve estimates
of effectiveness and behavioural
plasticity.
REDUCE impacts
through sustainable
sourcing of
ingredients
(for example,
buying certiied
biodiversity-friendly
products and using
local suppliers)
8.8% (2.4–1 8.8%)
(N=7) High uncertainty
due to lack of supply
chain transparency;
biodiversity-friendly
food production may
trade off against other
aspects of sustainability;
impacts may be
displaced to other
organizations (leakage).
More environmentally
sustainable produce
may have more limited
availability and higher
cost.
Avoids risk of choice
infringement.
May result in pricing
changes at the college
for consumers.
(1) Source environmentally
sustainable foods where
budget allows (for example, in
more commercial aspects of food
provision), particularly for impactful
foods that are unable to be avoided
entirely; (2) identify opportunities
for low-cost, seasonal and local
sustainable sourcing (for example,
allotment produce, repurposing
excess food and so on).
REDUCE impacts
by reducing
food waste (for
example, eficient
use of ingredients
and repurposing
leftovers)
Not possible to
calculate here
owing to data
limitations.
Reducing the amount
of food wastage
would help to reduce
biodiversity impacts.
There is uncertainty in
how much ‘repurposed’
wasted food would be
replacing rather than
adding to consumption.
The college already
implements several
measures to minimize
food waste;
resources would be
needed for monitoring
and communication
with food redistribution
networks.
No identiied negative
impact on consumers.
Consumers may beneit
from discounted
leftovers.
(1) Participate in existing lexible food
redistribution schemes as and when
required; (2) increase rates of onsite
composting;
(3) improve monitoring of waste
streams.
RESTORE impacted
biodiversity and
RENEW through
biodiversity
restoration offsets
to compensate for
residual impacts
in ecosystems
affected by college
food consumption.
Dependent
on the nature
and extent of
residual impacts
and the chosen
biodiversity
target.
High risk unless carried
out according to best
practice54, complicated
by a lack of supply
chain transparency and
assumptions made in
the biodiversity metric
calculation.
Likely to be expensive,
requiring novel funding
streams, availability of
expertize, and adequate
like-for-like offsets
on the market (for
example, biodiversity
credits).
May result in pricing
changes at the college
for consumers.
(1) Seek greater transparency in all
college food supply chains to enable
appropriate targeting of offsets to
the site of biodiversity impact.
Proactively
REFRAIN, REDUCE,
RESTORE and
RENEW biodiversity
(for example,
research,
innovation, and
local or global
conservation
initiatives).
Not normally
quantiiable. Must not be used as a
substitute for offsetting
of speciic, quantiiable
biodiversity impacts (to
prevent greenwashing).
Resource or funding
required but potential
for cost-effective
initiatives that produce
socio-economic value.
May deliver added
social beneits (for
example, opportunities
for college members,
or empowering local
communities in areas of
impact).
(1) Pursue proactive actions that
will help reduce future impacts
(for example, supporting research,
innovation and education on
sustainable food systems); (2)
maximize the social beneits of
proactive actions; (3) do not replace
quantiiable impact mitigation with
proactive actions.
Results of the technical and feasibility assessments for different categories of interventions, under the four steps of the MCH—Refrain, Reduce, Restore and Renew. Ranges provided for the
‘Technical mitigation potential’ column show the range of average biodiversity impact reduction across the number of interventions that were assessed (‘N’) and are not an indication of
statistical uncertainty (which cannot be estimated). Additional detail on speciic interventions, consumer group-speciic assessments and sources is provided in Supplementary Table 2.

Nature Food | Volume 4 | January 2023 | 96–108 102
Analysis https://doi.org/10.1038/s43016-022-00660-2
at lower socio-economic cost
57
. Measures will be required across all
scales (for example, government incentives, improved production and
procurement practices, improved market standards, increased moni-
toring and transparency, improved biodiversity metrics, systematic
approaches to offsetting and shifts in consumer behaviour) and will
need to be integrated with other societal goals34,51,58,59. Organizations
can make genuine, positive contributions to catalysing these changes
through mitigation of impacts and proactive conservation efforts23.
As new global biodiversity targets are being set, organizations have a
critical opportunity to change their food systems, bringing themselves
closer to a nature-positive future. If enough of them do this, societal
transformation can be provoked.
Methods
Context and participatory approach
This research was based at a University of Oxford college (Lady Mar-
garet Hall, Oxford, UK) and was carried out during July to December
2020. The college community consists of ~500 university undergradu-
ate and postgraduate students, ~130 members of support staff and
~75 members of academic staff. The college has a central canteen,
providing daily food and beverages (herein, ‘food’) at subsidized rates
for students and free of charge for staff. The college also runs a com-
mercial food service, frequently preparing food for conference, event
and annual summer school attendees. All food is ordered and prepared
through a central kitchen coordinated by a head chef and catering
manager. Along with the college’s domestic bursar, these individuals
were considered the primary end users of this applied research and
were involved throughout its inception and delivery in a participatory
and iterative manner.
Stage 1: estimating baseline quantities and biodiversity
impacts of food served at the college
Sources of food purchasing data. Our analysis was based on food pur-
chased by the college during the financial (academic) year of 2018/2019,
chosen as the most recent year for which a complete dataset was avail-
able and before the impacts of the coronavirus disease 2019 pandemic.
For feasibility, analyses were based on 3 months of invoice data from
September 2018, February 2019, and July 2019. These months captured
different aspects of the college’s food service: the main conference
season, standard term time and summer schools, respectively. Data
were primarily obtained from the college’s online procurement system,
supplemented with data from eight additional suppliers. All aspects
of food under the control of the college’s catering department were
captured; this did not include food prepared by students. The final
dataset consisted of 4,651 individual purchase records, 1,612 unique
food products and included information on date, supplier, product
code and description, number of units purchased, and cost per unit.
Information on product mass or volume per unit was included for 37%
0
25
50
75
100
A B C D E
Percentage of impacts from food consumption mitigated
Mitigation strategy
RENEW: osets + proactive actions under
the conservation hierarchy
RESTORE and oset residual impacts
REDUCE impacts through best-practice
sustainable sourcing
REDUCE consumption of impactful food
through behavioural interventions
REDUCE the amount of impactful foods
oered
REFRAIN completely from serving the
most impactful foods
Mitigation strategies:
(A) 'Preventable impacts'
(B) 'Avoidance-focused'
(C) 'Mixed approach'
(D) 'Reduce-focused'
(E) 'Behaviour-focused'
Top-down strategy:
Low biodiversity risk
High social risk
NG10
NNL
MNL75
MNL50
Baseline
Bottom-up strategy:
High biodiversity risk
Low social risk
Fig. 3 | Comparison of five strategies for mitigating the biodiversity impacts
of food served at the college. A breakdown of individual interventions per
strategy can be found in Supplementary Table 3. Each strategy is represented
by a bar covering 1 year’s worth of biodiversity impacts. The y axis shows the
approximated mitigation potential, such that 0% represents no action taken
(stage 1 baseline impacts) and 100% represents NNL of biodiversity resulting
from college food consumption. Dotted lines represent average annual impacts
required under the different cumulative targets described in Fig. 2 (recognizing
that in some years these targets may be missed, whereas in other years targets will
need to be surpassed to compensate for missed years). Interventions that refrain
from or reduce impacts (shown in blue and green) should be given precedence
over compensatory actions such as biodiversity offsets (orange and red). Here,
compensatory actions are considered only in terms of the extent of residual
impacts needing to be restored or offset, rather than the areas or types of species
or ecosystems to be targeted. Note that the values provided here are indicative
for use in informing pragmatic policy decisions and are based on numerous
assumptions, which are described in Methods and Supplementary Information.

Nature Food | Volume 4 | January 2023 | 96–108 103
Analysis https://doi.org/10.1038/s43016-022-00660-2
of products. Records were sorted into 63 product categories for further
analysis, based on food type and price of product.
Estimating quantities of food served. Owing to limited available
information on product mass or volume, these values were directly
estimated for the 200 most frequently purchased products (which
accounted for 70% of all purchased units). Estimates were based on
information within supplier catalogues, other college purchasing
reports or supermarket websites. All food categories included a mini-
mum of 50% of items with direct mass or volume estimates (average per
category: 89%). For the remaining 41% of unique products with no mass
or volume estimate, this was approximated on the basis of an average
cost-to-mass (or volume) ratio for each food category. For records
with an unknown product description (9% of records), mass or volume
of the product was approximated based on the known breakdown of
food categories and average gram or millilitre per £1 for the product
supplier. This approach was deemed reasonable, as most ‘unknown’
products came from specialized suppliers (for example, butchers).
However, most records had detailed estimates and final values were
sense checked by college stakeholders.
Estimating biodiversity impacts from food. The 1,589 unique prod-
ucts were matched to a corresponding item in one of the following
environmental datasets to obtain its biodiversity impact value:
(1) A dataset derived from Poore and Nemecek34,60, of 55 raw
ingredients with their associated environmental impacts,
including land occupation and transformation values, based on
a meta-analysis of global LCA studies. The LCA system boundary
ranged from agricultural inputs at farm stage to retail stage (for
details, see ref. 34). Environmental values were provided for each
ingredient at three levels of impact based on the range of pro-
ducers assessed: low (5th percentile), average (50th percentile)
and high (95th percentile). Biodiversity impacts were quantied
using a United Nations Environment Programme-recommended
metric developed by Chaudhary et al.40. This models the number
of expected global species extinctions (‘species extinctions
equivalent’) for a given area of land occupation (extent × time
occupied) and transformation (land use change) per food item
and country, based on the countryside species–area relation-
ship model and species vulnerability scores (including levels of
threat and endemism). Further details on underlying models are
provided in Chaudhary et al.40.
(2) An extensive database of products from six major UK online
supermarkets (‘foodDB’, developed by Harrington et al.39 and
used under license). An extract containing back-of-packet in-
gredients for 2,138 unique supermarket products was obtained.
Products consisted of composite food items in categories such
as ready meals, sandwiches, desserts and sweet treats, pies and
quiches, as well as specic vegan and vegetarian products. Raw
ingredients for each product were paired with corresponding
ingredients in dataset 1 and weighted according to quantity
to derive overall biodiversity impacts per 100 g of supermar-
ket product. Full details on the dataset and methodology are
described in Clark et al.61,62.
(3) A dataset derived from foodDB, with biodiversity impact values
aggregated at the supermarket ‘shelf’ level, as opposed to spe-
cic products. This dataset captured a broader range of food
items than dataset 2 (including 3,687 shelf categories) and was
used for more generic categories of composite food items (such
as ‘red wine’, ‘chocolate bars’, ‘dairy-free cheese’ and so on). It
was applied when an appropriate match could not be identied
in dataset 1 or 2, and where a specic product was not required
(that is, owing to low intra-category variation in environmental
impacts).
Biodiversity impacts could then be derived by multiplying the
mass or volume of each product by the impact values in the correspond-
ing databases. Generally, the median (50th percentile) impact values
were used (but see below).
Accounting for existing sustainable sourcing efforts. Consulta-
tions with college stakeholders identified that certain products were
routinely sourced with sustainable certifications (for example, all pur-
chased tea, coffee and sugar was either Rainforest Alliance or Fairtrade
certified). To broadly account for this, a weighted combination of 50th
(64%) and 5th (36%) percentile impacts was used for these certified
products. This was based on the assumption that certified products
were likely to have some positive impact on biodiversity relative to
average (50th percentile) products, but these positive effects could
be weak or uncertain. Percentage weightings were based on DeFries
et al.
53
. While this was a broad approximation, it was an evidence-based
way to account for actions already undertaken in the absence of more
detailed data on product-specific impacts.
Scaling up and allocating impacts. To calculate biodiversity impacts
for the full baseline year (2018/2019) and allocate portions of impacts to
each of the college’s main consumer groups outlined above, estimates
for the three focal months were factored up on the basis of a breakdown
of annual food costs provided by college stakeholders. This was done by
calculating the average quantity (mass or volume) or impact (species
extinctions equivalents) per £1 spent for each consumer group across
the three focal months, and then multiplying this value by the overall
amount spent per consumer group for the full year.
Key uncertainties and assumptions for stage 1. The methods used
here were the best available at the time and provided a set of estimates
from which decisions could justifiably be made by college stakeholders.
However, they are necessarily broad and based on several assumptions:
i. When factoring up estimates based on food costs, we
assumed that food consumed during the three focal months
was representative of the food consumed during the rest
of the baseline year (2018/2019). Thus, there may be some
unaccounted-for seasonal variation (although college stake-
holders stated this was minimal). Furthermore, this analysis
focuses on a single year.
ii. The chosen biodiversity metric only measures impact on the
basis of one component of biodiversity (species) ignoring
other components, such as habitats, ecosystems and genetic
or functional diversity. Its development was based on datasets
biased towards certain taxa (terrestrial vertebrates and vascular
plants) and geographic regions (often high income or high bio-
diversity), not accounting for the impact of agriculture on taxa
important to continued ecosystem functioning (for example,
soil microbes, plants and arthropods) or freshwater and marine
taxa. The metric is intended for calculating relative levels of
impact on biodiversity, and cannot be interpreted in terms of
absolute impact owing to assumptions made in its calculation40.
iii. The biodiversity metric focuses only on the eects of land
use on biodiversity, not directly including other important
pressures on biodiversity, such as over-extraction of water, pol-
lution, climate change, direct exploitation of wild populations
or invasive species. The relative biodiversity impact for food
products may therefore be under- or overestimated. For exam-
ple, relative biodiversity impacts of red meat might be higher if
the indirect impacts on biodiversity from climate change were
considered (Supplementary Information).
iv. In the absence of information on the region from which food
products were sourced, datasets were based on global average
LCA data per product. Any region-specic dierences in food

Nature Food | Volume 4 | January 2023 | 96–108 104
Analysis https://doi.org/10.1038/s43016-022-00660-2
product impacts were unable to be accounted for. While assum-
ing global average sourcing introduces uncertainty, high-level
sensitivity analyses indicated this assumption is unlikely to
change the main conclusions. Ingredients with the highest
environmental impacts in the baseline scenario tended to also
have the highest impact whether sourcing from best-case (5th
percentile impacts) or worst-case (95th percentile impacts)
producers (Supplementary Information). However, when
product-specic information on food sourcing becomes avail-
able, future analyses could incorporate biodiversity impacts
that reect an organization’s actual food supply chains. This
would have additional benets, for instance providing insight
into how an organization’s negative impact on biodiversity
might be compensated by actions with a ‘like-for-like’ biodiver-
sity benet (for example, in the same country and ecosystem
type).
v. Certain food categories are likely to be less accurate than oth-
ers. All sh was assumed to be produced through aquaculture
rather than wild caught (college stakeholders said this was
representative of their sh sourcing). Additionally, impacts for
lamb and mutton products were based on a single set of LCA
values, meaning that 5th and 95th percentile impacts could not
be estimated. Given that lamb was purchased in quantities an
order of magnitude lower than other meats, it is unlikely that
this had a substantial eect on the overall results.
vi. With regard to datasets 2 and 3, limited data were available
for ingredient composition of certain products, so estimated
environmental impacts have limited reliability. However, these
products tended to be ‘low-impact’ foods, so this lack of data is
unlikely to have substantial eects on overall results.
Stage 2: constructing target scenarios
Modelling the BAU scenario. The BAU scenario was calculated on
the assumption that the quantity and type of food served at the col-
lege would remain similar each year until 2035. Slight changes were
predicted to occur in certain food groups, based on trends in the UK
government’s National Diet and Nutrition Survey
44
, including average
annual increases of 2.7 g of vegetables and 1.1 g of poultry, and average
declines of 2.1 g of red meat and 4.1 ml of fruit juice per person per day.
These trends were converted to percentage changes, based on known
total average daily intakes per person per ingredient in the National Diet
and Nutrition Survey data, and yearly percentage changes were applied
to baseline impacts described in stage 1. This BAU scenario does not
consider possible changes to community numbers and associated con-
sumption patterns in future years, as this information was unknown.
Process-based target: modelling the ‘Healthy and Sustainable
Diets’ target (‘EL2035’). The ‘EL2035’ target was calculated on the
basis of best practice in healthy, sustainable diets, as set out by the
EAT-Lancet Commission on Healthy Diets from Sustainable Food Sys-
tems
33
. Food products purchased by the college were disaggregated
into their constituent ingredients, using datasets 2 and 3 where nec-
essary (that is, for composite food products). This provided a pro-
portional breakdown of the total mass or volume of raw ingredients
served at the college during the baseline year. To predict impacts under
an EL2035 target in 2035, these proportions were altered to match
the proportions of an EAT-Lancet ‘flexitarian’ diet (assuming 100%
uptake, specific dietary values taken from Supplementary Table 7 in
Springmann et al.31), with the assumption that overall mass or volume
of food served would remain constant. These new proportions were
combined with biodiversity impacts per food product (as described
above) to estimate overall impacts. Consumption of ingredients that
were not captured within the EAT-Lancet dietary recommendations (for
example, tea and coffee) was assumed to remain the same.
Outcome-based targets: modelling cumulative targets. The four
remaining outcome-based targets (‘MNL50’, ‘MNL75’, ‘NNL’ and ‘NG10’)
aim to reduce cumulative biodiversity impacts by a set amount by 2035
(respectively by 50%, 75%, 100% or 110%). Targets were modelled on the
basis of their overall cumulative outcome, with a period of capacity
building in the initial years of the scenario (Supplementary Table 1).
These scenarios were chosen to represent increasing ambition towards
achieving a nature-positive target.
Stage 3: assessment of interventions
Collation of interventions. Possible interventions to reduce biodiver-
sity impacts from food at the college were identified from a high-level
review of the academic literature (a full systematic review was out of
scope). The search strategy involved identifying existing systematic
academic reviews of behaviour-change interventions63–66, along with
key sources of grey literature
67
, and expanding the search following
a snowball sampling strategy (following references and citations)
until no additional intervention categories were highlighted. A cat-
egory could include several similar interventions, such as promoting
pro-environmental social norms through various forms of advertising.
These were further refined on the basis of consultations with stakehold-
ers to identify and exclude any interventions already in place at the
college (and therefore captured under baseline impact calculations).
The final list of interventions along with relevant sources is provided
in Supplementary Table 2.
Technical assessment. Percentage reductions in biodiversity impact
(technical potential values) were calculated for each intervention using
one of three approaches, depending on intervention type:
1. For top-down interventions that restricted quantity of certain
foods served at the college, the change in impacts was esti-
mated by substituting highly impactful foods with the same
mass or volume of their lower-impact equivalents (for example,
by replacing 100 kg of red meat with 100 kg of plant-based meat
alternative products). Replacements were made on the basis of
ingredient mass or volume, not on the basis of calories or nutri-
tion. This was considered practical, as it reects the functional
replacement of an ingredient in a recipe, which is more easily
communicated to catering sta.
2. For interventions focused on sourcing ingredients from produc-
ers with the lowest impacts on biodiversity, we used the range
of impact levels provided by dataset 1. Average (50th percentile)
impact values used in calculating the baseline were replaced
with low (5th percentile) impacts (for example, replacing 100 kg
of average red meat with 100 kg of red meat sourced from the
top 5% of producers in terms of reducing biodiversity impacts).
3. ‘Bottom-up’ interventions (for example, behaviour-change
interventions) were challenging to predict since these are
highly dependent on context and consumer responses. We used
indicative estimates of impact based on previous examples in
the literature (Supplementary Table 2), broadly assuming that
the proportional change estimated in these studies would apply
to ingredients and meals at the college. Literature sources were
identied from our review of behavioural interventions and
selected on the basis of similarities in context (that is, food in
a higher education setting)68–72. These estimates are therefore
only rough, order-of-magnitude indications used to inform
policy recommendations; trials and monitoring at the college
would be needed to assess the actual changes that these types
of intervention might achieve.
Feasibility assessment—stakeholder consultations. The feasibil-
ity assessment for interventions was completed using information
from interviews and follow-up discussions with three key food and

Nature Food | Volume 4 | January 2023 | 96–108 105
Analysis https://doi.org/10.1038/s43016-022-00660-2
operations staff members at the college (head chef, catering manager
and domestic bursar). Semi-structured interviews were carried out
in early 2020 as part of the Wellcome Trust -funded ‘Our Planet Our
Health’ (Livestock, Environment and People—LEAP) project. Interviews
were carried out with informed consent of participants and approval
from the University of Oxford’s Central University Research Ethics
Committee (reference number R68035 /RE002).
The interviews followed a set of questions that aimed to (1) under-
stand how the college’s food service operations work, (2) identify
facilitators and barriers to implementing environmental initiatives to
promote sustainable meals and (3) understand lessons learned from
environmental initiatives that have been tried previously in the college.
Interviewees were questioned on whether sustainability measures had
been discussed and implemented at the college, both generally as well
as measures relating specifically to food. They were also asked whether
responses to measures had been positive or negative for different
stakeholders, and the perceived reasons for those responses. Some
questions were specifically focused on approaches to meat consump-
tion at the college. Additional information on specific aspects of the
college’s operation, core values and food provision was gained through
further ad-hoc discussions.
Interventions were assessed against qualitative information in
notes from these interviews, to identify key relevant points relating
to social, logistical and financial risks or opportunities (as captured
in Table 2). Information regarding specific consumer groups was used
to categorize the level of social risk (in terms of poor, moderate or
adequate perceptions of an intervention; Supplementary Table 2),
with additional reference to responses when similar interventions have
been trialled in the past, or predictions based on stakeholder familiar-
ity with consumer requirements and values. As such, the categoriza-
tions used here are based on well-informed assumptions about likely
stakeholder responses. This was a pragmatic approach to prioritizing
interventions to aid decision-making processes; the relevant expertize
of stakeholders makes this a good basis on which to begin prioritizing
strategies and interventions.
Stage 3 limitations. As it was not within scope to conduct an in-depth
review and comprehensive analysis of interventions, it is important to
acknowledge the limitations of our approach:
Technical assessments for ‘top-down’ interventions involved tak-
ing an average value for plant-based meat alternatives, rather than
considering for example, direct vegetable substitutes. Further, the
approach to modelling sustainable sourcing assumed a switch from
50th percentile producers to producers with the minimum level of
impact (5th percentile). This therefore does not consider the actual
availability of such low-impact products to the college.
For bottom-up interventions, limiting the assessment to
behaviour-change studies in similar university settings meant that few
studies were used to inform estimates of technical potential, therefore
limiting consideration of behavioural plasticity (that is, the extent to
which the interventions actually change consumer behaviour
46
), which
may influence the effectiveness of interventions. Moreover, the risks
and benefits to consumers assessed as part of our socio-economic fea-
sibility assessment arose through interviews with college stakeholders,
and while the importance of these expertise-driven insights should not
be underestimated, some of the concerns raised were hypothetical, and
may not reflect actual behaviour should the interventions be enacted.
The latter two limitations emphasise the importance of (1) run-
ning surveys or focus groups with affected consumers to more directly
understand consumer perceptions, (2) trialling proposed initiatives
and gathering data to estimate behavioural plasticity, (3) applying spe-
cific behaviour-change frameworks for a comprehensive analysis (for
example, refs.
48–50
) and (4) monitoring changes in consumer behaviour
or perceptions and adapting the approach accordingly. We recommend
consideration of these points in future applications of our approach,
although the methods used here provide a solid, pragmatic basis for
prioritizing strategies for initial implementation.
Stage 4: modelling mitigation strategies
Strategy A. Strategy A assumes that the college serves no
animal-derived food products and sources all ingredients from pro-
ducers with the lowest impacts on biodiversity. Percentage reductions
in impact were calculated following a similar approach to the ‘Healthy
and Sustainable Diets’ target (EL2035): baseline food quantities were
disaggregated into individual ingredients and relative proportions
were altered to nutritionally balanced vegan dietary proportions
(obtained from Supplementary Table 7 in Springmann et al.31). Newly
proportioned food impacts based on average (50th percentile) values
were then replaced with low-impact (5th percentile) values to account
for environmentally sustainable sourcing.
Combining interventions in strategies B–E. Supplementary Table
3 lists the specific interventions included in strategies B–E. Interven-
tions were applied sequentially to avoid double counting of mitiga-
tion potential (for example, if a strategy contained more than one
intervention pertaining to the same category of food). For strategy E,
an assumption was made that the impacts of ‘bottom-up’ behavioural
interventions would be additive, that is, we do not account for any inter-
actions that might occur when co-implementing several behavioural
interventions. As such, percentage reduction values for these behav-
ioural interventions are indicative only, although this was considered
adequate for the purposes of informing policy decisions at the college.
Ethics statement
Interviews were carried out with informed consent of participants and
approval from the University of Oxford’s Central University Research
Ethics Committee (CUREC reference number R68035/RE002).
Reporting summary
Further information on research design is available in the Nature Port-
folio Reporting Summary linked to this article.
Data availability
Data on environmental impacts per food ingredient34 (dataset 1
described in Methods) is publicly available via the Oxford University
Research Archive depository60. Data providing environmental values
per food product linked to foodDB
39
(see description of datasets 2 and 3
in Methods) is described by Clark et al. (2022)
61
with an anonymized ver-
sion of this dataset freely available via the Oxford University Research
Archive depository
62
. Owing to legal constraints, non-anonymized
data from the foodDB database is available under license upon request
(foodDBaccess@ndph.ox.ac.uk). Datasets on food product quantities
and anonymized interview responses used in this study are available
from the corresponding author on reasonable request. For legal con-
fidentiality reasons, financial data from the college cannot be made
publicly available. Source data are provided with this paper.
Code availability
Code relating to calculations of environmental values per food product
(as per Clark et al.
61
) is available on the Oxford University Research
Archive depository62.
References
1. Mace, G. M. et al. Aiming higher to bend the curve of biodiversity
loss. Nat. Sustain. 1, 448–451 (2018).
2. Díaz, S., et al. Pervasive human-driven decline of life on Earth
points to the need for transformative change. Science 366,
eaax3100 (2019).
3. Díaz, S. et al. Set ambitious goals for biodiversity and
sustainability. Science 370, 411 (2020).

Nature Food | Volume 4 | January 2023 | 96–108 106
Analysis https://doi.org/10.1038/s43016-022-00660-2
4. Locke, H., et al. A Nature-Positive World: The Global Goal for Nature
(Wildlife Conservation Society, 2020); https://library.wcs.org/doi/
ctl/view/mid/33065/pubid/DMX3974900000.aspx
5. Open-ended Working Group on the Post-2020 Global Biodiversity
Framework. First Draft of the Post-2020 Global Biodiversity
Framework CBD/WG2020/3/3 (Convention on Biological
Diversity, 2021).
6. Open-Ended Working Group on the Post-2020 Global
Biodiversity Framework. Draft Recommendation Submitted
by the Co-Chairs CBD/WG2020/4/L.2-ANNEX (Convention on
Biological Diversity, 2022).
7. Environment Act 2021 (UK) (HM Government, 2021); https://www.
legislation.gov.uk/ukpga/2021/30/contents/enacted
8. Bull, J. W. & Strange, N. The global extent of biodiversity oset
implementation under no net loss policies. Nat. Sustain. 1,
790–798 (2018).
9. Prendeville, S., Cherim, E. & Bocken, N. Circular cities:
mapping six cities in transition. Environ. Innov. Soc. Transit. 26,
171–194 (2018).
10. de Silva, G. C., Regan, E. C., Pollard, E. H. B. & Addison,
P. F. E. The evolution of corporate no net loss and net positive
impact biodiversity commitments: understanding appetite
and addressing challenges. Bus. Strategy Environ. 28,
1481–1495 (2019).
11. zu Ermgassen, S. O. S. E. et al. Exploring the ecological outcomes
of mandatory biodiversity net gain using evidence from early‐
adopter jurisdictions in England. Conserv. Lett. 14, e12820 (2021).
12. McGlyn, J., et al. Science-Based Targets for Nature: Initial Guidance
for Business (Science Based Targets Network, 2020); https://
sciencebasedtargetsnetwork.org/resource-repository/
13. zu Ermgassen, S. O. S. E. et al. Are corporate biodiversity
commitments consistent with delivering ‘nature-positive’
outcomes? A review of ‘nature-positive’ deinitions, company
progress and challenges. J. Clean. Prod. 379, 134798 (2022).
14. Addison, P. F. E., Bull, J. W. & Milner‐Gulland, E. J. Using
conservation science to advance corporate biodiversity
accountability. Conserv. Biol. 33, 307–318 (2019).
15. Smith, T. et al. Biodiversity means business: reframing global
biodiversity goals for the private sector. Conserv. Lett. 13,
e12690 (2020).
16. Maron, M. et al. Setting robust biodiversity goals. Conserv. Lett.
https://doi.org/10.1111/conl.12816 (2021).
17. Newing, H. & Perram, A. What do you know about conservation
and human rights? Oryx 53, 595–596 (2019).
18. Standard on Biodiversity Osets (The Business and Biodiversity
Osets Programme, 2012).
19. Arlidge, W. N. S., et al. A mitigation hierarchy approach for
managing sea turtle captures in small-scale isheries. Front. Mar.
Sci. 7, 49 (2020).
20. Squires, D. & Garcia, S. The least-cost biodiversity impact
mitigation hierarchy with a focus on marine isheries and bycatch
issues. Conserv. Biol. 32, 989–997 (2018).
21. Booth, H., Squires, D. & Milner-Gulland, E. J. The mitigation
hierarchy for sharks: a risk-based framework for reconciling
trade-os between shark conservation and isheries objectives.
Fish Fish. 21, 269–289 (2020).
22. Gupta, T. et al. Mitigation of elasmobranch bycatch in
trawlers: a case study in Indian isheries. Front. Mari. Sci. 7,
571 (2020).
23. Budiharta, S. et al. Restoration to oset the impacts of
developments at a landscape scale reveals opportunities,
challenges and tough choices. Global Environ. Change 52,
152–161 (2018).
24. Bull, J. W. et al. Net positive outcomes for nature. Nat. Ecol. Evol.
4, 4–7 (2020).
25. Arlidge, W. N. S. et al. A global mitigation hierarchy for nature
conservation. BioScience 68, 336–347 (2018).
26. Milner-Gulland, E. J. et al. Four steps for the Earth: mainstreaming
the post-2020 global biodiversity framework. One Earth 4,
75–87 (2021).
27. Wol, A., Gondran, N. & Brodhag, C. Detecting unsustainable
pressures exerted on biodiversity by a company. Application to
the food portfolio of a retailer. J. Clean. Prod. 166, 784–797 (2017).
28. FAOSTAT Analytical Brief 15 Land Use and Land Cover Statistics:
Global, Regional and Country Trends, 1990–2018 (FAO, 2020).
29. Williams, D. R. et al. Proactive conservation to prevent habitat
losses to agricultural expansion. Nat. Sustain. 4, 314–322 (2021).
30. Leclère, D. et al. Bending the curve of terrestrial biodiversity
needs an integrated strategy. Nature 585, 551–556 (2020).
31. Springmann, M. et al. Health and nutritional aspects of
sustainable diet strategies and their association with
environmental impacts: a global modelling analysis with
country-level detail. Lancet Planet. Health 2, e451–e461 (2018).
32. Clark, M. A., Springmann, M., Hill, J. & Tilman, D. Multiple health
and environmental impacts of foods. Proc. Natl Acad. Sci. USA
116, 23357 (2019).
33. Willett, W. et al. Food in the Anthropocene: the EAT–Lancet
Commission on healthy diets from sustainable food systems.
Lancet 393, 447–492 (2019).
34. Poore, J. & Nemecek, T. Reducing food’s environmental
impacts through producers and consumers. Science 360,
987 (2018).
35. Wiedmann, T., Lenzen, M., Keyßer, L. T. & Steinberger, J. K.
Scientists’ warning on aluence. Nat. Commun. 11, 3107 (2020).
36. Benton, T. G. et al. A ‘net zero’ equivalent target is needed to
transform food systems. Nat. Food 2, 905–906 (2021). 2021.
37. Crenna, E., Sinkko, T. & Sala, S. Biodiversity impacts due to food
consumption in Europe. J. Clean. Prod. 227, 378–391 (2019).
38. Bull, J. W., et al. Analysis: the biodiversity footprint of the
University of Oxford. Nature 604, 420–424 (2022).
39. Harrington, R. A., Adhikari, V., Rayner, M. & Scarborough, P.
Nutrient composition databases in the age of big data: foodDB, a
comprehensive, real-time database infrastructure. BMJ Open 9,
e026652 (2019).
40. Chaudhary, A., Verones, F., De Baan, L. & Hellweg, S. Quantifying
land use impacts on biodiversity: combining species–area
models and vulnerability indicators. Environ. Sci. Technol. 49,
9987–9995 (2015).
41. Winter, L., Lehmann, A., Finogenova, N. & Finkbeiner, M. Including
biodiversity in life cycle assessment—state of the art, gaps and
research needs. Environ. Impact Assess. Rev. 67, 88–100 (2017).
42. Chaudhary, A. & Kastner, T. Land use biodiversity impacts
embodied in international food trade. Global Environ. Change 38,
195–204 (2016).
43. Lenzen, M. et al. International trade drives biodiversity threats in
developing nations. Nature 486, 109–112 (2012).
44. Bates, B., et al. National Diet and Nutrition Survey Years 1 to 9 of
the Rolling Programme (2008/2009–2016/2017): Time Trend
and Income Analyses (Public Health England & Food Standards
Agency, 2019).
45. Stewart, C., Piernas, C., Cook, B. & Jebb, S. A. Trends in UK meat
consumption: analysis of data from years 1–11 (2008–09 to 2018–
19) of the National Diet and Nutrition Survey rolling programme.
Lancet Planet. Health 5, e699–e708 (2021).
46. Nielsen, K. S. et al. Improving climate change mitigation
analysis: a framework for examining feasibility. One Earth 3,
325–336 (2020).
47. Selinske, M. J. et al. We have a steak in it: eliciting interventions to
reduce beef consumption and its impact on biodiversity. Conserv.
Lett. 13, e12721 (2020).

Nature Food | Volume 4 | January 2023 | 96–108 107
Analysis https://doi.org/10.1038/s43016-022-00660-2
48. Hollands, G. J. et al. The TIPPME intervention typology for
changing environments to change behaviour. Nat. Hum. Behav. 1,
1–9 (2017).
49. Marteau, T. M., Hollands, G. J. & Fletcher, P. C. Changing human
behavior to prevent disease: the importance of targeting
automatic processes. Science 337, 1492–1495 (2012).
50. Michie, S., van Stralen, M. M. & West, R. The behaviour change
wheel: a new method for characterising and designing behaviour
change interventions. Implement. Sci. 6, 42 (2011).
51. Moran, D., Giljum, S., Kanemoto, K. & Godar, J. From satellite to
supply chain: new approaches connect earth observation to
economic decisions. One Earth 3, 5–8 (2020).
52. Godar, J., Suavet, C., Gardner, T. A., Dawkins, E. & Meyfroidt, P.
Balancing detail and scale in assessing transparency to improve
the governance of agricultural commodity supply chains. Environ.
Res. Lett. 11, 035015 (2016).
53. DeFries, R. S., Fanzo, J., Mondal, P., Remans, R. & Wood, S. A.
Is voluntary certiication of tropical agricultural commodities
achieving sustainability goals for small-scale producers? A review
of the evidence. Environ. Res. Lett. 12, 033001 (2017).
54. Bull, J. W., Suttle, K. B., Gordon, A., Singh, N. J. & Milner-
Gulland, E. J. Biodiversity osets in theory and practice. Oryx 47,
369–380 (2013).
55. zu Ermgassen, S. O. S. E. et al. The ecological outcomes of
biodiversity osets under “no net loss” policies: a global review.
Conserv. Lett. 12, e12664 (2019).
56. Waddock, S. Achieving sustainability requires systemic business
transformation. Glob. Sustain. 3, e12 (2020).
57. Travers, H., Walsh, J., Vogt, S., Clements, T. & Milner-Gulland, E.
J. Delivering behavioural change at scale: what conservation can
learn from other ields. Biol. Conserv. 257, 109092 (2021).
58. Gaupp, F. et al. Food system development pathways for
healthy, nature-positive and inclusive food systems. Nat. Food 2,
928–934 (2021).
59. Astill, J. et al. Transparency in food supply chains: a review of
enabling technology solutions. Trends Food Sci. Technol. 91,
240–247 (2019).
60. Poore, J & Nemecek, T. Full Excel model: life-cycle environmental
impacts of food drink products. Oxford University Research
Archive https://ora.ox.ac.uk/objects/uuid:a63b28c-98f8-
4313-add6-e9eca99320a5 (2018).
61. Clark, M., et al. Estimating the environmental impacts
of 57,000 food products. Proc. Natl Acad. Sci. USA 119,
e2120584119 (2022).
62. Clark, M., et al. Supplemental Data for ‘Estimating the
environmental impacts of 57,000 food products’. Oxford
University Research Archive https://ora.ox.ac.uk/objects/
uuid:4ad0b594-3e81-4e61-aefc-5d869c799a87 (2022).
63. Bianchi, F., Dorsel, C., Garnett, E., Aveyard, P. & Jebb, S. A.
Interventions targeting conscious determinants of human
behaviour to reduce the demand for meat: a systematic
review with qualitative comparative analysis. IJBNPA 15,
102 (2018).
64. Bianchi, F., Garnett, E., Dorsel, C., Aveyard, P. & Jebb, S.
A. Restructuring physical micro-environments to reduce
the demand for meat: a systematic review and qualitative
comparative analysis. Lancet Planet. Health 2, e384–e397 (2018).
65. Hillier-Brown, F. C. et al. The impact of interventions to promote
healthier ready-to-eat meals (to eat in, to take away or to be
delivered) sold by speciic food outlets open to the general
public: a systematic review. Obes. Rev. 18, 227–246 (2017).
66. von Philipsborn, P. et al. Environmental interventions to
reduce the consumption of sugar-sweetened beverages
and their eects on health. Cochrane Database Syst. Rev. 6,
Cd012292 (2019).
67. Attwood, S., Voorheis, P., Mercer, C., Davies, K. & Vennard, D.
Playbook for Guiding Diners toward Plant-Rich Dishes in Food
Service (World Resources Institute, 2020); https://www.wri.org/
research/playbook-guiding-diners-toward-plant-rich-dishes-
food-service
68. Garnett, E. E., Balmford, A., Sandbrook, C., Pilling, M. A. &
Marteau, T. M. Impact of increasing vegetarian availability on meal
selection and sales in cafeterias. Proc. Natl Acad. Sci. USA 116,
20923 (2019).
69. Reinders, M. J., Huitink, M., Dijkstra, S. C., Maaskant, A. J. &
Heijnen, J. Menu-engineering in restaurants—adapting portion
sizes on plates to enhance vegetable consumption: a real-life
experiment. IJBNPA 14, 41 (2017).
70. Brunner, F., Kurz, V., Bryngelsson, D. & Hedenus, F. Carbon label
at a university restaurant—label implementation and evaluation.
Ecol. Econ. 146, 658–667 (2018).
71. McClain, A. D., Hekler, E. B. & Gardner, C. D. Incorporating
prototyping and iteration into intervention development: a case
study of a dining hall-based intervention. J. Am. Coll. Health 61,
122–131 (2013).
72. de Vaan, J. Eating Less Meat: How to Stimulate the Choice for
a Vegetarian Option without Inducing Reactance. MSc thesis,
Radboud Univ. (2018).
Acknowledgements
The authors acknowledge and thank interview participants for their
time and their helpful insights. We also thank J. Poore and the team at
the foodDB project39, whose datasets underpin the analyses carried
out here. This manuscript arises from research funded by the John
Fell Oxford University Press Research Fund (funding to E.J.M.-G.,
H.M.J.G., I.T., J.B. and E.B.). M.C. and C.S. were funded by the Wellcome
Trust, Our Planet Our Health Programme (Livestock, Environment and
People—LEAP, award number: 205212/Z/16/Z). N.G. was funded by the
Crankstart Scholarship scheme.
Author contributions
All authors have provided content, reviewed, edited and approved
this manuscript. I.T. coordinated the project, including conducting
analyses and initial drafting of the manuscript. E.J.M.-G. supervised the
project, with co-supervision provided by J.W.B. and additional project
coordination provided by H.M.J.G. E.B. and N.G. provided support with
data processing and impacts analysis. M.C. contributed and supported
the use of datasets relating to the environmental impacts of food
products. C.S. conducted and recorded participant interviews. B.A.
represented and liaised with the focal college, providing underlying
datasets as well as key contextual information.
Competing interests
The authors declare no competing interests, but we note for
transparency that B.A. is an employee of Lady Margaret Hall (the focal
organization of this study).
Additional information
Supplementary information The online version contains supplementary
material available at https://doi.org/10.1038/s43016-022-00660-2.
Correspondence and requests for materials should be addressed
to I. Taylor or E. J. Milner-Gulland.
Peer review information Nature Food thanks Brent Loken, Kristian
Nielsen and Martine Maron for their contribution to the peer review
of this work.
Reprints and permissions information is available at
www.nature.com/reprints.

Nature Food | Volume 4 | January 2023 | 96–108 108
Analysis https://doi.org/10.1038/s43016-022-00660-2
Publisher’s note Springer Nature remains neutral with
regard to jurisdictional claims in published maps and
institutional ailiations.
Springer Nature or its licensor (e.g. a society or other partner) holds
exclusive rights to this article under a publishing agreement with
the author(s) or other rightsholder(s); author self-archiving of the
accepted manuscript version of this article is solely governed by the
terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Springer Nature Limited
2023

1
nature portfolio | reporting summary March 2021
Corresponding author(s): Isobel Taylor (isobel.taylor93@outlook.com)
Last updated by author(s): Oct 22, 2022
Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection No software used for data collection. Collection methods for data extracts used in this study are described in prior works (Poore & Nemecek,
2018, Science; Clark et al., 2022, PNAS).
Data analysis All data handling was conducted in Microsoft Excel. Analysis for data extracts used in this study are described in prior works (Poore &
Nemecek, 2018, Science; Clark et al., 2022, PNAS).
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy
Data on environmental impacts per food ingredient (dataset (1) described in Methods) is publicly available via the Oxford University Research Archive depository
(DOI: 10.5287/bodleian:0z9MYbMyZ).
Data providing environmental values per food product linked to foodDB (see description of datasets (2) and (3) in Methods) is described, with an anonymised
version of this dataset freely available, in Clark et al. (2022, PNAS). Due to legal constraints, non-anonymised data from the foodDB database is available under
license upon reasonable request (foodDBaccess@ndph.ox.ac.uk).

2
nature portfolio | reporting summary March 2021
Datasets on food product quantities and anonymised interview responses used in this study are available from the corresponding author on reasonable request. For
legal confidentiality reasons, financial data from the College cannot be made publicly available.
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description This study explored how Nature Positive targets can be achieved for organisations through a novel application of the Mitigation and
Conservation Hierarchy Framework, focusing on the embedded biodiversity impacts associated with food consumption. The study is
split into four stages
Stage 1: Annual impacts from food consumed at a focal organisation were quantified by pairing consumption data with
environmental databases containing life-cycle biodiversity impacts of specific food products.
Stage 2: A series of biodiversity targets were defined, including ones aligned with a nature-positive target, and annual impacts under
each target scenario were modeled in terms of changes to annual and cumulative biodiversity impacts.
Stage 3: The feasibility and technical potential of interventions to mitigate impacts were assessed based on Stage 1 quantifications
and on qualitative interview data gathered from key stakeholders at the focal organisation.
Stage 4: Five mitigation strategies were explored by combining sets of interventions, outlining the risks/feasibility for each strategy,
and quantifying potential progress towards different targets - highlighting the challenges around achieving Nature Positive targets
(e.g., Biodiversity Net Gain) at an organisational level within the current food system.
Research sample Stage 1 consumption data consisted of 4,651 purchase records for individual products, including 1,612 unique food products. Each
product in the dataset included information on date, supplier, product code and description, number of units purchased, and cost
per unit. Data was gathered from three months (September 2018, February 2019, and July 2019)
Environmental datasets included biodiversity impact and carbon dioxide equivalent (CO2e) values for 55 raw food ingredients
provided by Poore & Nemecek (Science, 2018) and for extracts of 2,138 food products and 3,687 food 'shelves' listed in foodDB (data
provided by and described in Clark et al., 2022, PNAS). The foodDB extracts consisted of a specific subset of product categories
(including composite/'ready-made' meals as well as specific plant-based meat alternatives), aggregated at both product and
supermarket 'shelf' level. FoodDB data was provided under license from the authors (Harrington et al., 2019).
Stage 3 qualitative data (in the form of interview notes) were gathered from interviews from a sample of three targeted interviewees
at the focal organisation.
Sampling strategy Stage 1 consumption data was gathered for three representative months and then factored up to represent a year. Three months
were selected on the basis of consultation with stakeholders to ensure adequate representation of the different aspects of the
organisation's food operation while enabling a feasible level of manual data handling. These three months' of data were then
factored up to represent a year using a financial breakdown of organisational spending per operational area for the purposes of
establishing a single-year fixed baseline.
Stage 3 interviewees were selected using targeted/purposive sampling. Interviewees were limited to those meeting the required
criteria:
Inclusion criteria:
- Adults aged ≥18 years
- Able to speak and read English
- Having direct responsibility for management of the organisation's food operations
- Knowledge of the organisation's finances
- Having some interaction with the wider organisation's governance structures
Exclusion criteria: Unable or unwilling to provide consent for interview
Stage 3 interventions were collated from a non-systematic review of the literature. The search strategy involved identifying existing
systematic academic reviews of food consumption-related behaviour-change interventions, along with key sources of grey literature,
and expanding the search following a snowball sampling strategy (following references and citations) until no significantly new
additional intervention categories were able to be identified.
Data collection Stage 1 consumption data (food sales and purchase data) was provided directly by an authorised individual at the focal organisation
Stage 1 Environmental data was obtained directly from the authors of the relevant studies (Poore & Nemecek, 2018, Science;
Chaudhary et al., 2015, Environ. Sci. Technol; Clark et al., 2022, PNAS)
Stage 3 data was collected by the authors (C. Stewart) by conducting, recording, and transcribing interview notes.
Timing and spatial scale Consumption data covers three months over the 2018/19 academic year (September 2018, February 2019, and July 2019).
Consumption data relates to food prepared and consumed within the focal organisation.
Interviews were conducted in February 2020.
Data exclusions No data were excluded.
Reproducibility This study was exploratory, rather than experimental. Anonymised data/code for reproducibility of the impacts analysis can be

3
nature portfolio | reporting summary March 2021
Reproducibility provided (code is available from Clark et al., 2022, PNAS) however non-anonymised data extracted from foodDB cannot be provided
without license due to legal requirements.
Randomization Food consumption data was allocated to food products in environmental datasets based on the greatest degree of similarity between
products. Products were then allocated to food categories based on food group/composition. Randomisation was not relevant to this
study.
Blinding Interviews were recorded anonymously. Interview notes were applied in this study by a different individual to the interviewer, with
no way to link specific notes to the identity of single interviewees.
Did the study involve field work? Yes No
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
n/a Involved in the study
Antibodies
Eukaryotic cell lines
Palaeontology and archaeology
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
Methods
n/a Involved in the study
ChIP-seq
Flow cytometry
MRI-based neuroimaging
Human research participants
Policy information about studies involving human research participants
Population characteristics Three adult males (>18 years), able to speak and read English. All were employed by the focal organisation and had direct
responsibility for management of the organisation's food operations, knowledge of the organisation's finances and were
engaged with the wider organisation's governance structures.
Recruitment Participants were targeted through purposive sampling via email from a small number of individuals within the organisation
that fit the inclusion criteria (described in 'sampling strategy' above)
Ethics oversight University of Oxford’s Central University Research Ethics Committee (CUREC Reference number: R68035/RE002)
Note that full information on the approval of the study protocol must also be provided in the manuscript.