No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Probabilistic Modeling of a Sales Funnel to Prioritize Leads
Abstract
This paper shows how to learn probabilistic classifiers that model how sales prospects proceed through stages from first awareness to final success or failure. Specifically,we present two models, called DQM for direct qualification model and FFM for full funnel model, that can be used to rank initial leads based on their probability of conversion to a sales opportunity, probability of successful sale, and/or expected revenue. Training uses the large amount of historical data collected by customer relationship management or marketing automation software. The trained models can replace traditional lead scoring systems, which are hand-tuned and therefore error-prone and not probabilistic. DQM and FFM are designed to overcome the selection bias caused by available data being based on a traditional lead scoring system. Experimental results are shown on real sales data from two companies. Features in the training data include demographic and behavioral information about each lead. For both companies, both methods achieve high AUC scores. For one company, they result in a a 307% increase in number of successful sales, as well as a dramatic increase in total revenue. In addition, we describe the results of the DQM method in actual use. These results show that the method has additional benefits that include decreased time needed to qualify leads, and decreased number of calls placed to schedule a product demo. The proposed methods find high-quality leads earlier in the sales process because they focus on features that measure the fit of potential customers with the product being sold, in addition to their behavior.
... The total score is built by the (weighted) sum of scores. Categories and scores are "hand-tuned by experienced members of the marketing or sales team" [8] and comprise monetary and qualitative aspects as well as expected developments of the customer. Depending on the categories the scorecard method can also be applied to prospects. ...
... The stages and their names are often customized by sales teams and terms have different meanings. To give an example from our study, in [7] the stages "suspects, prospects, leads, customers" are used, in [8] the stages are "awareness, leads, marketingqualified leads, sales-qualified leads, won". However, the underlying logic is the same: If a customer or prospect reaches the next stage because the contact person responded to some marketing activities such as an e-mail, the probability to win a deal increases. ...
... The authors of [24,23] and [8] in row (5) predict both, the win propensity and the expected sales in case of a won deal. For predicting the win propensity the authors used logistic regression and a boosted tree classifier respectively. ...
A key task of sales representatives in operational planning is to select the most profitable customers to visit within the next few days. A strongly varying set of scoring methods predicting or approximating the expected response exists for this customer selection phase. However, in the case of field sales forces, the final customer selection is strongly interrelated with the tour planning decisions. To this end, we formalize variants of the profitable sales representatives tour problem as a multi-period team orienteering problem, thereby providing a unified view on the customer scoring and the tour planning phase. In an extensive computational study on real-world instances from the retail industry, we systematically examine the impact of the aggregation level and the content of information provided by a scoring method and the sensitivity of the proposed models concerning prediction errors. We show that the selection of a customer scoring and tour planning variant depends on the available data. Furthermore, we work out where to put in effort for the data acquisition and scoring phase to get better operational tours.
... The total score is built by the (weighted) sum of scores. Categories and scores are "hand-tuned by experienced members of the marketing or sales team" [8] and comprise monetary and qualitative aspects as well as expected developments of the customer. Depending on the categories the scorecard method can also be applied to prospects. ...
... The stages and their names are often customized by sales teams and terms have different meanings. To give an example from our study, in [7] the stages "suspects, prospects, leads, customers" are used, in [8] the stages are "awareness, leads, marketingqualified leads, sales-qualified leads, won". However, the underlying logic is the same: If a customer or prospect reaches the next stage because the contact person responded to some marketing activities such as an e-mail, the probability to win a deal increases. ...
... The authors of [24,23] and [8] in row (5) predict both, the win propensity and the expected sales in case of a won deal. For predicting the win propensity the authors used logistic regression and a boosted tree classifier respectively. ...
Field sales forces play an important role in direct marketing, especially for companies offering complex products, services, or solutions in the business-to-business context. A key task of sales representatives in operational planning is to select the most promising customers to visit within the next days. On an operative horizon, a key task for sales representatives is to select the most promising customers to visit within the next days. A strongly varying set of scoring methods predicting or approximating the expected response exists for this customer selection phase. However, in the case of field sales forces, the final customer selection is strongly interrelated to the tour planning decisions. To this end, we formalize variants of the profitable sales representatives tour problem as a multi-period team orienteering problem, thereby providing a unified view on the customer scoring and the tour planning phase. In an extensive computational study on real-world instances from the retail industry, we systematically examine the impact of the aggregation level and the content of information provided by a scoring method and the sensitivity of the proposed models concerning prediction errors. We show that the selection of a customer scoring and tour planning variant depends on the available data. Furthermore, we work out where to put effort in the data acquisition and scoring phase to get better operational tours.
... D'Haen et al. (2016) fed web crawling data into a Logistic Regression model, which generates a sales prospect list for selling energy retail plans. Duncan and Elkan (2015) used a three-class gradient boosted tree classifier to design two predictive lead scoring methods. Both methods were tested on sales data of two companies, and the experimental results highlighted the improvements in sales revenue when compared to a conventional rule-based approach (Duncan & Elkan, 2015). ...
... Duncan and Elkan (2015) used a three-class gradient boosted tree classifier to design two predictive lead scoring methods. Both methods were tested on sales data of two companies, and the experimental results highlighted the improvements in sales revenue when compared to a conventional rule-based approach (Duncan & Elkan, 2015). Benhaddou and Leray (2017) constructed a Bayesian network for lead scoring local companies based on news stories. ...
A fundamental component to managing a marketing campaign is identifying prospects and selection of leads. Current lead generation models focus on predicting the intention of a customer to purchase a product, however with financial products, particularly loans, this can be insufficient as there are many factors to consider, such as risk, utility, and financial maturity. Developing a marketing campaign for loan prospecting should consider not only customers who need a loan, but rather clients who need a loan and will also be approved. Otherwise, the marketing effort is deemed ineffective, if the lead cannot be converted into a sale. Although, a low response rate is expected for a marketing campaign for loans, we highlight a better approach in managing resources while maintaining a shortlist of high-quality leads. This manuscript introduces a bi-level approach to handle the complex nature of loan products. Two classifiers are built, one modelling loan intention and the other one modelling loan eligibility. We adopt convex combination to control weights of both problems, which in most cases results in an improved performance for identifying future successful loan applicants when compared to baseline models. Rank-based evaluation measures are also adopted to explore the performance of customer rankings. We find that soft classifiers, such as deep learning techniques, are ideal for ranking customers, achieving a superior performance when compared to other machine learning techniques. In addition, we conclude that an ideal cut-off for K customers is estimated to be between 20 to 25 customers, however our best model can maintain an Average Precision of greater than 0.85 when K approaches 50.
... Those predictions can be used to carry out an optimal amount of customer interactions, associated with the individual leads (Yan et al., 2015). Probabilistic classifiers are trained to rank initial leads based on their probability of conversion to a sales opportunity, probability of successful sale, and/or expected revenue (Duncan & Elkan, 2015). Data mining models, based on ANN are used to predict the behaviour of customers to enhance the decision-making processes for retaining valued customers (Bahari & Elayidom, 2015). ...
This paper considers Enterprise Information Systems functional architecture and carries out review of AI applications integrated in Customer Relationship Management, Supply Chain Management, Inventory and logistics, Production Planning and Scheduling, Finance and accounting, Product Lifecycle Management and Human Resources, with special attention to the manufacturing enterprises. Enhanced capabilities are identified and proposed as AI services. AI-enablement implements improved decision-making or automation by using Machine Learning models or logic-based systems. It is a process of the enterprise transformation leading to the convergence of the four major disruptive technologies, namely Industrial Internet of Things, Agent-based Distributed Systems, Cloud Computing and Artificial Intelligence.
... Yan et al. propose a win-propensity approach based on modeling the interaction of users with the sales support system as Hawkes Processes [9]. Duncan and Elkan propose a pure probabilistic model [48]. Compared with these approaches, our approach is theoretically either superior in terms of predictive or explanatory power, and always superior in balancing predictive and explanatory power. ...
... Yan et al. propose a win-propensity approach based on modeling the interaction of users with the sales support system as Hawkes Processes [9]. Duncan and Elkan propose a pure probabilistic model [48]. Compared with these approaches, our approach is theoretically either superior in terms of predictive or explanatory power, and always superior in balancing predictive and explanatory power. ...
MAN Energy Solutions, one of the largest ship engine manufacturers in the world, is looking into further improving its hit rate of through-life engineering services and spare parts quotations. We help to solve this relevant field problem by building a novel machine learning based sales win-propensity prediction system that utilizes the lightGBM algorithm, SHapley Additive exPlanations, and a second layer conditional probability model of quotation age. Moreover, we build an implementation method for the broader class of such systems and extend the scientific literature on explainable machine learning by abductively developing and instantiating the design principles (DPs) of local contrastive explainability, global explainability, selective visualization, causality, confirmatory nudging, and accountability in a sales win-propensity system.
... This means that the conversion to the next purchase funnel steps cannot be appreciated. Databases comparison on a digital footprint can only show averaged statistics, and the classical marketing research methods at the last stage of the purchase funnel give a very serious distortion [10,11,12,13]. ...
... In the context of professional social networks there has been work on identifying key decision makers within an organization (via declared user profiles) for the purpose of B2B ad targeting [24]. In the B2B marketing industry, identifying such leads [9] and key decision makers in an organization is widely seen as an effective approach for B2B marketing [24]. Compared to prior work in professional social networks (where users explicitly declare their roles in an organization), we do not focus on identifying generic decision makers in an organization. ...
Online purchase decisions in organizations can go through a complex journey with multiple agents involved in the decision making process. Depending on the product being purchased, and the organizational structure, the process may involve employees who first conduct market research, and then influence decision makers who place the online purchase order. In such cases, the online activity trail of a single individual in the organization may only provide partial information for predicting purchases (conversions). To refine conversion prediction for business-to-business (B2B) products using online activity trails, we introduce the notion of relevant users in an organization with respect to a given B2B advertiser, and leverage the collective activity trails of such relevant users to predict conversions. In particular, our notion of relevant users is tied to a seed list of relevant activities for a B2B advertiser, and we propose a method using distributed activity representations to build such a seed list. Experiments using data from Yahoo Gemini demonstrate that the proposed methods can improve conversion prediction AUC by 8.8%, and provide an interpretable advertiser specific list of activities useful for B2B ad targeting.
... Thus, probabilistic optimization methods have been used by several researchers for various cases. In the field of business, probabilistic optimization models can be used in cases such as service identification, process standardization [2], and sales modeling [3]. These methods can be useful in many decisionmaking cases in business as they are built upon uncertainty [4]. ...
This paper deals with a probabilistic problem in which there is a specific probability for a customer to meet a seller in a specified area. It is assumed that the area in which a seller acts follows an exponential distribution and affects the probability of meeting with a customer. Furthermore, the range in which a customer can meet a seller is another parameter which affects the probability of a successful meeting. The solution to the problem is based on a bomb fragmentation model using Lagrange equations. More specifically, using Lagrange equations, the abovementioned dimensions will be calculated in order to optimize the probability of a customer meeting a seller.
Although lead scoring is an essential component of lead management, there is a lack of a comprehensive literature review and a classification framework dedicated to it. Lead scoring is an effective and efficient way of measuring the quality of leads. In addition, as a critical Information Technology tool, a proper lead scoring model acts as an alleviator to weaken the conflicts between sales and marketing functions. Yet, little is known regarding lead scoring models and their impact on sales performance. Lead scoring models are commonly categorized into two classes: traditional and predictive. While the former primarily relies on the experience and knowledge of salespeople and marketers, the latter utilizes data mining models and machine learning algorithms to support the scoring process. This study aims to review and analyze the existing literature on lead scoring models and their impact on sales performance. A systematic literature review was conducted to examine lead scoring models. A total of 44 studies have met the criteria and were included for analysis. Fourteen metrics were identified to measure the impact of lead scoring models on sales performance. With the increased use of data mining and machine learning techniques in the fourth industrial revolution, predictive lead scoring models are expected to replace traditional lead scoring models as they positively impact sales performance. Despite the relative cost of implementing and maintaining predictive lead scoring models, it is still beneficial to supersede traditional lead scoring models, given the higher effectiveness and efficiency of predictive lead scoring models. This study reveals that classification is the most popular data mining model, while decision tree and logistic regression are the most applied algorithms among all the predictive lead scoring models. This study contributes by systematizing and recommending which machine learning method (i.e., supervised and/or unsupervised) shall be used to build predictive lead scoring models based on the integrity of different types of data sources. Additionally, this study offers both theoretical and practical research directions in the lead scoring field.
Getting visitors to register is a crucial factor in marketing for online news portals. Current approaches are rule-based by awarding points for specific actions [3]. Finding efficient rules can be challenging and depends on the specific task. Registration is generally rare compared to regular visitors, leading to highly imbalanced data. We analyze different supervised learning classification algorithms under consideration of the data imbalance. As case study, we use anonymized real-world data from an Austrian newspaper outlet containing the visitor’s session behavior with around 0.1% registrations over all visits. We identify an ensemble approach combining the Balanced Random Forest Classifier and the RUSBoost Classifier correctly identifying 76% of registrations over five independent data sets.
In this study, the impact of customer agility and digitalisation in restructuring an automotive company and its daily operations was explored. Drawing on agility literature, we provide an empirical case study with the purpose of exploring how agility and digital transformation interconnect with one another. Our research confirms the importance of a digital mindset, flexible and agile company structure, and extensive digital skills for digital transformation. Furthermore, modern automotive companies must leverage strong IT infrastructures to maintain massive data gathered daily, both online and offline.Theoretically, our research highlights the interplay between digitalisation and customer agility and contributes to said literature stream, as it supports existing theories on organisational traits that favour agility and digitalisation. Specifically, this study extensively reveals how IT systems allow companies to interact with their environments quickly, manage real-time data streams and adapt accordingly. In practice, this study can serve as the best example for managers and practitioners, especially considering the context in which it occurred. In addition, the automotive industry is, until now, one of the most active, competitive and ever-changing sectors. Finally, this study's limitation lies within its scope; its implications may not be generalised since it focuses on the automotive industry.
In this paper, we present a method to use advance demand information (ADI), taking the form of request for quotation (RFQ) data, in B2B sales forecasting. We apply supervised machine learning and Natural Language Processing techniques to analyze and learn from RFQs. We apply and test our approach in a case study at a large after-sales service and maintenance provider. After evaluation we found that our approach identifies ∼70% of actual sales (recall) with a precision rate of ∼50%, which represents a performance improvement of slightly more than a factor 2.5 over the current labor-intensive manual process at the service and maintenance provider. Our research contributes to literature by giving step-by-step guidance on incorporating artificial intelligence in B2B sales forecasting and revealing potential pitfalls along the way. Furthermore, our research gives an indication of the performance improvement that can be expected when adopting supervised machine learning into B2B sales forecasting.
Sales professionals help organizations win clients for products and services. Generating new clients starts with identifying the right decision makers at the target organization. For the past decade, online professional networks have collected tremendous amount of data on people's identity, their network and behavior data of buyers and sellers building relationships with each other for a variety of use-cases. Sales professionals are increasingly relying on these networks to research, identify and reach out to potential prospects, but it is often hard to find the right people effectively and efficiently. In this paper we present LDMS, the LinkedIn Decision Maker Score, to quantify the ability of making a sales decision for each of the 400M+ LinkedIn members. It is the key data-driven technology underlying Sales Navigator, a proprietary LinkedIn product that is designed for sales professionals. We will specifically discuss the modeling challenges of LDMS, and present two graph-based approaches to tackle this problem by leveraging the professional network data at LinkedIn. Both approaches are able to leverage both the graph information and the contextual information on the vertices, deal with small amount of labels on the graph, and handle heterogeneous graphs among different types of vertices. We will show some offline evaluations of LDMS on historical data, and also discuss its online usage in multiple applications in live production systems as well as future use cases within the LinkedIn ecosystem.
Gradient boosting constructs additive regression models by sequentially fitting a simple parameterized function (base learner) to current “pseudo”-residuals by least squares at each iteration. The pseudo-residuals are the gradient of the loss functional being minimized, with respect to the model values at each training data point evaluated at the current step. It is shown that both the approximation accuracy and execution speed of gradient boosting can be substantially improved by incorporating randomization into the procedure. Specifically, at each iteration a subsample of the training data is drawn at random (without replacement) from the full training data set. This randomly selected subsample is then used in place of the full sample to fit the base learner and compute the model update for the current iteration. This randomized approach also increases robustness against overcapacity of the base learner.
The input to an algorithm that learns a binary classifier normally consists of two sets of examples, where one set consists of positive examples of the concept to be learned, and the other set consists of negative examples. However, it is often the case that the available training data are an incomplete set of positive examples, and a set of unlabeled examples, some of which are positive and some of which are negative. The problem solved in this paper is how to learn a standard binary classifier given a nontraditional training set of this nature.
Under the assumption that the labeled examples are selected randomly from the positive examples, we show that a classifier trained on positive and unlabeled examples predicts probabilities that differ by only a constant factor from the true conditional probabilities of being positive. We show how to use this result in two different ways to learn a classifier from a nontraditional training set. We then apply these two new methods to solve a real-world problem: identifying protein records that should be included in an incomplete specialized molecular biology database. Our experiments in this domain show that models trained using the new methods perform better than the current state-of-the-art biased SVM method for learning from positive and unlabeled examples.
This paper presents approaches to semi-supervised learning when the labeled training data and test data are differently distributed. Specifically, the samples selected for labeling are a biased subset of some general distribution and the test set consists of samples drawn from either that general distribution or the distribution of the unlabeled samples. An example of the former appears in loan application approval, where samples with repay/default labels exist only for approved applicants and the goal is to model the repay/default behavior of all applicants. An example of the latter appears in spam filtering, in which the labeled samples can be out-dated due to the cost of labeling email by hand, but an unlabeled set of up-to-date emails exists and the goal is to build a filter to sort new incoming email.Most approaches to overcoming such bias in the literature rely on the assumption that samples are selected for labeling depending only on the features, not the labels, a case in which provably correct methods exist. The missing labels are said to be "missing at random" (MAR). In real applications, however, the selection bias can be more severe. When the MAR conditional independence assumption is not satisfied and missing labels are said to be "missing not at random" (MNAR), and no learning method is provably always correct.We present a generative classifier, the shifted mixture model (SMM), with separate representations of the distributions of the labeled samples and the unlabeled samples. The SMM makes no conditional independence assumptions and can model distributions of semi-labeled data sets with arbitrary bias in the labeling. We present a learning method based on the expectation maximization (EM) algorithm that, while not always able to overcome arbitrary labeling bias, learns SMMs with higher test-set accuracy in real-world data sets (with MNAR bias) than existing learning methods that are proven to overcome MAR bias.
Machine learning methods are powerful tools for data mining with large noisy databases and give researchers the opportunity to gain new insights into consumer behavior and to improve the performance of marketing operations. To model consumer responses to direct marketing, this study proposes Bayesian networks learned by evolutionary programming. Using a large direct marketing data set, we tested the endogeneity bias in the recency, frequency, monetary value (RFM) variables using the control function approach; compared the results of Bayesian networks with those of neural networks, classification and regression tree (CART), and latent class regression; and applied a tenfold cross-validation. The results suggest that Bayesian networks have distinct advantages over the other methods in accuracy of prediction, transparency of procedures, interpretability of results, and explanatory insight. Our findings lend strong support to Bayesian networks as a robust tool for modeling consumer response and other marketing problems and for assisting management decision making.
Scikit-learn is a Python module integrating a wide range of state-of-the-art
machine learning algorithms for medium-scale supervised and unsupervised
problems. This package focuses on bringing machine learning to non-specialists
using a general-purpose high-level language. Emphasis is put on ease of use,
performance, documentation, and API consistency. It has minimal dependencies
and is distributed under the simplified BSD license, encouraging its use in
both academic and commercial settings. Source code, binaries, and documentation
can be downloaded from http://scikit-learn.sourceforge.net.
Accurate, well-calibrated estimates of class membership probabilities are needed in many supervised learning applications, in particular when a cost-sensitive decision must be made about examples with example-dependent costs. This paper presents simple but successful methods for obtaining calibrated probability estimates from decision tree and naive Bayesian classifiers. Using the large and challenging KDD'98 contest dataset as a testbed, we report the results of a detailed experimental comparison of ten methods, according to four evaluation measures. We conclude that binning succeeds in significantly improving naive Bayesian probability estimates, while for improving decision tree probability estimates, we recommend smoothing by -estimation and a new variant of pruning that we call curtailment.
Function estimation/approximation is viewed from the perspective of numerical optimization iti function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest-descent minimization. A general gradient descent "boosting" paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are regression trees, and tools for interpreting such "TreeBoost" models are presented. Gradient boosting of regression trees produces competitives highly robust, interpretable procedures for both regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire and Friedman, Hastie and Tibshirani are discussed.
Deemed "one of the top ten data mining mistakes", leakage is essentially the introduction of information about the data mining target, which should not be legitimately available to mine from. In addition to our own industry experience with real-life projects, controversies around several major public data mining competitions held recently such as the INFORMS 2010 Data Mining Challenge and the IJCNN 2011 Social Network Challenge are evidence that this issue is as relevant today as it has ever been. While acknowledging the importance and prevalence of leakage in both synthetic competitions and real-life data mining projects, existing literature has largely left this idea unexplored. What little has been said turns out not to be broad enough to cover more complex cases of leakage, such as those where the classical i.i.d. assumption is violated, that have been recently documented. In our new approach, these cases and others are explained by explicitly defining modeling goals and analyzing the broader framework of the data mining problem. The resulting definition enables us to derive general methodology for dealing with the issue. We show that it is possible to avoid leakage with a simple specific approach to data management followed by what we call a learn-predict separation, and present several ways of detecting leakage when the modeler has no control over how the data have been collected.
The leading introductory book on data mining, fully updated and revised!When Berry and Linoff wrote the first edition of Data Mining Techniques in the late 1990s, data mining was just starting to move out of the lab and into the office and has since grown to become an indispensable tool of modern business. This new edition—more than 50% new and revised— is a significant update from the previous one, and shows you how to harness the newest data mining methods and techniques to solve common business problems. The duo of unparalleled authors share invaluable advice for improving response rates to direct marketing campaigns, identifying new customer segments, and estimating credit risk. In addition, they cover more advanced topics such as preparing data for analysis and creating the necessary infrastructure for data mining at your company. Features significant updates since the previous edition and updates you on best practices for using data mining methods and techniques for solving common business problemsCovers a new data mining technique in every chapter along with clear, concise explanations on how to apply each technique immediatelyTouches on core data mining techniques, including decision trees, neural networks, collaborative filtering, association rules, link analysis, survival analysis, and moreProvides best practices for performing data mining using simple tools such as ExcelData Mining Techniques, Third Edition covers a new data mining technique with each successive chapter and then demonstrates how you can apply that technique for improved marketing, sales, and customer support to get immediate results.
Due to the proliferation of information systems and technology, businesses increasingly have the capability to accumulate huge amounts of customer data in large databases. However, much of the useful marketing insights into customer characteristics and their purchase patterns are largely hidden and untapped. Current emphasis on customer relationship management makes the marketing function an ideal application area to greatly benefit from the use of data mining tools for decision support. A systematic methodology that uses data mining and knowledge management techniques is proposed to manage the marketing knowledge and support marketing decisions. This methodology can be the basis for enhancing customer relationship management. q 2001 Elsevier Science B.V. All rights reserved.
The objective of this paper is to inform the information systems (IS) manager and business analyst about the role of machine learning techniques in business data mining. Data mining is a fast growing application area in business. Machine learning techniques are used for data analysis and pattern discovery and thus can play a key role in the development of data mining applications. Understanding the strengths and weaknesses of these techniques in the context of business is useful in selecting an appropriate method for a specific application. The paper, therefore, provides an overview of machine learning techniques and discusses their strengths and weaknesses in the context of mining business data. A survey of data mining applications in business is provided to investigate the use of learning techniques. Rule induction (RI) was found to be most popular, followed by neural networks (NNs) and case-based reasoning (CBR). Most applications were found in financial areas, where prediction of the future was a dominant task category.
Sales departments tend to believe that marketers are out of touch with what's really going on in the marketplace. Marketing people, in turn, believe the sales force is myopic--too focused on individual customer experiences, insufficiently aware of the larger market, and blind to the future. In short, each group undervalues the other's contributions. Both stumble (and organizational performance suffers) when they are out of sync. Yet few firms seem to make serious overtures toward analyzing and enhancing the relationship between these two critical functions. Curious about the misalignment between Sales and Marketing, the authors interviewed pairs of chief marketing officers and sales vice presidents to capture their perspectives. They looked in depth at the relationship between Sales and Marketing in a variety of companies in different industries. Their goal was to identify best practices that could enhance the joint performance and increase the contributions of these two functions. Among their findings: The marketing function takes different forms in different companies at different product life cycle stages. Marketing's increasing influence in each phase of an organization's growth profoundly affects its relationship with Sales. The strains between Sales and Marketing fall into two main categories: economic (a single budget is typically divided, between Sales and Marketing, and not always evenly) and cultural (the two functions attract very different types of people who achieve success by spending their time in very different ways). In this article, the authors describe the four types of relationships Sales and Marketing typically exhibit. They provide a diagnostic to help readers assess their companies' level of integration, and they offer recommendations for more closely aligning the two functions.
Function approximation is viewed from the perspective of numerical optimization in function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest--descent minimization. A general gradient--descent "boosting" paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least--squares, least--absolute--deviation, and Huber--M loss functions for regression, and multi--class logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are decision trees, and tools for interpreting such "TreeBoost" models are presented. Gradient boosting of decision trees produces competitive, highly robust, interpretable procedures for regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire 1996, and Fr...
Knowledge management and data mining for marketing, Decision Support Systems, v.31 n.1
- J Michael
- Chandrasekar Shaw
- Gek Subramaniam
- Michael E Woo Tan
- Welge
Leakage in data mining: Formulation, detection, and avoidance
- Saharon Shachar Kaufman
- Claudia Rosset
- Ori Perlich
- Stitelman