Editing Market segmentation (section)

== Segmentation: algorithms and approaches ==

The choice of an appropriate statistical method for the segmentation depends on numerous factors that may include, the broad approach ([[A priori knowledge|a-priori]] or [[Post hoc analysis|post-hoc]]), the availability of data, time constraints, the marketer's skill level, and resources.<ref>Myers, J.H., ''Segmentation and Positioning for Strategic Marketing Decisions'', American Marketing Association, 1996</ref>

=== A-priori segmentation ===

A priori research occurs when "a theoretical framework is developed before the research is conducted".<ref>Market Research Association, ''Glossary of Terms'', Online:http://www.marketingresearch.org/issues-policies/glossary</ref> In other words, the marketer has an idea about whether to segment the market geographically, demographically, psychographically or behaviourally before undertaking any research. For example, a marketer might want to learn more about the motivations and demographics of light and moderate users to understand what tactics could be used to increase usage rates. In this case, the target variable is known – the marketer has already segmented using a behavioural variable – '''user status'''. The next step would be to collect and analyze attitudinal data for light and moderate users. The typical analysis includes simple cross-tabulations, frequency distributions, and occasionally logistic regression or one of several proprietary methods.<ref>Wedel, M. and Kamakura, W.A., ''Market Segmentation: Conceptual and Methodological Foundations,'' Springer Science & Business Media, 2010, pp 22-23.</ref>

The main disadvantage of a-priori segmentation is that it does not explore other opportunities to identify market segments that could be more meaningful.

=== Post-hoc segmentation ===

In contrast, post-hoc segmentation makes no assumptions about the optimal theoretical framework. Instead, the analyst's role is to determine the segments that are the most meaningful for a given marketing problem or situation. In this approach, the empirical data drives the segmentation selection. Analysts typically employ some type of clustering analysis or structural equation modeling to identify segments within the data. Post-hoc segmentation relies on access to rich datasets, usually with a very large number of cases, and uses sophisticated algorithms to identify segments.<ref>Wedel, M. and Kamakura, W.A., ''Market Segmentation: Conceptual and Methodological Foundations,'' Springer Science & Business Media, 2010, pp 24-26.</ref>

The figure alongside illustrates how segments might be formed using clustering; however, note that this diagram only uses two variables, while in practice clustering employs a large number of variables.<ref>Constantin, C., "Post-hoc Segmentation using Marketing Research," ''Economics'', Vol 12, no 3, 2012, pp. 39–48.</ref>

=== Statistical techniques used in segmentation ===
[[File:Clustering.jpg|thumb|300px|Visualisation of market segments formed using clustering methods]]

Marketers often engage commercial research firms or consultancies to carry out segmentation analysis, especially if they lack the statistical skills to undertake the analysis. Some segmentation, especially post-hoc analysis, relies on sophisticated statistical analysis.

Common statistical approaches and techniques used in segmentation analysis include:

* Clustering algorithms<ref>https://inseaddataanalytics.github.io/INSEADAnalytics/CourseSessions/Sessions45/ClusterAnalysisReading.html., ''Cluster Analysis and Segmentation'', Online: inseaddataanalytics.github.io/INSEADAnalytics/Report_s45.html [with worked example]</ref> – overlapping, non-overlapping and fuzzy methods; e.g. [[K-means]] or other [[Cluster analysis]]
* [[Conjoint analysis]]<ref>Desarbo, W.S., Ramaswamy, V. and Cohen, S. H., "Market segmentation with choice-based conjoint analysis," ''Marketing Letters,'' vol. 6, no. 2 pp. 137–147.</ref>
* Ensemble approaches – such as [[random forest]]s<ref>Perbert, F., Stenger, B. and Maki, A., "Random Forest Clustering and Application to Video Segmentation," [Research Paper], Toshiba Europe, 2009, Online: https://mi.eng.cam.ac.uk/~bdrs2/papers/perbet_bmvc09.pdf</ref>
* [[Chi-square automatic interaction detection]] – a type of decision-tree<ref>Dell Software, ''Statistics Textbook'', Online: https://documents.software.dell.com/statistics/textbook/customer-segmentation {{Webarchive|url= https://web.archive.org/web/20161022161158/https://documents.software.dell.com/statistics/textbook/customer-segmentation |date=2016-10-22 }}</ref>
* [[Factor analysis]] or [[principal components analysis]]<ref>Minhas, R.S. and Jacobs, E.M., "Benefit Segmentation by Factor Analysis: An improved method of targeting customers for financial services", ''International Journal of Bank Marketing,'' Vol. 14, no. 3, pp. 3–13.</ref>
* [[Latent class model|Latent Class Analysis]] – a generic term for a class of methods that attempt to detect underlying clusters based on observed patterns of association<ref>Wedel, M., and Kamakura, W.A., ''Market Segmentation: Conceptual and Methodological Foundations,'' Springer Science & Business Media, 2010, p. 21.</ref>
* [[Logistic regression]]<ref>Burinskiene, M. and Rudzkiene, V., "Application of Logit Regression Models for the Identification of Market Segments", ''Journal of Business Economics and Management'', vol. 8, no. 4, 2008, pp. 253–258.</ref>
* [[Multidimensional scaling]] and [[canonical analysis]]<ref>T.P. Beane and D.M. Ennis, "Market Segmentation: A Review", ''European Journal of Marketing'', Vol. 21 no. 5, pp. 20–42.</ref>
* [[Mixture model]]s – e.g., EM estimation algorithm, finite-mixture models<ref>Green, P.E., Carmone, F.J. and Wachspress, D.P., ''Consumer Segmentation Via Latent Class Analysis, ''Journal of Consumer Research, December, 1976, pp. 170–174, DOI: https://dx.doi.org/10.1086/208664</ref>
* Model-based segmentation using simultaneous and [[structural equation modeling]]<ref>Swait, J., "A structural equation model of latent segmentation and product choice for cross-sectional revealed preference choice data," ''Journal of Retailing and Consumer Services,'' Vol. 1, no. 2, 1994, pp. 77–89.</ref> e.g. [[LISREL]]
* Other algorithms such as [[artificial neural network]]s.<ref>Kelly E Fish, K.E., Barnes, J.H. and Aiken, M.W., "Artificial neural networks: A new methodology for industrial market segmentation," '' Industrial Marketing Management,'' Vol. 24, no. 5, 1995, pp. 431–438.</ref>

=== Data sources used for segmentation ===

Marketers use a variety of data sources for segmentation studies and market profiling. Typical sources of information include:<ref>{{cite web|work=U.S. Small Business Administration |url= https://www.sba.gov/blogs/conducting-market-research-here-are-5-official-sources-free-data-can-help |date=10 September 2014 |first=Caron |last=Beesley |title=Conducting Market Research? Here are 5 Official Sources of Free Data That Can Help |archive-url= https://web.archive.org/web/20141113211626/http://www.sba.gov/blogs/conducting-market-research-here-are-5-official-sources-free-data-can-help |archive-date=13 November 2014 |access-date=3 June 2022}}</ref><ref>{{cite web|last=Marr |first=Bernard |title=Big Data: 33 Brilliant and Ad Free Data Sources for 2016 |work=Forbes |date=12 February 2016 |url= https://www.forbes.com/sites/bernardmarr/2016/02/12/big-data-35-brilliant-and-free-data-sources-for-2016 |accessdate=3 June 2022}}</ref>

==== Internal sources ====
* Customer transaction records e.g. sale value per transaction, purchase frequency
* Patron membership records e.g. active members, lapsed members, length of membership
* Customer relationship management (CRM) databases
* In-house surveys
* Customer self-completed questionnaires or feedback forms

==== External sources ====
* Commissioned research (where the business commissions a research study and maintains exclusive rights to the data; typically the most expensive means of data collection)
* Data-mining techniques
* Census data (population and business census)
* Observed purchase behaviours
* Government agencies and departments
* Government statistics and surveys (e.g. studies by departments of trade, industry, technology, etc.)
* Omnibus surveys (a standard, regular survey with a basic set of questions about demographics and lifestyles where an individual can add specific sets of questions about product preference or usage; generally lower cost than commissioned survey methods)
* Professional/Industry associations/Employer associations
* Proprietary surveys or tracking studies (also known as ''syndicated research''; studies carried out by market research companies where businesses can purchase the right to access part of the data set)
* Proprietary databases/software<ref>Wedel, M. and Wagner, A., ''Market Segmentation: Conceptual and Methodological Foundations,'' Kluwer Academic Publishers, 1998, See Chapter 14</ref>