Parametric’s principals, Chris Robson and Scott Laing, offered attendees at this year’s MRA National Conference an approach to enhancing the relevance of traditional segmentation research. “Marketers are increasingly attracted to the use of large-scale data assets such as CRM and POS databases, as well as commercial sources like D&B,” Chris told attendees. Rather than producing segmentation models that are purely descriptive personae, they offered an approach to give primary research a central role in program development and targeting. You can find the full presentation here.
We often hear from clients that they have spent a lot of time and money developing a segmentation model, and yet are struggling to find ways to make it relevant to their business. Unfortunately, not all segmentations are created equal — and different segmentation types are relevant to different business problems. One of the biggest mistakes we find is that people develop purely descriptive segmentations and expect them to inform all their business problems. However, it’s not always enough to say what people within a segment are like. To remedy this we often recommend a two-layer approach, combining a descriptive segmentation (with an emphasis on motivations and gap analysis) with a reach strategy, an optimum way to reach individuals within key descriptive segments through targetable attributes.
- Descriptive Segmentation: The first layer we propose is to develop a descriptive segmentation that allows us to discuss, reason and quantify the primary groups of interest. In addition to the usual attitudinal and behavioral factors typically included in such descriptive segmentations, we recommend including a motivational dimension, that measure what factors motivate individuals to participate in a relevant activities, e.g., purchasing, community involvement, etc.. From this we are not only be able to describe and size the segments, but we are able to identify gaps that show where the current offerings are meeting the motivational needs of individuals, and where there is opportunity for deeper engagement.
- Reach Strategy: While descriptive segmentation is a powerful tool for describing segments and reasoning about opportunities for improved penetration, the factors used are often not targetable from a marketing viewpoint. We recommend that a reach strategy algorithm is developed in order to inform how best to reach specific groups of interest. While this is informed by the descriptive segmentation, it is distinct from it as it focuses solely on targetable attributes. Often, this insight can be joined with database information to develop concrete targeting strategies using CRM or other commercially-available data.
- Descriptive Segmentation: In addition to standard demographic and geographic questions we include short batteries of Likert scale attitudinal and behavioral questions. These are analyzed in the usual way through Exploratory Factor Analysis (EFA) and rotation. These question sets are often augmented by a Discrete Choice Exercise in order to measure motivational factors. Respondents would be grouped into segments using standard clustering techniques.
- Reach Strategy: Reach strategies are then developed by using appropriate analysis techniques, e.g., Tree Analysis (CRT, CHAID or similar) to identify where to find segment members based on targetable factors that optimize a specific variable (e.g., likelihood to purchase or participate in offers). Strategy can be documented and reach efficiency (effectiveness) measured through ROC curves or similar.
One of the great things about today’s Marketing Research industry is the level of partnership that exists. Whether it’s sharing ideas, learning new techniques, referring clients or good old-fashioned friendly competition, the industry comes together and works together in a unique way.
At Parametric, we love working with other MR industry partners, either as suppliers, consultants or primary contractors. We believe that by working with others to provide the highest quality solutions we all benefit, and the industry itself grows stronger.
Here are some examples of how we’ve worked with MR partners in the past:
- As methodology consultants for advanced techniques
- Conjoint and discrete choice studies,
- Custom simulations
- Financial analysis of customers, brand and products
- Financial analysis of programs and promotions
- Clustering, segmentation and targeting strategies
- Dynamic modeling of markets
- As a complete design-to-simulator supplier of advanced analysis for your engagements
- As a supplier of end-user custom tools
- As CE/Technology domain specialists
These are just a few ways we can help you grow your business by expanding your capabilities. Don’t feel shy about exploring other options and ideas with us, we are always keen to share ideas (and coffee and beer) with others who are passionate about this industry.
To talk with us about how we might drive each other’s success, contact us.
We’re often asked by client interested in “taking the plunge” into more advanced methods whether they should use Choice Based Conjoint (CBC) or MaxDiff for their studies. Of course – as with all research questions – the answer is entirely dependent on what insight is needed. And of course, that’s rarely a satisfying answer for clients who are struggling with multiple stakeholders! We find a typical problem for product managers we work with will center on prioritizing features and understanding the effect on buyer preference (choice).
To select between CBC and MaxDiff often depends on the nature of the feature “list” to be tested. CBC works well when there is a natural grouping of mutually exclusive features, e.g., color can be red, blue or green, and processor speed can be 1, 2 or 3 GHz. If, instead, we’re working with more of a laundry list of features that are not mutually exclusive, the prioritization problem lends itself well to MaxDiff. Interestingly, price sensitivity is intrinsic to CBC, but has to be added to MaxDiff. We’ve done this many times with excellent results. Likewise, the concept of simulation is closely linked to CBC, but we find that there is often an opportunity to use MaxDiff data to greater effect, developing simulators for evaluating preference for feature bundles.