Welcome back to Marketing Insights from Bedford Street Marketing. Hope you had a great weekend and a productive Monday!!! Tonight’s post is on Sub-Segmentation and it’s importance in marketing campaigns and strategy.
Everyday my Twitter feed and Instagram feed is filled up with offers to get more likes or followers using new advertising methods. Be it a digital ad strategy or some platform that delivers a new and improved creative that is sure to get 1,000 likes in 10 seconds. I see these and think – you’re selling the body of the car (the creative aspect of advertising) without an engine. The engine that makes all marketing campaigns successful are analytics and data-driven segmentation of customer or prospect universes. The creative may impress and influence the decision to purchase but it’s the solid foundation of using segmentation to find the right pool of consumers that drives the campaign.
Sub-Segmentation is one of these latest “buzz” words I’ve been hearing lately. In honesty, it’s a new way to describe what database management companies have been doing for a long time. It’s a new term for an old tried and true way to maximize marketing efforts. The definition of Sub-Segmentation is finding targeted and responsive populations of prospects/customers out of a larger database. Imagine you have a in-house customer file of 1 Million records, you don’t want to market to that entire audience the same way. You want to find pockets of consumers who are most responsive like Soccer Moms, Millennial Shoppers, Gen X Movers or target by specific geographic criteria. These are all examples of Sub-segments of your larger base universe. Once you have found your best sub-segments, then you can accordingly design your marketing strategy and what channels to use based upon this learning.
Bedford Street Marketing recommends three distinct methods to sub-segment your data set or contact us at firstname.lastname@example.org, and we’d be happy to do this for your company.
Method 1: Empirical Sub-Segmentation
Who knows your customer base better than you? Daily interaction with your customers gives you great insight into who are your best customers and prospects. This method is the least analytical, but effective in meeting it’s end goal.
There are some great tools out there to help you with your empirical sub-segmentation. For instance, if you are tracking hits and conversions on your website – Google Analytics is a great tool and should be used. It allows you to see simple reports on your audience, including demographics and new vs returning users. Additionally, you can use already created Audience groupings or create your own custom ones to see how well these groups convert on your site. Further you can capture how you acquiring these hits (channel analysis) and also the behavior on your site (how long each session is, what page are they landing on and what page do the exit on). It’s a great tool and very easy to set-up.
If you want to dig a little deeper into your customer base using this empirical method, Excel is a great option. Using pivot tables and some of the advance formula options, you can slice and dice your data into smaller, more targeted segments. The only draw back to using Excel is the limit on total number of records allowed – which is 1 million.
The Empirical method is good for initial testing, and can be used as a basis of further analytic methods mentioned next.
Customer Profile and Chaid Decision Tree Analysis:
Let’s say you’ve run a few tests using the Empirical method and the results have not been where you want them to be. A next step to further enhance sub-segmentation is running a Customer Profile and utilizing Decision Tree Analysis.
Customer Profiles are done by taking your in-house file and matching it up against a sample of a large demographic database. A Customer Profile will index how your customers look compared to the national sample and clearly point out your best customers. A word of caution when using this method – national demographic files have hundreds of elements, all of which may not pertain to your business. Make sure the demographic elements you are using in a profile are the right ones. There are large companies who have API’s that allow you to do a profile – like Acxoim or smaller database management firms can do offer profile services as well. Bedford Street Marketing suggests looking at all your partner options before making a choice.
The next step, after you have your Customer Profile, is to do further analysis. A Profile will show you the demographic elements that your customers index well against – but they won’t show the correlation of these elements or how to effectively combine them to find the best sub-segment groups. Using CHAID Decision Tree Analysis, available in tools like IBM’s SPSS Statistics, will find the right combinations of elements and create these prime responding segments.
Decision Tree Analysis is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences, including chance event outcomes, resource costs, and utility. It is one way to display an algorithm that only contains conditional control statements. That’s a fancy way of saying, it will take all the data into account, using whatever parameter is the target goal (sales, calls, conversions, net response, etc) and calculate algorithms that combine these elements or create the sub-segments that will look-like your best customers. As noted, this analysis can take into account internal costs, ROI and all the important factors to the target goal of your business.
Finally, we get to the last method to sub-segment your customers and prospects. There are many types of models available to businesses that can get find the right prospects to contact. These include Mailed Regression (looking at Responders vs Non-Responders from a given campaign(s) and Clone Models (looking at customers or response data against a national sample). Unlike the Decision Tree Analysis, models produce Segment Ranks and model scores. The final output of a model can be a 10 Rank or 20 Rank universe, where the top ranks are the best prospects/customers. Models are more in-depth and require working with statistical analysts to get the job done.
The benefit of a model is that it reports the top 10 to 15 elements/descriptors of a best responder or customer. As noted above, it gives each record a raw score and groups these prospects into ranks – which allows for a robust testing strategy. For instance, the initial test could be one where you test Rank 1 vs Rank 2 vs a cross-section of the top 10 Ranks, using the same creative. As you get results back in, you can roll out with the winner and always test deeper into the model. Bedford Street Marketing highly recommends using models in your marketing strategy.
Well, hope you’re all still awake. 🙂 The three methods mentions above are fantastic ways to improve your marketing strategy and increase the results for any marketing channel. Working with a Marketing Partner familiar with all three and who can walk you through these models and Profile processes is something that is highly recommended.