We all love the term relevance, but is it really relevance we are striving for? What most people don't realize is that in email marketing, segmentation is not about defining all your customers and how to treat them; instead, it's about helping you make decisions at a very granular level. Being too high-level with your approach is useless in day-to-day email marketing. If you are too granular in your definitions, you'll struggle to get programs out the door. While you may have a very clear view of your highest valued customers, the question is: Should you focus on these highest-valued customers, or should you work harder at gaining a more granular picture of the less-valued, less-defined segments?
I've seen great uses of various types of segmentation. I've seen demographic segments work really well for businesses that need baseline qualifiers to assess the degree of promotion, incentive and interaction with the brand. Yet household income, age, gender, presence of kids, education and geo-targeting are only part of the way to this elusive stage of "relevance" we seek. Using approaches like Personicx and Prizm Clusters will help you in targeting, but they are less effective at helping you understand motivations -- those beliefs that drive behaviors (frequency of visit, spend, and potential cross-product interests). This approach can be amazingly expensive to keep updated, but again, there are economies of scale if you set up an effective approach to data collection in the first place.
Transactional-based approaches are much more empirical in terms of identifying past purchase histories, along with variables that influence increased purchase, purchase value and category base spend and predictors. Certain categories of business are no-brainers for this model, yet it becomes a really daunting task to manage your email programs with this approach without a very refined approach to modeling that continually updates itself.
You can spend years trying to do predictive modeling, but if your products turn over quickly, or you have thousands of SKUs, it can be impossible to manage iteratively outside of simple targeting or high/low transactional periods. While this is used in email most frequently for personalization and promotional strategies, still, unlike other marketing forms, email marketers still have to deal with the entire database. So the challenge is to have enough resources to consistently execute against the database.
What about taking it a step further with behavioral data? This is sometimes considered the same as RFM-type approaches. A segment is generated based on some behavior: a purchase, an Interaction or event. This is what the online space has been struggling with for years as the customer journey is so much more fragmented between the web, store, call center and outbound channels. Behavioral segmentation is an operations persons' nightmare -- mainly due to lack of agreement on how targeting decisions are made in companies, with a lack of understanding of the customer experience making it even harder to execute at the channel level (who drove the sale?).
Attitudinal segmentation is the coolest form, in my opinion, but also the hardest to operationalize. Now you know who your customers are and you know how to target them, and you may even have some idea of the types of motivations that drive types of promotions, categories of interest and emotional drivers to purchase.
The challenge with this approach is twofold. One, it's expensive to commit to. I've seen many companies do so -- and it just sits there for the annual meeting. I've seen others try to operationalize this without the commitment, funding, resources or technology to extend this past the Business Intelligence groups. This makes decisions that much harder; if you don't have a clear understanding of the values your business or service delivers, along with competitive considerations, it can be a fruitless cause done in a vacuum.
David your article is insightful. I have found companies and organisations to be very much into segmentation, dare I say it, TOO much. They heavily dissect their databases into categories and try to target very finely who their messages are sent to. Email Marketing IS NOT a media which requires infinite segmentations. It is a media which is "Interruptive" in the life of the recipient therefore your RELATIONSHIP with your list and the way you interrupt them with the subject line, and email content header and topic is important.
Email Marketing serves only two purposes: 1. To build a relationship or 2. To make a sale.
Too many email marketers concentrate on always trying to make a sale. This is my terminology is DUMB. Sales have never changed regardless of the media used or the sales person selling. PEOPLE BUY FROM FRIENDS. If are continually sending out SALES emails you will damage your relationship with the recipient and maybe loose their friendship. Think before you email - 'Why am I sending this email to me list?" Cheers Kurt "Email Mastery" http://www.kurtjohansen.com
Well Put Kurt....
David, thank you. The best, most honest, article on segmentation I've read in years. It got me thinking. Too often segmentations, and ways to implement them, do not realize their potential. There are many hazards on which they founder, which you clearly outline.
Throughout your example, there is one common reason why segmentations fail: There is always an easier course of action for an Organization to take. Segmentations often reflect the real complexity of our markets, and living by them requires much (sometimes too much) organizational commitment.
So let's keep it simple. Taking your example, and focusing on the *objective* of each day-to-day communication: A segmentation should tell you 1) where to invest your marketing dollars, and 2) how to make each individual communication more engaging to each customer.
Focusing on simple actionable objectives helps avoid making things too damn difficult for an organization to do. And the simple plan that's implemented beats the 'perfect' one that stays in the drawer. Every time.
What a treat to read an article about segmentation from someone who obviously has been working in the trenches. I can’t resist tossing my two cents’ worth into this discussion:
• “…the challenge is balancing the right segmentation methodology with an organization's ability to work within these guides”. --- Right on! In our experience, it is not the segmentation that’s hard. Executing against it is difficult. Companies struggle to go from spray and pray to writing copy and making offers for even as few as three segments, plus they need to launch the separate mail streams. We became an existing customer marketing solutions company to help our clients with these challenges.
• “If you are too granular in your definitions, you'll struggle to get programs out the door.” --- The segmentation can be broad—perhaps a handful of groups—but each customer can be approached in a personalized way, with personalized offers, by using purchase probabilities and variable data.
• “..yet it becomes a really daunting task to manage your email programs with this approach without a very refined approach to modeling that continually updates itself.” --- Management is simplified with marketing automation through rules-based triggers. And if the analysis is based on behaviors, you don’t need continuous updates to the models.
• “…if your products turn over quickly, or you have thousands of SKUs, it can be impossible to manage iteratively outside of simple targeting or high/low transactional periods.” --- Here, I strongly disagree. We routinely deal with client companies that have well over 100,000 SKUs. We predict purchase probabilities for each of their customers, but it would be unwieldy to do so for each SKU. Instead, we use the client company’s hierarchical division of their SKUs into groups and subgroups, and analyze at that granularity. Most clients can market effectively when there are purchase predictions for up to several hundred groups of SKUs. Further, SKU-intensive and customer-intensive companies have tons of transactional data, which makes the analysis more accurate.
• “…behavioral data?” --- We know this type of data, transactions, makes for the most accurate predictions of what a customer will buy next or which customers are likely to defect. We put data on a scale from passive to active, with demographic data at the passive end, attitudinal data next, then browser clicks, with transactions at the active end. The more active the data type, the more accurate the predictions. We have quantitative data to prove that customer satisfaction survey results do not correlate w/ future buying behavior. RFM is a really weak, low accuracy form of transactional behavior analysis, but still better than some other types.
• “The challenge with this approach is twofold.” [expensive and difficult to operationalize] --- By totally automating the analysis process, we’ve been able to reduce the time and cost dramatically. By automating the execution with triggers and workflow, delivery is simplified. Our clients are implementing differentiated contact strategies based on continued analysis and automated execution and doing this at very reasonable rates. The end result is an increase in customer lifetime value.
I like your approach Kurt. However. I come from the general advertising world - and I learned long ago that people expect an advertisement is designed to sell something. If you don't make an offer or give them an opportunity to respond, they're bewildered. The problem is that what passes for selling today is an inheritance from the era of painless dentistry ads. I devoutly agree that people buy from friend. People also ask friends for advice about products/services. The best selling tactics are built around helping people vs. trying to 'sell' them. Help them find their way in a crowded marketplace and you've probably made a loyal customer. Then try not to screw up the relationship :)
Well put again Kurt... I'd like to add your information to my blog, great insight !!!
Sorry, meant well put Mark... :))