- Image by mac steve via Flickr
Most marketers, when they are faced with the problem of retaining customers, usually think of developing a statistical model to predict which customers are likely to leave their franchise in the coming months/year. Such models score customers, usually from 1000 to 1 based on their likelihood to stop doing business with the company. The model analyzes a combination of customer behaviors, types of purchases and demographics to find patterns among customers left in the past. Those patterns are then used to find customers who are more likely to leave in the future. This approach is widely considered to be a Best Practice, even more so in businesses that have a subscription model representing the bulk of their revenue.
These models have proven consistently to show positive ROI, since by saving a customer who might’ve left, particularly the subscription business, you save not only next month’s revenue but that customer’s future revenue as well.
Recently, I found myself recommending to a client that they do NOT take such an approach, despite the fact that that approach is a clear Best Practice. The reasons why is illustrative of a problem that many of our clients face in growing their business.
The the main reason that we recommended not pursuing a retention model had nothing to do with analysis, but had everything to do with execution. When we probe deeper into how our client was going to act on the retention risk analysis, we discovered that they had relatively few levers to push:
- Basic trigger emails. The primary vehicle that they could use was an off-the-shelf trigger e-mail program based on basic business rules, entered manually. So instead of sending e-mails based on a high risk score, the system could only accept rules such as “when a customer has not come in in 35 days, and had a first purchase under $50, you should e-mail them X offer.”
- Store manager calls. The other lever they had was having store managers call customers directly. Given the limited time that a store manager has available, you can imagine how difficult it is to have that manager called the bulk of the customers who are at high attrition risk. So in summary, what you have is two constraints – restrictions of the rules–based trigger e-mail tool, and of the store manager’s time.
While it may be exciting and financially rewarding for consulting firm to develop a more elaborate model for clients based on best practices, the real proof is always “in the pudding.” What I mean is: If you can’t execute based on the results of a model, don’t build one. All that effort would wind up with some “nice to knows,” an expensive bill, and a lack of clear verifiable results. Net of everything, this formula equals disaster.
Instead, we built cross–tabs that allowed us to explore the relationship between behaviors and customer retention quickly, and provide business rules at the end. We also worked to make sure that, when we provided those rules, we could do so in a format that the software would accept. We also focused on how to prioritize the list given to store managers, so that they could, in their limited available time, only call the highest likely to stay, highest value customers. The result was a lot less sexy, but a lot more effective than a statistical approach. The results were measured in the only metric that really matters, incremental revenue and ROI on the investment.
Many times, executives will get excited with the “shiny new thing.” That can take the form of a new channel (e.g. social media), or a technique like attrition modeling. The goal of successful marketers is to temper the executive’s enthusiasm about the shiny new thing and refocus them on approaches that can be executed and measured, and will drive ROI.
Ultimately, the final arbiter of any analysis is not what you learned from it, but what you did. “Nice to knows” do not pay the bills and do not keep the business (and your career) growing. Base the level of complexity of what you do on what you can execute and you will find yourself saving time and making money.
Related articles by Zemanta
- Putting the Customer back into CRM (customerthink.com)
- How to estimate Lifetime Value; Sample cohort analysis (lsvp.wordpress.com)
- The experience trap (cio.co.nz)
{ 2 comments }
