The following is from the March 2007 Newsletter. Got a question about Customer Measurement, Management, Valuation, Retention, Loyalty, Defection? Just ask your question. Also, feel free to leave a comment.
Want to see the answers to previous questions? The pre-blog newsletter archives are here, “Best Article” reviews here.
Optimizing Mail Drops for Consumables
Please note: The business discussed below is a “continuity business”, where customers consume the product and need to either reorder from the company every few months or seek alternatives sources for the product. In this scenario, the behavior of customers is generally governed by the Latency Metric.
Q: Currently we mail our current customers direct mail every 6.5 weeks. We have a new VP and he is asking if that is the optimal spacing of mailings. I’m wondering if there are any best practices for setting up frequency tests? If you can shed any light on how to set up such a test I would greatly appreciate it.
A: Well, do you know how you got to the 6.5 weeks number in the first place? Somebody must have thought it was a good idea based on some kind of data (I hope)!
Obviously, there is some significant financial risk in simply “moving the drop around” and testing results that way. You can do it, often by slivering off parts of the drop and dropping then at different times, but there could be a substantial financial penalty for approaching the problem this way – both on the cost and sales sides. This is especially true when you have a current schedule that seems to be working.
The first thing I would do, if possible, before taking on the risk of messing with the mailing is to see if you can find any segmentation / frequency that makes more sense from the customer data itself. Since you also have a web site, there probably is evidence of “natural purchase cycles” the customer engages in that operate outside the mail drop – customers ordering “when and how they want to”.
Can you find evidence that the average purchase cycle is more like 5 weeks or 7 weeks? How does this differ by product line, or packaging of the product? Both segmentation by actual customer behavior and segmentation by product line will generally provide increased profits, provided the cost of dropping different mail streams does not overpower the increased sales.
For example, if someone can buy a “90-day supply”, well, 6.5 weeks is a bit early for the mailing, I’d think. If they can only buy a 30-day supply, well, it seems to me that 6.5 weeks could be a bit late. Look to actual purchase cycles by product line / supply length and see if you can find any patterns in the purchase behavior.
The key to this kind of analysis is to line up all the customers so that the purchase cycles match. In other words, you need to enforce the same start date. One way to do this, for example, is look at all new customers who started in January 2007; of the ones that bought again, when did they purchase – 5 weeks, 6 weeks, 7 weeks out? What percentage of new starts in January (or any other month) purchased in each of the subsequent weeks? Be aware choosing a single month may create results that have a seasonal bias, but I’m not sure that is relevant in a product line like yours.
A more complex but possibly more accurate way to do this is to “normalize” the start date of all new customers in 2006 and then look at the subsequent purchase patterns – given the same start date, what percent bought again 5 weeks out, 6 weeks out, 7 weeks out? You can achieve virtually the same thing by taking each month of 2006 and running it through the same drill as the one described above for January 2007, though it won’t be as accurate.
Once you have nailed the cycle for new customers, you can move on to see if there is any change in optimal cycle date as customers age. My guess is the cycle probably gradually lengthens until the customer defects. If this is true, it might be worth it to do two mailings with different cycles – one cycle for customers who became new customers in the past (say) 6 months and all other customers. It’s likely in this business there could be an important behavioral difference between new and current customers that would allow you to deliver a more optimized mailing cycle.
Failing access to any analytical means to drill down into the data first, because either you lack the resources or simply don’t have the time, set up your next drop with flagged segments based on “weeks since last purchase” and look at profit per customer. You could also back into this if you have good promotional history on your customers.
In other words, if you are going to drop “everybody” at the same time, there must be a segment where for this single drop, the time since last purchase based on arrival of the mail is 5 weeks ago, 6 weeks ago, 7 weeks ago, and so forth. If you flag these segments before the drop in the database, you should be able to go back and determine sales per customer mailed for each segment. This will tell you if your timing should be adjusted. Further, you might divide these time-based segments, if there are enough members in the segment, along various product lines.
Then, once you have a handle on the general cyclicality of different segments, you can get to profit per segment by using control groups to measure the lift and profit by segment.
A careful analysis of the next drop (or as I said, a previous drop if you have good history) should tell you which drop cycle for each product line is optimal. From there, you have to look at economies of scale and decide if you can afford that kind of segmentation. You may find that due to the economies of scale in the mailing, you simply cannot drop 50% of your mail one week and the other 50% the next, for example. But you might find enough support in your analysis to either justify the current 6.5 week drop as the most efficient, or to move it up or back somewhat.
Another way to approach the “timing problem” relative to economies of scale would be to try “reminder to re-order postcards” instead of mail or catalogs to some members of the group that require special timing considerations. For example, new customers might not really need a catalog on their first drop, a postcard driving them to the phone or web site to reorder might be enough.
No silver bullets, I’m afraid. Just good ‘ol fashioned sloggin’ through the data ought to get you to where you want to go!