Always remember, I advocate matchbacks for customer acquisition activities. I recommend mail/holdout tests for housefile names. Matchbacks, in and of themselves, are not evil. How your vendor chooses to apply matchbacks to your business can result in a worst practice. I am not saying you shouldn't do matchbacks. I am saying you should apply them correctly, not in the fashion that some in the vendor community recommend.
Today, we'll briefly chat about the methodology I use, strongly influenced by mail/holdout results, with some matchback information folded in, to determine the optimal number of catalog contacts a customer should receive in a year.
Here's my goal for you. My goal is to reduce your housefile catalog marketing advertising expense by about 20% while losing less than 5% of your sales. My goal is for you to take this money, and re-invest it in catalog customer acquisition (even if at a loss, because that will be more profitable for you than the 27th catalog to an existing customer this year), e-mail marketing, paid search, affiliate marketing, other online marketing, mobile marketing, social marketing, or to have you simply pocket the extra profit. That's what I'm trying to do for you.
Ok then, here's the methodology that I use.
Step 1: Model The Likelihood Of Annual Repurchase. I build a Logistic Regression model that predicts the probability of a customer buying again in the next twelve months. Let's pretend that the likelihood is 30%.
Step 2: Model The Amount A Customer Will Spend, Annually, If The Customer Repurchases. This is a Ordinary Least Squares Regression model that calculates annual spend. Let's pretend that the amount is $150.
Step 3: Multiply Step 1 by Step 2. This yields the predicted demand amount expected of each customer. In our example, this is 0.30 * $160 = $48.00.
Step 4: Calculate The Organic Percentage. This is my proprietary trick. I use Digital Profiles, mail/holdout results, some matchback information, Multichannel Forensics data, and a series of slick tricks to calculate the percentage of demand that a customer will spend if the customer does not receive one single catalog all year long. This is where all of the magic happens! In our example, let's pretend that the percentage is 75% ... meaning that if we don't mail a single catalog all year, the customer will spend $48.00 * 0.75% = $36.00.
If the customer spends $36.00 organically, that means the customer will only spend $12.00 because catalogs are mailed.
If you have mail/holdout testing, a better practice than matchback programs, I fully incorporate the results of mail/holdout tests in this step.
Step 5: Simulate The Profitability Of "X" Catalogs. More magic happens here. I use a "power curve", often derived from mail/holdout tests, but if you don't have a mail/holdout test, I've got a few tricks at my disposal. I create a table that illustrates what will happen at different contact frequencies --- it turns out that customers will shop on their own without the need to receive a lot of catalogs, and they will shop increasingly on their own as the number of contacts are reduced. Therefore, we simply need to find the number of contacts that yields the maximum level of profitability. In the table below, the optimal strategy happens when you mail just two catalogs --- clearly, this customer will continue to spend money if catalogs aren't mailed.
This, Catalog Executives, is what I do. I combine a few tricks, take some common-sense methods, and end up with a clever way to save you 20% or 25% of your budget dollars while losing a very small percentage of your demand, causing you to be much more profitable.
Hint --- this methodology tends to be more accurate and generates more profit than using matchback methodology to evaluate how many catalogs to mail to housefile segments.
This type of project is very popular in 2010 --- so if you want in on the excitement, please contact me for details!