Hey,
For many, attribution in CRM Marketing is simply a “last-touch” model, where all transaction value is attributed to the last message the customer engaged with before completing a desired action. In Acquisition Marketing, the term Multi-Touch Attribution is many times used with various half-baked solutions trying to solve it by, for example, setting predefined weights to each touch point, or using a “time-decay” approach.
But there’s no reason to default to last-touch attribution in CRM marketing. Many times, such a model may mislead you into making wrong decisions about your strategy and cost you dearly. Why would you attribute weeks’ worth of multichannel efforts to encourage a customer’s next purchase only to credit the last SMS they received just before making that purchase? Similarly, there is also no reason to have to “guesstimate” the relative contribution of each touch point.
You have a better option, right at your fingertips. Optimove’s unique scientific approach to incrementality measurement. You see, using control groups on your campaigns, sets you up for measuring the exact attribution of each campaign with statistical credibility. Even in instances when customers are exposed to multiple campaigns on the same day.
Let us explain in detail below while we highlight for you the ultimate multi-touch attribution solution.
Cheers,
Shai Frank, VP of Product
Setting you up for success: control groups, representativeness, and microsegments
Ok, it isn’t as simple as saying the words “test and control.” Why? Because for control groups to be worth their perceived “sacrifice”, they have to be representative. This means that for each customer receiving your campaign, there is a “twin” in the control. Same as with political polls or clinical trials. Otherwise, you have the risk of dividing your target group in a way that places all similar customers on one side of your test, leading to biased results.
Not an easy feat unless you can identify the hundreds of different customer archetypes you have in your database. But lucky for you, you already have these hundreds of microsegments already built into Optimove. If you’ve never seen them, go to your Customer Model and scroll to the bottom. Microsegments stay up to date based on how customers engage with your campaigns, perform new transactions, and more.
Only once you establish a representative control group, you can confidently measure the difference between sending and not sending a campaign. That is, how much value should be attributed (the average incremental uplift per customer) to each specific campaign.
Let us explain this all with an example
Imagine you want to target all your customers in the New Lifecycle (yes, we preach granularity and not mass batch-and-blast but roll with us here) with a campaign measured on the revenue it generates. This target group probably includes customers from a good few dozen of microsegments (i.e. customer archetypes).
Before executing your campaign, Optimove will allocate a proportional number of customers from the various microsegments into the control group, to make sure it’s a representative sample of the customers in the test group.
Once your campaign is sent, Optimove will measure the average difference in customer spend between those who received the campaign and those who didn’t. That is called the average uplift per customer.
And that is exactly how much revenue can be attributed to this specific campaign.
Taking attribution multi-touch
Now you probably are thinking: “well Shai, that is easy, but what happens when I want to look at multiple campaigns?”
Well, that’s the beauty of average uplift per customer. It works for multi-touch attribution as well.
Let’s imagine you have a Stream with two campaigns. We’ll call them A and B. And let’s also assume that Daniel, one of your customers targeted with this Stream spends $100 by the end of it.
Traditional “last-touch” attribution would show you that campaign B generated $100 of revenue, whereas campaign A had zero contribution. That’s probably severely overestimating the impact of campaign B while underestimating the impact of campaign A.
But by going into Optimove’s Stream Analysis you can see the average uplift per customer for campaigns A and B. Imagine that Campaign A’s results are $5 and Campaign B’s $10.
The question then is – how much money of the overall $100 can be attributed to each campaign?
The answer: $5 can be attributed to Campaign A, $10 can be attributed to Campaign B, and the remaining $85 can be attributed to Daniel’s spontaneous behavior, irrespective of any campaigns (meaning, Daniel would have spent $85 anyway, even without receiving any of these campaigns).
The same is also true if Daniel would be targeted by multiple campaigns on the same day or any other multicampaign scenario. This is because, for each Daniel, in every campaign, there is a Daniel-x (a twin) placed in the control group, which has been taken from the same microsegment and has been targeted with all the rest of the campaigns except for this one. If Daniel received three campaigns that day, there is a Daniel-x that received none. If Daniel received only 2, then there is a Daniel-x that didn’t receive these 2. And so on.
So, no matter what combination of campaigns you are trying to measure, if you are using Optimove’s control groups, you are measuring your marketing’s multi-touch attribution, on steroids.
If you’d like a bit more information, we’ve outlined it in this Academy article under the “Interpreting Uplift and Attribution” section.
> From our blog
Celebrating CRM Excellence
For the second year in a row, Optimove celebrated the best in CRM Marketing with its Heptagon Awards. Read all about the winners in our latest blog.