January 29, 2020
By: Jeff Fowler
Gather ‘round girls and boys and let the Old Dog learn you another trick! Today’s discussion is about another age-old marketing campaign practice called a Split. A split is exactly what it sounds like: it simply divides an audience into two or more subsets. So – if you have a segment of 100 people, you might split it into two groups of 50 (called an “A/B split”) so you can test two different offers and see which one works better. Simple, right? Well, not so fast Little Grasshopper…
You should know that not all splits are created equal. Here are some fun facts about splits that will make you the life of your next party:
- Much like one sock can’t be a pair, it is impossible to have a single split. You have to have at least two. Oddly, my washing machine doesn’t seem to understand this.
- While there’s no physical limit on the number of splits you can create, it’s tough making a hundred splits out of a segment of 50 people. But 3-way splits are very common.
- Splits don’t have to be proportional. 90/10 splits are often used for something called a Holdout Group. What’s that? A holdout group is a subset – in this case 10% – of people set aside to NOT solicit, so you can measure the Lift of your offer. And what is lift? You sure ask a lot of questions. Lift is the difference in response between these two groups. Simply calculate response percentages and average order sizes for people who were sent the offer vs. those who were not, then subtract the holdout numbers from the offer group numbers. This is your lift. If both these numbers are negative, you’re in trouble.
- Splits are almost always generated randomly. However, MarketWide gives you a few more options, which I’ll talk about later.
So now that you know everything there is to know about splits, here’s a question to see how well you’ve been listening: “How do you test two different creatives to two different segments, without creating A/B splits of each segment?” Continue reading for my exciting answer below! But first, here’s a shameless plug by our Top Sponsor (us).
Did you know that people who use MarketWide™ are happier than folks who use competitive systems? That’s right! We’ve run countless google searches across the entire Internet, looking for any marketing system that creates more happiness than ours, and found nothing. Further, we’ve learned that our customers tend to be among the smartest and best-looking people in the Marketing Industry, if not the entire universe. So, if you’re smart and good-looking (we know you are) and you want to be happy: give us a call!
MarketWide = Happiness
Now before we tackle my question, let’s talk about how MarketWide handles splits. Typically, marketers think in terms of splitting segments, and many systems (including an ancient version of ours, back before we became smarter, better-looking, and happier) are designed with this in mind. With these systems, you basically choose a segment and click a “split” button. Simple enough… but MarketWide takes it several steps further.
Rather than limiting you to splitting only segments, we let you split ANYTHING. Here are some of the cool ways you can split an audience with MarketWide.
- Split by Field – this breaks your audience into separate splits for each value of the field. For example, splitting on Gender generates separate splits for males vs. females.
- Split by Condition(s) – a somewhat advanced feature that lets you enter conditions for each split. For example, you might create separate splits for 12-month multi-buyers, 12-month one-offs, inactives, and prospects. These are defined using the MarketWide filter function.
- Split a Split (of a split, of a split…) – yes, you can split a split. Or split by gender first, then make A/B splits on males and females.
- Split Everything – you can split across the entire campaign, by splitting the universe.
- Split Nothing – as an extra bonus, you don’t have to split at all! To take advantage of this exciting feature, simply don’t click the split button.
Just to brag a little more, if you’re not splitting on a field or using a condition, MarketWide’s Split Manager gives you options to control HOW it should split. Notice in the screenshot below, in addition to making typical Random splits you can choose Distributed or Ranked. Ranked sorts the audience across all splits, such that the first one gets the “best” (or worst) people while the last split gets the worst (or best). And Distributed ensures that the selected field is equally distributed across all splits. For example, if you create 10 splits and distribute by Gender, each split receives the same percentage of males vs. females.
And there it is – the answer to my question! To test two creatives against two segments:
- First, build out the two segments and run your selection. Assign a unique code to each segment; we’ll call them A and B.
- Next, add your touch (output), by dragging over a touch action.
- Now edit your touch and select the first row (called the “Universe”), then open Split Manager as shown above.
- And finally, set the number of splits to 2, choose Distributed as your Type, and select Segment_Code as your distribution (“Based On”) field. This ensures that each segment is represented equally in each split. Now label each split; we’ll refer to them as 1 and 2.
The result? One campaign, two segments (A and B), and two splits (1 and 2). Each split will have an equal percentage of A’s vs. B’s.
That’s a Wrap
For a tiny 2-segment campaign, all this may seem trivial, but it’s a much bigger deal if you have campaigns with hundreds of segments and need to split most or all of them. Using other marketing systems, creating A/B splits of 500 segments generates 1,000 splits, while with MarketWide it’s just two. There’s even more nifty time saving features I could talk about, but that’s for another day.
Want to see MarketWide in action? Contact us.