This little piggy…

Ho21rvpSow do you slice up which of your marketing channels get what slice of the attribution pie?

Hopefully we have moved beyond simply assigning everything to last click. And hopefully we’re well beyond first click.

So which direction do you take?
Well, let’s take what Google Analytics has to offer. If you haven’t explored, you can find it in “Conversions” > “Attribution” > “Model Comparison Tool”.
I’ve already talked about last click. It’s just plain wrong.
And first click. It’s just wrong too.
Linear attribution is less wrong that first and last, but frankly there’s gold, silver and bronze and someone deserves the gold. It’s a little mean not to give last click the gold, right?
The time decay model says the more recent the click, the higher up the leaderboard (the gold medal!). You just need a half-life. 7 days, 30 days? 90? And that depends on your product.
What about a position based model? 40% to the last, 40% to the first and the rest distributed to all the other channels? Well, we’re getting there!

So what about a more motive based attribution model? Taking into account a little bit of research before purchase, this would be a position based model, giving a last click the gold medal (say 40%), and first click a runner’s up prize (say 10%) because without that first click you wouldn’t even have had the first date. And then giving the rest to the channels in between.
But then how do you distribute among the other channels? What about an engagement metric? How deep into your site did they get? How long did they spend. Maybe adding extra weighting to social if you were running a big social campaign, and correlating any timings of your offline campaigns with online activity.
The upshot is, you can have an attribution model like the one I suggest above, but the metrics that make up that model may well have to change with every campaign you run.

There are certainly logarithmic models that have been developed and continue to be developed (ask your agency!). But be wary that a one size fits all might not be suitable.

I hate “big data”

mf6CKEMBig data? That’s a term we hear mentioned a lot, but what does it really mean? As businesses we have loads of data? Well that’s true. In truth we’ve always had lots of data to play around with, it just came in different formats and from different sources (sales data, footfall, even web log files).

Yes, we now do have so much more. But it’s only as big and scary as you want it to be. “Big data” is a term that, in my view, is all too often  used to sell you more services to try and make sense of it all. I’m not saying these are wrong though, but…

…the danger is running before you can walk. It doesn’t matter if you have loads of data that you’re not even touching right now if you don’t understand the questions you are being asked, or the questions you should be answering without being asked.

Don’t add on expensive session replay if you still have a poorly implemented web analytics service that you’re not really using to full effect. It’s a guarantee that there will be both quick wins and deeper levels of optimisation that you can extract from your web analytics tool without trying too hard.

If you’re ready to add extra levels of digital analytics, whether session replay or surveys or testing capabilities, then you won’t be fearful of “big data” because you’re already scaling your capabilities.

Don’t let anyone tell you you need a full suite of analytics from day one to gain a deep understanding of user behaviour on your site or app. Eventually you will have a full suite, but build it up in layers.