Monday, 23 November 2009

The click path money pit

There's a problem with Holy Grails. In the end they're just trinkets, golden ornaments that shine but are not useful. In Arthurian legend huge, wasteful campaigns were mounted to recover them.

So it is with click path, the Holy Grail of web analytics.

On the web, give or take a blocked cookie or two, we can track an individual's viewing of ads across multiple websites and over time. Then we can see what they did on our own website - did they go on to buy?

Did they see a branded ad, then visit us and a few of our competitors' sites, then see a price offer for our product and then purchase? And if they did, wouldn't it be amazing to know exactly how that worked?

Projects like this have been tried off-line once or twice and resulted in a horrendously expensive waste of time.

But online it's easier, online all the data is at our fingertips...

Which makes it an easier decision to pour money into the pit.

The first 'surprise' of click path analysis is quite how much data it generates. I was talking to a planner the other day who'd requested the full history for a few thousand product purchases as an experiment to see what he could do with it. The nine million rows that came back were somewhat beyond Excel's capabilities to analyse. If you haven't got some serious SQL programmers on your team, forget it.

So now you've got yourself a major IT project - data collection, processing and analytics. Front ends and SQL databases. This is going to be expensive.

But the results will be worth it, no? The results will put us on the cutting edge.

Well actually, no. Not yet.

If you get to the big database without running out of budget or management patience (which is going to take a miracle) then you've got... not a lot. Now you need to analyse it.

There will be a few big wins. Some sites and ads will stick out as converting well and some general paths to sale will emerge. You could have found that out a lot more cheaply with well-designed surveys though and the rest of the database will just be noise. Lots and lots and lots of noise.

The only way to make use of the little bits of gain that exist in the noise is buying systems that can optimise tiny pieces of the schedule automatically - serving bespoke brand ads and price offers based on browser profiles and the ads they've already seen. Have you got one of those? No, me neither.

If you don't believe me, look at the long tail of a search campaign. You've generated clicks from thousands of words, sure, but 90% of your clicks came from the top twenty terms. The only reason it's worth having the long tail is because it's generated automatically.

Automatic click path analysis will get here, eventually, but the wins won't actually be that big. It's not the holy grail. If you're a media agency don't even try to build it, Google will get there, or a startup will get there, and it will be worth it for them because they can sell the solution to everyone. For most of us, it's a money pit and a reason to defer, prevaricate and delay doing the best that you can without it. In the words of Monty Python...

No comments: