Data’s Messy — Get Over It

Last week was a busy one for online marketers in London, with the conference season getting into full swing. An inaugural Internet Retailing conference was followed by two days of Ad:Tech. By all accounts, both were well attended. This is good news. It shows the industry is thriving and there are more people out there wanting to do more things.

I was at Ad-Tech last Wednesday running a couple of workshop sessions. One was entitled “Data Integration: Why does it have to be so difficult?” I moderated a panel made up of representatives from Tangozebra and Logan Todd here in the UK, and from Sitebrand for a North American perspective.

We didn’t really know what to expect from the audience. We were worried we’d fail to keep them engaged on what might be considered somewhat dry and turgid subject matter.

We needn’t have worried. After a cautious start (we are talking about Britain here, after all) the audience soon warmed up to the theme, which was basically: “How do I make sense of all these numbers? Especially when I have different tools, supposedly showing measuring the same thing, showing me different numbers!”

A show of hands revealed the vast majority of end users were using at least two Web analytics systems to track their online businesses, as well as having other data flowing in from search engine marketing systems, ad-server reports, affiliate systems and so on. That’s probably fairly typical of most online businesses today.

So, how do you deal with it?

Probably the most consistent piece of advice that came back from the panel was that it’s messy and difficult, so don’t spend loads of time and effort trying to get it all absolutely right. At the same time, be absolutely clear about what it is you’re measuring in the first place. That way, you’ll understand the difference between a click recorded on an advertising system and a visit recorded on a Web analytics system. They’re not the same thing.

One of the biggest online marketing fallacies is that’s totally measurable and accountable. I think we all know that just isn’t true. Measuring online marketing performance is fuzzy and complicated.

How do you attribute sales properly, for example? Typically, you may have someone first visit your site on a search referral, followed by a subsequent visit from an affiliate before converting on a later visit by coming directly to the site. Each acquisition marketing channel vendors may have their own rules claiming the sale. How do you avoid paying two or three times over?

The panels’ answer was to set your own rules and decide how you’re going to pay for traffic and conversions. To do this, you also need good data, so you need to spend some time and effort getting tracking systems sorted out so you fully understand the dynamics of customer acquisition on your site.

As with many things in life, there’s no silver bullet for solving data integration and data validation problems. One key aspect that arose in the session is the importance of looking at relative performance rather than absolute performance all the time. We used to have a saying when I was at ACNielsen many years ago: “a trend is a friend.” If data are consistent in a certain direction, it’s probably true. If there’s a sudden change, it’s either as a result of something happening or there’s an issue with the data. Either way, you want to find out why it’s happening.

The same principal can be applied to online marketing data. Rather than worry too much about why “System A” shows a different level of traffic than “System B,” ensure the trends are at least in the same direction. Also, rather than focus too much on whether it’s 10,000 or 12,000 visits, look at what’s happening between different visitor segments, for example.

In such a data rich environment, it’s potentially easy to be consumed by the numbers. The feeling from the session in London last week was you just need to accept that it’s messy and horrible. Get over it and move on. There’s a cost to perfect information that’s probably not worth paying for.

Related reading

tencent_emily-ma_featured-image
kenneth_ning_emarsys_featured-image
bounce-370x229
site search hp
<