Wherever you look in digital, data is the new black. Big data, smart data, data that helps make you the hero, rather than having to bow to the opinions of the “HIPPO” – the Highest Paid Person’s Opinion.
However, working with a couple of clients over the past few weeks has been a stark reminder that data on the web is not always what it seems.
Data does not compute, captain
First up, there’s the frustration of data that can’t be triangulated. When data in one source doesn’t match data in another.
For example, if Facebook is telling you that you’ve spent (say) £1,000 on advertising, driving 500 clicks to your website, you expect to see those 500 visits coming up in Google Analytics. And if Facebook also tells you that you made 5 sales from those 500 visits, you’d expect to see those 5 sales in Google, clearly attributed to your wonderful Facebook advertising. Right? Wrong.
Google and Facebook do not report in the same way. Why?
- Facebook measures any click (and any view) of an ad leading to a sale in N days. (You can set the N timeframe for both views and clicks yourself). So even if your sale came days after the user had quit their Facebook session, then they searched for your site on Google and converted, Facebook would still register that as a win for your Facebook ad. Hoorah for Facebook!
- Facebook users must be logged-in, so that means Facebook can track users while they switch devices from mobile to desktop. Typically, Google Analytics can’t do that (unless you have similarly logged-in users). Google will report one user over multiple devices as multiple users.
- When users move from a logged-in Facebook page to an ordinary website, details of where that traffic has come from (the referrer details) can be lost. Google will not show that traffic as coming from Facebook.
So when someone tells you that you’ve had 5 sales from your Facebook advertising, make sure you’ve got the follow up questions to hand:
- Is that tracked in Facebook or Google? (Replace Google with your analytics tool of choice as required)
- Is that last click or view-through conversions? (View-through conversions are users who converted having only viewed the ad rather than clicking on it)
- What’s the attribution window on those conversions?
Google is not your friend
If cross-tool reporting inconsistencies are difficult enough for you, there’s the problem that Google Analytics itself doesn’t report data in a consistent way.
If you flip between E-Commerce reporting and Multi-Channel Funnel reporting in Google Analytics, it will give you different numbers. Wow.
This is because by default, Google uses a reporting approach that favours paid-for campaigns. Like Google’s own advertising mechanic AdWords, for example… in fact, especially Google’s AdWords.
Let’s say a user has visited your site from an AdWords pay-per-click ad, and then later goes directly to your site and converts. It’ll be the Adwords campaign that gets the conversion, not the direct visit. Why? Because it’s always been that way. It’s a legacy issue from the days when digital marketers had to fight over budget with finance departments.
But a different approach is used in the Multi-Channel Funnel part of Google Analytics. Attribution is based on the path to conversion with everything included (as long as you’ve tagged it correctly).
This means Multi-Channel reports are more accurate – but here’s the clincher: since Multi-Channel Funnels are relatively new they lack some of the features of the rest of Google Analytics. So in certain circumstances you’re stuck between the two approaches, giving a slightly different answer depending on the finer nuances of the problem you’re trying to solve.
It helps to clarify exactly what you’re asking and what action you might take as a result of the answer. Then choose the suite of reports that best suit your needs.
Bots oh crazy bots
Ad fraud is a growing problem among media sites and those selling advertising online. It’s estimated that around a half of all web traffic globally doesn’t come from humans. It comes from automated programs – known as bots – that scour the web.
This is a problem for those selling ads because their performance is usually measured in pageviews. If those pageviews are not coming from actual live humans then, well… what are the advertisers paying for?
While Google Analytics will help you block a whole range of known bot sources, new ones spring up every day. And tracking down bots in Google Analytics is hard because due to data privacy restrictions, GA does not track IP addresses – and those are needed to investigate any questionable activity on your site.
If you suspect something’s up with your web traffic, you’ll need to delve into the murky depths of your server logs to find out. Given the right tools, there you’ll be able to identify which IP addresses have high traffic volumes or particularly long or short sessions from the same IP. These IPs can then be filtered from your reporting, or from visiting your site in the first place, depending on how draconian you want to be with them.
Lies, damned lies, and… web data
Fortunately though, these people usually represent only a small fraction of a total population on a popular website – maybe 3% at worst. Still, it is worth remembering that web data isn’t ever usually 100% accurate.
Remind yourself of this the next time you’re provided with a slew of data or reports from which to make decisions from. And come armed with the follow up questions to clarify. Then your data can be more powerful than a HIPPO.