A couple of weeks ago I wrote a post about some of the problems I’ve encountered when trying to gauge accurate traffic from our social media channels, particularly Twitter.
Based on the overwhelming response in the comments section, and feedback from our own recent social media measurement roundtables, it appears this is a very common problem.
Lots of you took the time to offer up ideas and recommendations, so I thought it would be useful to recap (and test) some of the suggested solutions here…
This is the most obvious problem here, so I ran a quick test to see if we could make an appreciable difference. On the 18th February, I appended every tweet from @Econsultancy with the following code:
Now I can visit the ‘campaigns’ section of Google analytics and all the visits generated by the tweets we sent will appear there:
Utm tracking is probably the most common way to keep track of outgoing campaigns, so you’d think it would be fairly accurate.
Here’s a tweet sent on the day:
(we send most links twice for the benefit of followers in different timezones).
According to Twitter, 699 people clicked those links.
And here’s the amount of visitors we received from that tweet, as defined by Google Analytics:
408 visits. So we’re currently missing 291, or 41% of our ‘should-be’ visitors.
Reasons the numbers might not add up:
1: People love us
As several of you pointed out. ‘Clicks’ don’t necessarily equate to ‘visits’.
It’s certainly possible that users were visiting as part of the same session. So they’d visited a post, left, and visited another within a 30-minute period.
At 41%, the numbers seem a bit high for this to always be the case, but in general that’s good news for us as it indicates a highly engaged following.
2: Twitter’s click measure is inaccurate
This seems extremely unlikely. If there’s one thing that Twitter should be able to measure, it’s the ‘physical’ action of somebody clicking on a link.
If we were promoting these posts using a CPC campaign, that click count would be used to charge us.
Incidentally, if you are running promoted tweets campaigns, Twitter’s ad platform doesn’t seem like it’s penetrated popular awareness to suffer from significant click fraud (yet…) so while I’m sure it may go on to a certain extent, I’d say Twitter’s billing is pretty sound.
3: Operating system meltdown
A lot of people mentioned this. Here are our all traffic figures for the same period site by OS:
4: Accidental attribution
The other day we noticed something unusual in our Google results.
We Googled ‘Econsultancy mobile marketing’.
We got this link in the search results: http://econsultancy.com/uk/blog/62128-six-examples-of-mobile-marketing-excellence?utm_medium=email&utm_source=daily_pulse
That link is appended with this utm code: utm_medium=email&utm_source=daily_pulse
That code is assigned to our Daily Pulse newsletter. So how did it get on Google?
Frankly, I’m not too sure, but that does mean that people reaching that article from Google may be attributed to the Daily Pulse.
And Google indexes Tweets. So some of our Twitter links might well be attributed to Google, and vice versa.
At this point in the analytics funnel I usually recommend a stiff whisky.
We also had a lot of people point out some salient points about Twitter that all social campaign managers would do well to remember:
You can’t please all the people…
We have 129,000 followers on Twitter.
None of our tweets are ever, ever seen by them all.
Wednesdays at 2pm is our highest potential reach period, with around 26% of our total audience active on Twitter at that time.
Assuming they are all even human, that doesn’t mean that if I tweet I can expect 33,000 people to see it, because they might be eating a sandwich, or staring aimlessly out the window, or looking at another tweet at the time. And even then they aren’t going to all click on it.
Simple stuff but still not focused on enough by a lot of people.
RTs don’t always equate to visits:
We generate a lot of Retweets, which is geneally a good thing as it gives us a chance to expand our sphere of influence, but it’s important to remember that some of those RT’s will be from bots, or from RSS feeds, so many people are curating and sharing without actually visiting all the links they share.
In general you can expect to get more than one visit from one RT, but not always, so keep this in mind when you’re looking at engagement figures.
It’s also worth mentioning that analytics is measuring decodes, which are just a touch different to actual visits by living, breathing humans.
And even the live ones don’t always stick around long enough to do anything useful.
The vast majority of tweets go through bit.ly or t.co shorteners. As commentor ‘Jevgenijs’ pointed out, these are fast, but they still might increase bounce rates.
On ecommerce sites, a one second delay in page-load can cause 7% loss in customer conversions, while 40% of visitors will leave a page that takes more than 3 seconds to load.
Suddenly, that high bounce rate from BlackBerry devices makes a lot more sense.
What’s the solution?
There are some services emerging that are designed to track content in a more accurate way as it wends across the social web.
To do this properly, social needs to be factored in to your overall approach to digital tracking. Although there are plenty of ‘Heads of Digital’ out there, social media manager is still often a junior position in many cases, so implementing this can be problematic.
In addition, if like Econsultancy, you are looking for a variable set of metrics for different on-site pages (for example, from our blog, or from report pages), then this may be further complicated.
Make sure you have a clearly deliniated set of tracking codes for each channel. It won’t be perfect, but it can help to clear up some of the more unambiguous channels.
Burn your silos
While I was testing all this, I spoke to our own Data and Digital Marketing Manager, Ben Barrass, who made a straightforward point about analytics data that is increasingly the elephant in the room:
These numbers will never match.
We can look at attribution from trackable traffic and say “Ok, so social is responsible for 20% of attributional conversions”.
What we can’t do is then say “that means 20% of direct attributional conversions are also from social”.
In Econsultancy’s case, I’m fairly sure it’s much higher than that, but there’s no possible way I could ever prove that statement.
Overall, trying to track social in a granular way is a useful guide, but it’s not the be-all and end-all.
As Stuart Hall put it: Whatever the stats do or don’t say, how do they help ‘move the needle’ to meet business objectives.
It’s here that social media represents a unique problem. Although I’m generalizing massively, I don’t think it’s unfair to say that a majority of businesses consider social to be a function of marketing.
Unfortunately marketing tends to revolve around short term campaigns, which demand quicker, measurable results.
Social media keeps running, all the time, so more valuable metrics can only be identified when you have (at least) several months of data, and need to revolve around long-term figures like overall growth, brand sentiment, yearly revenue increases and lifetime customer value, rather than “we did A, and B happened the next day”.
These metrics tend to come from having a unified vision of your business.
I’m reminded of a social media cliche: You can’t control the message. This originally referred to the way people interacted with your communications. In this case, it also covers the way that message is taken apart and seeded across countless channels. You have no control over this, and in most cases your best efforts can only give you a guide. Likewise, while different analytic systems excel or lag in various areas, none are quite equipped to deal with the wonderful chaos of the internet.
Social may be a function, but so is everything else you do, so the ultimate measurement here is probably “How well is your business doing?” We need to be looking at You can do all the research into individual channels you like, but they need to be completely aligned if you want an accurate picture of your progress.
Thanks again to everyone who offered advice on this last time around, as always I’d love to hear any comments you have.