It has been a couple of years since we really started to make the most out of Twitter. Since then we have experimented with a live Twitter feed on our homepage and alongside our blog articles. We have hired a dedicated social media producer. And most recently we joined the Twitter Promoted Products beta.
I don’t want to talk about Promoted Products just yet, as we’re still making sense of things, but we’ve been using Google Analytics to measure Twitter since we started the @econsultancy account and I have a few insights to share. And some numbers too…
I’d like to start with a few caveats.
Firstly, Econsultancy is a B2B-focused brand and our audience, goals and tactics are somewhat different from those of a B2C company. I’m sharing this information mainly because that’s what we try to do, but also to attract more insight from people like you.
Secondly, I don’t think you should read too much into some of these numbers. Not yet, anyway. There is a real customer engagement story here. It’s not all about the easy-to-spot numbers, but sometimes about how they’re interpreted. See my wishlist at the foot of this post for what we should also be looking at.
Thirdly, we haven’t truly optimised our site to make the most of Twitter referrals. Our Twitter account is largely powered by our blog content, and blog articles are certainly not the best converting pages on our (or any) site. We plan on doing a better job of integrating the Twitter API in the months to come, to create new functionality that will increase onsite engagement.
Ok, so onto what we measure…
This is the easiest and most obvious thing to measure, as far as social media is concerned. Let’s take a look at the last 30 days worth of data:
Now let’s see how this compares to the same period in 2010?
Note that ‘Twitterfeed’ is shown as the top referrer. These are links that have had a tracking code appended to them (shown in bold) before we shorten and push them out onto our Twitter stream, e.g…
As such, we might surmise that visitors who enter via our ‘Twitterfeed’ links are either following @econsultancy on Twitter, or may have stumbled upon a retweet by somebody who is.
Further analysis shows that ‘Twitterfeed’ referrals have a higher average order value than both ‘Twitter.com’ (which can be defined as non-‘Twitterfeed’ links on Twitter.com) and also – perhaps surprisingly – customers referred via natural search. This backs up my hunch that customers who have clicked on our Twitterfeed links are far more highly engaged than those who visit via ‘Twitter.com’ (ditto search). And also that it pays to engage, as the theory goes.
Other year-on-year observations…
- The percentage of new visitors referred by Twitter is increasing. Perhaps not surprising given Twitter’s vast growth.
- The overall volume of visitors is also going up… but then again we have been active in Twitter’s Promoted Products beta, and have bought traffic that way. That will account for some of this near-30% growth.
- Relatively few visitors find their way to Econsultancy via the Twitter mobile site. This number seems low, but then most of us use Twitter clients on our smartphones. Mobile users tend to be new visitors.
- Time on site is falling. This could be due to any number of factors, particularly our content.
What we’re especially interested in is whether or not this is new revenue, or revenue generated by existing Econsultancy users who use Twitter to stay tuned into to what we’re doing (much as they might once have done via email, which remains a powerful channel for us).
So, to find out we can look at the New vs Returning Visitors data via an advanced segment called ‘Twitter Referrers’. This data is for the last 12 months, to provide us with a bigger sample (in this period more than 382,000 visitors were referred by Twitter). You can see that 41% of Twitter referrals were classified as ‘New Visitors’:
Now it’s time to see what ‘New Visitors’ contribute to our revenue. There’s a big drop off here, between New and Returning visitors…
So roughly 60% of our Twitter visitors are a) Returning Visitors, and have previously been on the Econsultancy website, and b) Returning Visitors generate 90% of the transactional e-commerce revenue we allocate to the Twitter channel.
It may seem like Twitter isn’t all that great for Econsultancy as far as new customer acquisition business is concerned, and that we should be having a major rethink. But scratch beneath the surface and there’s much more to it than that. It’s not at all easy to dismiss Twitter as a source of value to our business.
THE BIGGER PICTURE
Firstly, the key thing to note here is that we’re only counting e-commerce revenue, which is defined as transactions that take place onsite. Visitors who become customers typically buy subscriptions, book training courses, or sign up to paid events. All of these things have a modest value, typically less than £2,000.
Higher value purchases tend to be invoiced, and e-commerce revenue, as defined by our Google Analytics filter, doesn’t include anything paid via an invoice. So if you’re a new visitor who subsequently asks Econsultancy to create and deliver a global in-company digital training programme, or who sponsors one of our major events, then that’s not going to be counted by Google Analytics if you ask to pay by invoice. These things fall outside of the boundaries of ‘e-commerce’, as far as our tracking is concerned.
Also, we do not include advertising revenue in e-commerce revenue, because it can’t really be considered as ‘e-commerce’. Econsultancy isn’t massively reliant on advertising: we normally show one ad in the sidebar of our blog pages, and that’s pretty much it. That said, I can tell you that Twitter users have accounted for at least £30k in ad revenue over the past year.
As such I’m in no doubt that that revenue derived from Twitter users is in the six-figure range, which is enough to justify the effort and the budget we spend on it.
What we’re less clear about is whether the 90%-10% Returning-New ratio is consistent across the invoiced revenue. If anything I’d expect the higher value accounts to be more likely to be based on existing relationships, rather than new ones. That said, we can allocate around 40% of advertising revenue to new visitors referred by Twitter.
SOME MORE NUMBERS
No more charts, for fear of overkill, and to maintain just a little mystique, but here are some more findings:
Twitter visitors are becoming more valuable. In February 2011 Twitter visitors were worth, on average, 14p per visit value. That’s twice as much as they were worth in February 2010, though is currently about a third of the average site visitor.
We can see that Twitter visitors typically spend less than the average customer (£84 vs £116). Search traffic lives somewhere between the two, with customers referred by a search engine spending around £100 on average. Search volumes are of course significantly higher.
Twitter visitors convert to customers about 56% less than the average site visitor. Search traffic is, perhaps rather obviously, twice as likely to convert as Twitter traffic. I do think that Twitter is becoming more important as a search tool, at least as far as my own behaviour is concerned, but that it remains largely a browse-based medium (unlike Google, which requires a keyword query in order to be useful, and is much more intent-orientated).
Measuring hard metrics such as traffic and sales is just the start. We must step back and see the bigger picture when measuring social media.
For example, I haven’t (yet) measured the accrual of inbound links and any acceleration in link velocity since we started to seriously use Twitter. I’d like to be able to do that, to see how all of the chatter and retweeting on Twitter flows out onto other sites, and how that affects search rankings and referrals.
I haven’t yet looked at the softer brand metrics. I know for a fact that Econsultancy’s brand positioning and reach has improved in the past two or three years, which just so happens to coincide with the rise of Twitter (and our adoption of it as a platform). I just don’t know exactly how much it has improved, nor what the value of that might be.
I’d like to be able to split out Twitter users in more detail, for example by Twitter client, e.g. ‘Tweetdeck’ vs Twitter.com. I have a strong suspicion that advanced Twitter users behave differently from Twitter.com users.
I’ve mentioned Twitter Search, which is a) useful but b) rather limited. As such we’re watching the rise of Google Realtime closely and will soon start tracking it in Google Analytics, to see what proportion of Twitter-related traffic may originate from it.
And finally, I want to look at the customer lifetime value of new visitors referred by Twitter, to see whether they’re any more or less loyal than the average Econsultancy subscriber.
There are all kinds of other things to share with you, and plenty more that I’d like to be doing, but for now let’s wrap this up. I’d love to hear about your own experiences of making sense of Twitter. Do leave a comment below.