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...
MULTIPLE CAVEATS!
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...
TRAFFIC
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.
NEW CUSTOMERS?
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:
VALUE
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.
RELATIVE VALUE
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.
CONVERSION RATES
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).
WISHLIST
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.
Chris Lake is Director of Product Development at Econsultancy, an entrepreneur and a long-term internet fiend. Follow him on Twitter, Google+ or connect via Linkedin.
Recommended reports
Twitter for Business
This report is aimed at companies and individuals who are thinking of joining Twitter and want to find out more about the social platform, as well as people who are currently using Twitter for business, and want a deeper understanding of best practice.
Online Measurement and Strategy Report 2011
An analysis of the online measurement industry based on a survey of digital marketers. The 67-page report and 29-slide summary presentation contain valuable insights into the use of paid-for and free web analytics tools. Beyond web analytics, the report also looks at a range of data sources and information requirements, and the extent to which companies are able to develop a coherent measurement strategy.
5:34PM on 28th February 2011
This is a terrific blog post. The big issue with Twitter is squarely: can I attribute revenue to activity on Twitter? If the answer is yes, then serious investment will follow. Until the question is answered it will be seen as experimental medium. My hunch is that Twitter is about a year away from being taken seriously as a revenue-driving channel. But you may have just brought the day a bit closer.
9:50PM on 28th February 2011
Chris,
Thank you for writing this article. I found great value in your assertions and I am pleased to see more ROI Twitter information.
We recently finished a case study, soon to be white paper, that shows ROI for Twitter in a direct to consumer setting within the wine industry.
Excellent work!
10:29AM on 1st March 2011
Thanks for the article Chris, I've been waiting for someone to write about this for a while.
10:45AM on 1st March 2011
Great post! I belive that tweets will increase e-commerce revenue.But it will take some more time.
Robert
11:03AM on 1st March 2011
@ Mark - with a standard Google Analytics implementation it uses a last click wins model without any attribution - to be able to see conversions off the back of twitter you need something that will allow you to see a sum of all visits, Omniture or something like Atlas would allow you to do that, but GA wouldn't without some serious customisation.
If I visit via Twitter and then come back (and convert) after doing a bit of thought, I may come via direct, google (brand) or something completely different... digging out that old tweet just doesn't really fit with the way most people use twitter!
Great post btw.
E-Business Consultant at Dan Barker
12:51PM on 1st March 2011
hi, Chris,
wow - I love this post. Great information, presented really well.
Have you thought about adding some custom variables to your google analytics setup? Eg. to track 'first ever visit was via twitter', 'has visited at some point via twitter', 'member/non-member', etc?
dan
Director of Product Development at Econsultancy
1:02PM on 1st March 2011
Thanks for the feedback.
@Dan - yes, we want to start drilling down and segmenting, so great ideas.
@Mark - I think it's a clear revenue-driver, it just depends on how you choose to use it, and for a B2C non-media brand there are some obvious things to think about, especially relating to content strategy (we blog every day... many retailers don't, so they'd need to figure out how to feed Twitter to get the best out of it).
Digital Consultant at Form Digital Consulting, www.formdigital.co.uk
1:32PM on 1st March 2011
Great post Chris. And excellent points from Dan and Gerry. In terms of assessing the value of Twitter in getting new customers for Econsultancy, you need to know whether new customers tend to have followed tweeted links *at some point* previously, not just in that visit. The fact that Econsultancy customers are likely to use multiple devices, laptops, PCs, etc, also adds complexity to any cookie-based analysis!
Another thing you could look at is whether your new followers tend to go on to become new customers (and whether this is a better predictor of future purchase than other predictors, like subscribing to your emails or attending an event). This is possible if you ask for Twitter ID at registration/purchase.
Director and Co-Founder at 4Ps Marketing / SEM agency London
2:02PM on 1st March 2011
A really open and interesting post. I look forward to finding out how you as you optimise the site for Twitter referrals.
I always find it interesting to look at the ratio of comments of an Econsultancy post to the number of tweets. For example this article has 9 comments and 245 tweets. I wonder if Econsultancy could find a way of reducing the entry barrier to posting a comment via Twitter. It can sometimes be quite time consuming to post a comment so I will RT with comments but as you can imagine that only gives you a few words.
Keep up the good work.
2:02PM on 1st March 2011
Hi Chris - A really illuminating post - I hope you follow up with the other analysis - will be interesting to see that too. Interesting bounce rates - very high all round.
I guess this is because you state that this traffic is largely generated via blogs and so the user is reading the blog and then closing out. So surely there's work to be done here with getting some signposting/incentives from the blog through to your site - and leveraging the traffic even further.
Ecommerce Manager at Swell UK
2:08PM on 1st March 2011
I've done similar things on my site. Another useful thing to do would be to be add a "follow" button to the site and attach a Google Analytics virtual pageview to it. You can then create a goal in GA to track the number of new follows you gain via the site. You can do the same with the retweet button to measure retweets from the site.
2:10PM on 1st March 2011
I have just started measuring the impact Twitter makes for one of my clients simply because I have begun using it as a real driver of traffic and (hopefully) an aid to SEO. Devoting this time to it, I want to be able to show the client that it pays for itself. It's early days yet, but I believe I'll be able to do that using GA for direct sales.
However, capturing information for those who discover the client via Twitter but return later from another medium is proving a real barrier - any help on that would be much appreciated!
Ecommerce Manager at Swell UK
2:14PM on 1st March 2011
Ben, you can do that with GA's custom variables. Basically, you detect when a user comes to your site from Twitter and then you set a custom variable (stored in one of the GA cookies) which adds the user to a group indicating that they've originally come from Twitter. They'll remain in this group until you remove them, allowing you to see if the customer ever visited via Twitter.
The neat thing to do is to create an advanced segment and then you can easily view all of the people in this group and how they interact with the site.
The snag, I have found, is that you quickly run out of advanced segments (you're limited to 100) and of custom variables (limited to 5) so you need to plan ahead a bit.
3:12PM on 1st March 2011
I have experimented with this for other clients in a different manner and found it very useful. I didn't realise it could be set up in this manner, however. I shall look into, thanks very much, Matt.
4:00PM on 1st March 2011
great stuff! - have a toy with BTbuckets GWOptimizer plugin - it allows you to isolate A/B tests by inbound referrer, so you can really undertake some great Twitter testing.
Director of Product Development at Econsultancy
4:01PM on 1st March 2011
@Adam - Absolutely... there's still some finessing to be done, and I'm a big fan of attribution rather than last click (TagMan is worth looking at as well as the others mentioned by Gerry). As you can probably tell, while I'd like to see *all* the detail, I'm ok with interpretation and extrapolation too. I've previously suggested that social media lives inbetween paid search and TV advertising, as far as measurement goes. There is some room for movement in how we interpret things.
@Matthew - Spot on. I agree that the retweet-to-comment ratio is a very interesting metric for publishers, and one thing we're looking at is integrating the Twitter API to allow people to comment (and maybe to sign up as a Bronze member, or 'Bronze-lite'... I really like how Quora and Empire Avenue do this). Some pros and cons but I think it makes plenty of sense. Anything that a) removes barriers to engagement and b) cleaves that blessed CAPTCHA in two is a good thing.
@Michelle - Blog-based sites often have fairly high bounce rates, depending on how visitors access the site. For example, the rise of news aggregators has meant that readers often find their way in via a clutch of stories / sources (and some will back out and check out multiple versions of the same story). The same applies with social media sites like Digg and Reddit, and also with Twitter. I'm not sure about you, but I get most of my reading matter from Twitter recommendations these days, and I hop in and out of articles. But yes, you're right... we need to get better at signposting, both to membership and further reading (there are no 'related articles' right now, and our 'Learn More' pods are manually-created, which is far from optimal). We'll be optimising the blog in the next three months, all being well, and experimenting a little along the way. Stay tuned!
@Matt - Nice idea - I like it. We do have a 'follow' button on the page, if you can find it, but I don't think we're tracking it via GA. Will ask the question. Good insight too for Ben.
4:43PM on 1st March 2011
Matt - I wonder if you'd be able to help with the setting up the custom variables step by step with regard to this? Google is it's usual technical but unhelpful self on this matter!
Specifically setting up the Twitter 'group' and then adding users to it.
Ecommerce Manager at Swell UK
5:09PM on 1st March 2011
@Ben Yep, no worries. (I'm actually based down the road from you and our sites were originally built by Fluid.) Send me a DM on Twitter @TechPad
Director of Product Development at Econsultancy
5:25PM on 1st March 2011
@Matt - And / or write a guest blog post for us and we'll share your walkthrough with our readers ; )
Ecommerce Manager at Swell UK
5:31PM on 1st March 2011
Yeah, I'd love to. Thanks for the offer!
Director of Product Development at Econsultancy
5:42PM on 1st March 2011
@Matt - If you email me (chris@) we'll pick it up from there. Thanks!
12:08AM on 2nd March 2011
The bounce rate for the Twitter traffic is very high, and higher than the site average - so most of your twitter traffic is exiting upon entering the site or after only visiting one page.
The analysis needs to look at the Twitter referral traffic on a page by page or content view. Which content is engaging the Twitter traffic or is any?
Director at WORKSsitebuilder
6:30PM on 2nd March 2011
Did somebody mention Twitter?
Linkbait!
Only kidding, interesting post Chris.
Interesting to analyse the content of the referring tweets too.
One hypothesis might be that 'real-time' language / hashtags used in referring tweets that are trending can be used in other content to influence organic search (or PPC for that matter)
I bet there are an interesting correlations between the authority / experience of the referrer and consequent engagement. Also, who are the influencers, what do they talk about / link to, and what language do they use?
I suspect that as with search traffic, signposting recent, contextually and temporally relevant related content above the fold will decrease your bounce rate. People read the first line and will head back out to twitter if it is not obviously relevant, but if you can show other contextually relevant content above the fold, and page design supports it being obviously linked to that article then they stay longer.
we've noticed people are blind to 'featured content', 'most commented' boxes, but respond well to pull quotes (in article) with obviously related content.
interesting stuff.
angus
5:47PM on 3rd March 2011
Any word on when that guest blog post with a walk through for your readers is surfacing? :)
9:51PM on 3rd March 2011
I would love to see the walk through. I've been out of heavy GA use for a year or two and am a bit rusty...but picking up quickly. I have a few tagged links out in Twitter, but I am not clear on how to get reports on them exactly…yet. I bet the walk through would help.
7:19PM on 6th March 2011
I have experienced that the bounce rate through twitter is very high. In my opinion we should ignore the Google analytic as twitter depresses Google analytic report.
4:09AM on 7th March 2011
Some information is easy to measure. Some is hard. Some is cheap. Some is expensive.
That means actions that are easy and cheap to measure get measured. Others are pushed aside and people who object are said not to be rigorous enough.
Twitter can cause a lot of cool things to happen. But it probaby costs more to measure it than do it. Which most people don't figure into their ROI.
7:36AM on 16th March 2011
Nice article, we have just started to do the same thing with analytics, and had the same question on if we can track what the users are using to interact with twitter. I totally agree, twitter.com is for newbies, seasoned users and professionals use tweetdeck, hootsuite etc.
Looking forward to the guest blog post.
Marketing & PR Executive at Abel & Cole
11:43AM on 28th March 2011
Has Matt's guest blog post for the walk through been posted yet?
Marketing & PR Executive at Abel & Cole
1:17PM on 28th March 2011
It's here i believe - http://econsultancy.com/uk/blog/7310-tracking-on-site-tweets-and-follows-with-google-analytics
4:07PM on 12th October 2011
A totally thorough and insightful post Chris. Looking at all this data I can't help but think that at times we are guilty of over complicating and over analysing everything to the point where the data no longer becomes as useful?