You have probably see this reported survey on lots of marketing sites.
“A UK based study by Hitwise analyzed data from the top 100 retailers to find how much additional web traffic was generated by each Facebook fan. The metrics showed that each fan of a Facebook page produced an extra 20 visits to the website.”
I do not believe it and here’s the math.
I don’t disagree with the raw figures, which are summarized here:
The methodology is explained for a previous survey.
“A first port of call is usually to look at the downstream traffic from a particular website (in this case Facebook), to see who is benefiting from that traffic … 12% of all visits immediately after visiting Facebook go straight to Google, which would indicate that people have seen something on Facebook which they like, and they want to research that product or brand further through search. Using Search Sequence we can see what people were searching for online immediately after they visited Facebook to capture the data not just from Google, but from Bing and Yahoo! Search as well”
The research reportedly compares the sort of searches that average people do at all times, with the sort of searches that people do after using Facebook, and assumes that any difference is caused by the marketing that people saw there.
Here’s the problem
It is not a safe assumption that people leaving Facebook are like average Web users.
This means you cannot simple compare the raw numbers in the columns labelled “Share of clicks in all Categories” and “Share of clicks after a visit to Facebook” and claim the difference is down to Facebook marketing.
Possible sources of error include:
These groups are not matched by gender: Normal Google searchers are about equally divided between men and woman, but “fifty-eight percent of all Facebook users are women”.
In fact the gender difference is even greater because, “Women spent an average of 16.3% of their online time on social networks in April 2010, compared to 11.7% of men.” This is a huge difference.
These groups are not matched by age. Facebook users are younger - the percentage of Internet users of different ages who use Facebook is… 18-29:86%, 30-49:61%, 50-64:47%, 65+:26%.
- The groups are not matched by intent. Facebook use is often a non-work activity (except for digital marketers!), whereas people use the Web for work and leisure. It seems likely that people do different searches during non-work times (presumably including more shopping) than at work times.
So Facebook users are not average. It’s plausible that Facebook is used disproportionally by young-ish women, who also do disproportionately more shopping searches.
Update on 13 September 2011: Confirmation! “The average social network user is most likely to be female, aged 18 to 34 … Active social network users were … 47 percent more likely to spend heavily on clothing, shoes, and accessories” - allfacebook.com based on a survey from Neilsen.
I am not surprised, in the slightest, to see above-average searches and clicks through to fashion retailers such as “topshop”, “boohoo”, “jack wills” and “river island”. As we’ve seen, Facebook appeals to much the same demographic as these companies, so this is much as I would expect, independent of any marketing.
When I fix the math, this effect goes away. There are 0 extra Website visits.
Fixing the math
Is there any effect? Let’s try and fix the maths. (See further below for the sources of the estimates.)
- Averaging the numbers for all the sites in the table, gives 133% increase in fashion searches after using Facebook
- Estimate women doing Google searches: normally=50%, after Fb=58.2%.
- Estimate women do 155% fashion searches compared to men
- Fashion searches after FB are increased by 103.5% because of the greater number of women.
- These brands mostly target young people.
- Estimate Internet users of all ages who visit Facebook regularly = 55%. But Internet users 18-29 = 86%.
- Guess young people are 2x as likely to search for brands that target them.
- Fashion searches after FB are increased by 114% because of the greater number of young people.
- 50% of employers block Facebook and 50% of people do not shop at work. This generates a correlation between Facebook and shopping at other times. I assume 1/3 of internet use is for work purposes.
- Fashion searches after FB are increased by another 114% because both are more common at non-work times.
- Uplift due to statistical errors = 103.5% * 114% * 114% = 134.5 %
(compared with a total gain of 133% in the reported figures).
Total uplift due to Facebook likes = 133% - 134.5% = approximately 0
I see no evidence here that 1 Facebook Fan equals Any Website Visits.
What does this Mean for Experian and Hitwise?
I think you need to address the issue identified above and ensure that you’re comparing like-with-like when you take the difference of the clicks for the Facebook and non-Facebook groups. I think this will mean that the value of a Facebook fan is much smaller than your initial estimate of 20 “visits”, but of course I have not seen the detailed figures. It may even be zero.
If you like, I would be happy to check any revisions to the math - Pete
Assumptions and Calculations.
NB: If you don’t like any of my assumptions, please provide recent evidence with a link to the source. Unsupported rhetoric will be ignored. Thanks in advance.
- Guestimate for the proportion of women doing typical Google queries: 50%.
- Estimate for proportion of women doing searches after Facebook: 58.2% (based on point 2 in the first list): 16.3/(11.7+16.3) *100% = 58.2%
- Estimate for male/female searches. Audience numbers for fashion are in the proportions 14.6 : 9.4, so assume searches are too. So number of fashion searches by a woman, compared to a man = 14.6/9.4 * 100% = 155.3%
- Calculated as “men equivalents”, fashion searches before Facebook =
50 (for men) + 50*1.553 (for woman) = 127
Calculated as “men equivalents”, fashion searches after Facebook =
41.8 (for men) + 58.2*1.553 (for woman) = 132%
So the uplift for fashion searches after Facebook, due to more women = 103.5%
- Topshop targets 16-35, boohoo is a direct competitor of “miss selfridge” which targets 18-30, jack wills targets students, River Island targets 18-30, and so on. The point is, these are “young” brands.
- Estimate Internet users of all ages who visit Facebook regularly by averaging the percentages by age group =(86+61+47+26)/4% = 55%.
- Guestimate young Internet users are 2x as likely to search for brands targeted at them. 86% of young Internet users visit FB regularly, so search uplift due to this is 100% + 25%*((86-55)/55) = 114%.
- More Than Half of Employers Now Block Twitter, Facebook, MySpace at work. Also 50% of people only do shopping at home. There is a correlation here. More people are forced/choose to shop in an environment where Facebook is also available.
- I can’t find any online figures for the balance of Internet use between leisure and work, so I’ll assume it is 2:1 and activity levels are constant across both.
- This means that 1/8 of the time people people are blocked from facebook and avoiding shopping, so the search uplift apparently associated with Facebook at other times = 100/(1-0.125)% = 114.2%.
Update on 29 June 2011: Fixed about ten typos and added a missing link to some evidence. Apologies to early readers! - Pete
Update on 18 August 2011. Good article that explains why facebook “fans” are not the same as real-world “fans”: What Is The Value Of A Facebook Fan? Zero! - Augie Ray