# How can social networks easily generate an extra \$200 million per year

Here I focus on LinkedIn and how they can monetize their groups via charging a fee for email blasts, but the same applies to Google+, Twitter, Facebook etc. In short, LinkedIn alone could generate an extra \$50 million per year, thought the best implementation would probably involve LinkedIn outsourcing email blasts to a vendor such as MailChimp or VerticalResponse: it would probably mean that LinkedIn would earn only \$25 million a year, the vendor would earn \$25 million a year, but for LinkedIn, it would mean no more spam issues (and no more technical support, complaints), and email blasts totally outsourced and automated.

So how would this work?

LinkedIn has 1.9 million groups today. Most are very small, some are large (> 100,000 members), some are very large. On average, based on our estimates (we checked 20 randomly selected LinkedIn members to see how many groups they belong to - the average was 24), each LinkedIn member belongs to more than 20 groups. Right now, there are about 300 million members. Let's say, conservatively, that about 100 million members are active and accept to receive email blasts, about once a week from each group. In short, it means that LinkedIn could potentially monetize 100 million members x 20 groups per year. Assuming only one out of 10 groups would accept to pay a fee to LinkedIn (or its vendor) for email blasts, we are at 100 million users x 2 groups/user, in terms of monetizing capability.

We pay \$15,000 a year to our vendor (VerticalResponse) to send 20 million messages to 100,000 subscribers each year. Since we are a rather large client, let's assume that the aggregated fee for 100 small clients reaching out to 100,000 users (when combined together), would be like \$25,000 a year.

Interesting numbers

From this analysis, LinkedIn could generate \$25,000 x 100,000,000 (users) x 2 (two groups out of twenty per user, that accept to pay a fee) divided by 100,000 (our VerticalResponse subscriber base). That's \$50 million a year. My guess is that LinkedIn would have to share 50% with their vendor (VerticalResponse or MailChimp) unless they want to manage this whole thing themselves. The same applies to Google+, Facebook, Twitter and other social networks, thus my \$200 million figure in the subject line.

Interestingly, LinkedIn is about 1,000 times bigger than we are (in terms of members). Their revenue is about \$2 billion per year (2,000 times our revenue). Our community is a leading niche network that should command much higher ad revenue per pageview than LinkedIn (not sure what LinkedIn's margin is, ours is 80%). My guess is that LinkedIn is significantly overpriced (or we are underpriced, or a combination of both).

Note

Another potential source of revenue would be for LinkedIn to share the email addresses of members (of the groups that you own on LinkedIn) with group owners only, for a fee - except those from members who explicitly refuse to have their emaill address shared with third parties. Actually, they did that in the past, but it was free. It is clear that LinkedIn attaches a lot of value to email address of its members, their legal team is very active on this to prevent thieves from stealing these email addresses and reselling them..

Views: 1220

Comment

Join Data Science Central

Comment by Vincent Granville on April 3, 2014 at 7:00am

I've tested many LinkedIn blasts. One sent to 100k users typically generates between 200 and 800 clicks, in line with other newsletter performance, and justifying earning potentials.

You don't need to gather stats on millions of users to estimate average number of groups per user. If you think 20 is not a big enough sample,  maybe you should learn basic statistics and sampling theory. I used two different samples, each with 10 users. In both cases the average number of groups per user was above 20. I have more confidence in my numbers than in your link. And the purpose here is to show that approximations like mine, should be part of the decision process, along with more conventional data science. I use such approximations all the time for my business; sometimes real data is not available or flawed.

Also note that my 20 groups per user applies to the 30% most active, responsive users, as described in my post: it applies to 100 million (out of 300 million) LinkedIn users. The remaining 200 million subscribe to very few groups, probably well below 7 groups per user. So my number is not as far off as you think.

Comment by Swamy Narayana on April 3, 2014 at 4:44am
I think you underestimate the negative effect this would have on users. If I signed up for e-mails, but couldn't see emails from all my groups (because most didn't pay), I'd be forced into two avenues of consumption to get updates from what should be a single source (groups on LinkedIn). This is very jarring from a user perspective, and issues like this lead to loss of members and subscriptions, dropping the value for everyone.

I'd also guess that you overestimate the adoption rate. LinkedIn has a large percentage of its user base that is not very active. Monthly unique visitors is barely two-thirds of the total user base (http://expandedramblings.com/index.php/by-the-numbers-a-few-importa...), and from the time spent on the site stats I'd expect most of the active users are pretty casual. Also, there are fewer groups than business pages, and its unclear to me who you would charge for bulk messaging from groups... other than the direct sponsored groups (and I can't find stats on that) many of them would undoubtedly not be able to or inclined to pull the cash together to send e-mails.

According to the link I sent, the average number of groups per user is 7 -- less than 25% of the 24 you estimated (and shame on a data-science person for estimating something like that from a sample of only 20; that's a ridiculously underpowered sample). You'd have to account for multiple languages as well; and there's an adjustment that needs to be made for users who were willing to sign up for e-mails but only from groups that wouldn't be willing to pay for them -- that would be non-zero and potentially large given your 2-in-20 assumption.

In fact, most of your numbers are just assumptions with no explained basis. This isn't remotely data scientific. Shame on you!