How much is a Stack Exchange reputation point worth, roughly, to the company?

Question

The Stack Exchange websites are run for profit, and they work by publishing questions and answers that are valuable by two measures: the number of clicks they gain, and how many points they get given by users. It's in the company's interest for the two measures to stay correlated; and while the first correlates with profit, the second correlates with reputation points because every time a contribution is marked as good its author gains reputation.

So I wondered whether the company uses a rough conversion rate between reputation points and US dollars, and if so what that rate is, or if they don't use a rate or don't want to reveal it then what the rate might reasonably be. Of course the relationship might not be linear, so for example if a 1k-rep user's work is worth $100, a 100k-rep user's work might be worth $20000 and Jon Skeet's 1.1M rep...well I don't know!

For comparison, here are some articles on the value of a "like" on other social media sites:

"What is the Real Value of the Facebook “Like”?"
"What is the value of a ‘Like?’"
"What’s the Value of a Like?" (Harvard Business Review - they cite a recent report showing that 80% of the US companies surveyed were "unable to quantify the value of their social media efforts", but they are looking at companies that advertise on Facebook rather than Facebook itself)

Edit
A quick and dirty calculation using

(estimated value of company, $500m or £400m) / (estimated total rep, 2 x 10⁹ points)

gives a figure of $250 or £200 per 1000 rep.

Answer 1

The following suggests that a user with a reputation score of 100,000 has done about £20,000 ($27,000) of work for the company.

Pending the investigation of the applicability of Bathsheba's excellent suggested model, I took a sample of 25 users of Stack Overflow and for each of them I looked at their reputation score and the number of "people reached". The sample comprised the authors of the five answers posted here prior to this one, along with the lowest-score users on each of pages 1-5, 51-55, 101-105 and 201-205 of the listing of SO users by rep.

A linear regression with the intercept set at 0 rep = 0 views gives a goodness of fit R² = 0.833 and a figure of 77 views per rep point. So if 1000 views are worth $1 to an advertiser, that's $77x per 1000 rep where x is the average number of pages a visitor views here.

R² jumps to 0.89 if two very high views-to-rep users are removed.

An estimate based on what's been written so far would be maybe $80-$400 (£60-£300) per 1000 rep, or $16-$80 (£12-£60) for a full 200-rep day. Since the estimate of the company's value at $500 million is based on the idea that it's going to be around for some time, and since the number of views of any much-viewed page is going to increase during that time, a reasonable estimate of the amount of value corresponded to by reputation should be in the higher part of that interval. I would suggest a ratio giving the following figures:

• a 200-rep day gives the company about £40 ($54).
• a rep score of 1000 is for about £200 ($270) worth of work
• a rep score of 100,000 is for about £20000 ($27000) worth of work

(I'd be grateful for some help making a data query for "people reached" against "reputation" points for a bigger sample.)

Answer 2

In the last round of funding, 2015, the Stack Exchange valuation was around $460 million [USD].

Total reputation in each of the top ten sites^* by number of users:

• Stack Overflow: 864,255,312
• Super User: 36,328,065
• Ask Ubuntu: 25,036,905
• Mathematics: 46,546,816
• Server Fault: 25,748,853
• Software Engineering: 22,078,489
• Unix & Linux: 17,511,360
• Ask Different: 10,203,160
• English Language and Usage: 15,221,216
• Android Enthusiasts: 6,027,366

So if we assume that reputation value is equal across all sites (it's not, but this is meant to be a quick assessment), and we ignore the fact that we're using 2017 reputation counts applied to a 2015 valuation, and we assume that the content is what makes the company valuable, and that the reputation roughly correlates to content value, then we can sum the reputation, divide by the valuation, and provide a rough ballpark estimate for reputation per USD:

1,068,957,542 / $460,000,000 = 2.32 rep/$

Approximately $0.43 per reputation point.

Of course there are many, many flaws to this method, but it's one way to assign arbitrary value to an arbitrary fake internet point system.

If, for instance, I attempted to sell my network account (all sites) to another person, with its 224k of total reputation, I couldn't command a price of $0.43 * 224,000 = $96,320^†.

If someone purchased Stack Exchange the valuation would be much greater than $460M [USD], however they wouldn't necessarily be buying the reputation of the users, but the content and the userbase itself. When Microsoft purchased Minecraft and LinkedIn they weren't buying a product, they were buying players and users - established customer bases and communities.

So the $0.43 is hugely inflated, however there really is no easy way to accurately gauge the value since there's no market. We can make up any sort of calculation that we want, but until reputation becomes something easily and widely traded then there's no real way to evaluate its actual value.

^*_{I'm ignoring meta.SE for the purposes of this post.}

^†_{Interested parties may inquire within.}

Answer 3

I wondered whether the company uses a rough conversion rate between reputation points and US dollars

No, we don't. We don't attach monetary value to reputation. Never did, never will.

It doesn't make any sense to do so.

In the same way that Twitter likes don't have any inherent monetary value for Twitter.

Answer 4

I have 100k reputation on Super User. Quite literally the only material benefit I have gotten from it is a mug, a T shirt and assorted swag.

SE doesn't make money directly off our answers, but indirectly off the 'reputation' of the site - as a good place to get answers, and leverages that for ancillary products, like careers and advertising.

So, your premise is probably flawed, and there's no way to set a dollar value on a user's reputation. This isn't even taking into account things like moderation activity or goodwill from other users.

Answer 5

The rough estimation of $250 or £200 per 1000 rep, or $250K or £200K for 100K rep, should be updated in view of the company's recent sale for $1.8bn. That figure is exactly 3.6 times the figure of $500m that was used.

So, using 100K rep as our unit, we can estimate that a user with 100K rep has done about $90K or £72K of work for the company.

Or to put it another way, if you've got 110K rep, you've done about 100k USD of work for the company - in some cases for the pleasure of being allowed to use the company's tools while working for it, plus the feeling of camaraderie with other workers.

This is based on a figure of 2 x 10^9 for the total number of reputation points. That has probably gone up, which will of course increase the number of reputation points that correspond to a million dollar's worth of work. I am not sure how to estimate the current total number of reputation points.

Edit
The above is one part of an outside view. As @Luuklag has rightly said in a comment, among the company's assets there is also the Teams platform which of course has a considerable financial value. It would be good if someone could come up with a reasonable lower bound of an outside view for the value of the work corresponding to 100K reputation points.

Answer 6

The business model of SO as a company is structured in such a way that its income isn't correlated with reputation, so you cannot attach a monetary value to reputation.

SO has two main ways of making money. The first is ads. The income money SE gets from ads is going to loosely correlate to the amount of views on the site (particularly from anonymous users). Views however, particularly views from anonymous users, isn't going to correlate to reputation. Lots of actions will earn people lots of reputation, while keeping a fairly low anonymous user count, while other posts will get enormous numbers of views, particularly views from anonymous users, without earning the site as much rep (proportionally). This is in large part because anonymous users can't vote, so their views don't generate rep.

Next there's the Jobs portion of the site. That generates money by matching people looking for jobs with people looking to hire. There's even less correlation here between the revenue generated and rep increases in the site, mostly as a result of the fact that such a small portion of the site's users actually use this feature of the site, but also because usage rates aren't going to correlate with the user's rep (if you only count "active" users).

Answer 7

My thought: it's not worth anything.

A comparison between currency and reputation makes little sense so therefore you can't think of the value of Stack Exchange being in "reputation". The mathematicians amongst us would not consider a substitution of reputation for currency as a valid change of measure. A couple of reasons in non-technical language:

1. Reputation is not-transferable (with the exception of the bounty "give") between parties so therefore cannot be used as fiat money.

2. Reputation is far too inflationary. Processes that deflate reputation are rare and include downvoting questions (-2), answers (-3), and question and user removal.

It's probable that the rate of change of total reputation is correlated with site activity, but that neglects the important facet of folk visiting the sites that don't vote. So I'd submit to you that the rate of change of reputation might be broadly linked to the value of Stack Exchange, not the reputation level itself. In other words, a starting point for a model might be

v(t) = a (dr(t) / dt) + b(t)

where v is the value, r the reputation, t time, a is a constant and b a function that might end up being too time-dependent for anything useful to be gleaned here. If b turns out to be practically constant or even linear in time then you have stumbled upon something truly brilliant. Not sure where you'd get the data from though to test this model; you can probably get r(t) from a fancy data query, and v(t) from the stock price and shares outstanding data. Good luck!