31

Giving What We Can is still growing at a surprisingly good pace

Giving What We Can (GWWC) used to have a graph on its site showing how its membership had grown over time. After a website redesign about a year ago it was removed for technical reasons and bringing it back hasn’t yet hit the top of the tech team’s list of priorities. As a result people haven’t been able to follow how GWWC is doing.

Well, I bring good news! Despite having much less direct attention paid to it than it used to receive, membership continues to grow at a healthy pace. In fact a surprisingly healthy pace (click for a higher res version):

Since CEA stopped focussing on growing GWWC’s membership in mid-January 2017, it has attracted an additional 1,152 members.

That compares to 1,528 over the same period before mid-January 2017. By this measure deprioritizing GWWC has only slowed the pace of membership growth by 25%.

We can look at the change in trajectory some other ways too. GWWC’s membership seemed to be growing faster than linearly up to mid-January 2018, but linearly since then. If we model the growth as a second order polynomial - remember your quadratics from high school - growth is down a third. If we model it as a third order polynomial - which I think is the best fit for the data up to January 2018 - it’s down about half. If we model it - somewhat ambitiously - as an exponential, it’s down more like 70%.

A few other observations:

  • We can see that it took a while from when GWWC hired its first staff member for this to pay off in higher membership growth. I was around working at GWWC for the first half of this, and it looks to me like a pretty classic case of it taking time to develop ‘product-market fit’ as Y Combinator calls it. It took time to learn i) what would encourage the sorts of people who would stick with the pledge to actually sign up, ii) what kind of research or other content would attract the kind of readers who want to take the pledge, iii) how to make best use of local GWWC groups, iv) plenty of other things besides.
  • The impact of the three holiday season pledge drives in 2014, 2015 and 2016 is very visible.
  • The period when a reorganised Centre for Effective Altruism as a whole was focussed on growing GWWC’s membership - the third holiday drive from October 2016 to January 2017 - was quite successful. It resulted in 80% more membership growth than the second best four-month period.

Growth not slowing down as much as as I would have predicted it would seems like evidence that previous efforts to grow GWWC’s membership were not having as much impact as I thought.

This doesn’t gel with my personal experience working there in 2012 and 2013.

Firstly, we can see what membership growth without any effort to draw attention to the pledge looks like in the very gradual increase over 2009-2012. This was generated by organic spread through social networks, occasional news stories, and some internet search traffic. Only 35 people joined in 2010 and 98 in 2011 - an 18th and a 6th of the rate today, respectively.

Secondly, the impact of the holiday campaigns in 2014, 15 and 16 speaks to the effectiveness of encouraging people to make an active decision on whether to join or not. Note that without an active campaign, there was no noticeable spike in pledges during 2012, 13 or 17, which shows that it was these efforts rather than some general 'holiday season effect'.

An alternative explanation is that GWWC’s larger membership base means it now spreads organically through social networks much faster than it did in the past.

The background rate of people finding out about GWWC through its website hasn’t much changed: since January 2017 the number of new users visiting the site has declined only 15%, despite there being little new content. This reflects the large fraction of people coming in through search traffic, which has been roughly stable.

Another alternative interpretation is that efforts to promote the ideas of effective altruism more broadly - without encouraging people to join GWWC specifically - may be creating GWWC members at quite a high rate, and higher than we saw back in the early days.

Whatever’s going on, don’t count GWWC out. Congrats to all the people who have helped to build it over the years, and welcome to all these new members!

--

While I'm at 80,000 Hours now, and worked at GWWC back in 2012 and 13, I threw this post together in a personal capacity.

Comments (6)

Comment author: tobyjolly 17 September 2018 12:36:42PM *  3 points [-]

Are there any estimates of GWWC attrition other than this one?

Linear growth in new members with a fixed attrition rate would result in active members curving towards a plateau (~12,000 assuming ~600 new members and ~5% attrition annually).

Comment author: Julia_Wise  (EA Profile) 17 September 2018 01:26:39PM 1 point [-]

We haven't updated attrition rates since then, no. We're hoping that having the donation recording system (pledge dashboard, formerly My Giving) integrated with the EA Funds login will yield us better data this year.

Comment author: PeterW 16 September 2018 07:50:46AM 0 points [-]

Thanks for this interesting post. I just thought I'd say it looks to my eye like there is actually an increase in the rate of gaining new members at the end of 2017 - so could there be evidence of a speed up caused by giving season after all?

Comment author: lincolnq 14 September 2018 09:30:33AM 3 points [-]

I think you're right that hiring your first staff/online signup put you onto the exponential curve. And you didn't fall off of the exponential until you de-emphasized it -- if you fit the exponential from 2014-2017 and extrapolate to today, you might hit something like 8000 members. So if you think staying on-curve seems plausible if you were to have continued working on it, I would guess that de-emphasizing growth still cost you users, even as you continued to grow linearly.

Comment author: Robert_Wiblin 15 September 2018 05:55:39PM *  1 point [-]

If you look at these graphs ending in January 2017 I think you'll agree that a polynomial of degree 3 (cubic) seems like the best fit: https://imgur.com/a/9SlFZd9 .

If that's right we would expect something like 5,000 members by now.

It occurs to me now that all of these trend-lines are a bit biased towards forecasting rapid growth, as they finish right at the end of the 2016 holiday campaign which absorbed substantial resources. This was the highest period of growth, and likely not sustainable. It might be more reasonable to put the end-date in ~April and then we can fit the trend-line to a less cyclical curve.

Comment author: Peter_Hurford  (EA Profile) 17 September 2018 03:48:56AM 3 points [-]

Luckily we have metrics that can evaluate goodness of fit and don't have to rely on our eyeballs. :)

More seriously, thanks a bunch for putting this together. I want to revisit the original growth metrics post sometime in January 2019 (I think Jan-Dec metrics are better than Aug-Jul metrics) and I'll definitely include this post.