Effective Altruism Forum web traffic from Google Analytics

Note: If you are us­ing uBlock you may not be able to see the images, be­cause they have ”/​google-an­a­lyt­ics” in the name and uBlock is a bit … para­noid … about block­ing file names with that sub­string. Please dis­able uBlock for effec­tive-al­tru­ism.com or tem­porar­ily dis­able it to see images.

As 2016 comes to a close, I thought it’d be in­ter­est­ing to look at traf­fic trends on the Effec­tive Altru­ism Fo­rum (which I’ll call EAF for this ar­ti­cle) since its launch around Septem­ber 7, 2014. Some key take­aways fol­low, af­ter which I dis­cuss more of the data.

Key takeaways

  • An­nual cy­cle and long-term growth trends (more):

    • De­cem­ber and Jan­uary see the high­est traf­fic in the year, as ex­pected con­sid­er­ing it is Giv­ing Sea­son (though the rel­a­tive role of de­mand for con­tent and sup­ply of con­tent is un­clear). The sum­mer months see the low­est traf­fic. This sea­sonal fluc­tu­a­tion is more sig­nifi­cant than year-over-year growth.

    • Over­all, year-over-year growth in to­tal pageviews has been about 10% per year. It is swamped by other sources of fluc­tu­a­tion (so for a given month it could even be nega­tive, and for some months it could be as high as 30%). From late 2014 to late 2015, there was a sig­nifi­cant in­crease in ses­sions ac­com­panied by a de­crease in pages/​ses­sion that roughly can­celed. In 2016, ses­sions and pages/​ses­sion have re­mained roughly con­stant.

    • The year-over-year growth in unique vis­i­tors is com­pa­rable to that of GiveWell’s web­site, but much less than that of 80,000 Hours. GiveWell and 80,000 Hours are the only two or­ga­ni­za­tions I know of that cater to a similar au­di­ence and re­lease web traf­fic data.

  • Im­pact of EA events (more): EA Global and other similar events have neg­ligible traf­fic im­pact, whether we look at im­pact dur­ing the event, shortly be­fore or af­ter the event, or long-term be­fore ver­sus af­ter changes.

  • Post traf­fic life­cy­cle (norms and ex­cep­tions) (more):

    • The ma­jor­ity of posts get al­most all their traf­fic in the first few days af­ter they are pub­lished. The traf­fic may come from so­cial me­dia and other refer­rals, or peo­ple dis­cov­er­ing the con­tent from the home page.

    • A few pieces of con­tent get sus­tained traf­fic over a long pe­riod of time. This could be steady traf­fic or traf­fic re­sult­ing from mul­ti­ple traf­fic spikes. Some of these pieces are generic EA clas­sics, and oth­ers are about tan­gen­tially re­lated sub­jects of wider in­ter­est. No ar­ti­cles ex­clu­sively about EA or­ga­ni­za­tions or per­sonal dona­tion de­ci­sions have made it to this list.

    • The size of the ini­tial traf­fic to a post cor­re­lates far more closely with other in­di­ca­tors of post trac­tion (such as com­ments and net up­votes on the post) than the long-term traf­fic to the post. On the other hand, Face­book likes + com­ments + shares cor­re­late more with long-term traf­fic than ini­tial traf­fic, though I have less con­fi­dence about that con­clu­sion (due to more limited data).

I close with some spec­u­la­tive thoughts on im­pli­ca­tions.

Some ad­di­tional graphs and data (that weren’t im­por­tant enough to make it to the main post) are in the Mis­cel­lanea sec­tion.

Here’s a graph of monthly pageviews in Google An­a­lyt­ics. Note that since the site was launched af­ter a few days of Septem­ber 2014 were already over, the Septem­ber 2014 num­ber rep­re­sents a pe­riod shorter than a month.

You can see that De­cem­ber and Jan­uary have higher traf­fic than the sur­round­ing months for the end of 2014 end-of-year Giv­ing Sea­son. For the 2015 end-of-year Giv­ing Sea­son, Fe­bru­ary (so Fe­bru­ary 2016) ac­tu­ally has slightly higher traf­fic than Jan­uary, but you can still see the same qual­i­ta­tive pic­ture: a sharp rise in traf­fic in De­cem­ber, a de­cline in Jan­uary, and a re­turn to nor­malcy soon. For the 2016 end-of-year Giv­ing Sea­son, we only have De­cem­ber data, and we see a pretty huge rise from Novem­ber to De­cem­ber.

An year-over-year com­par­i­son be­tween EAF’s first and sec­ond year (Septem­ber 7, 2014 to Septem­ber 6, 2015, com­pared with Septem­ber 7, 2015 to Septem­ber 6, 2016) shows a 9.86% in­crease in pageviews (from 314,789 to 345,832), a 21.53% in­crease in ses­sions, and a 26.97% in­crease in users. Pages/​ses­sion and av­er­age ses­sion du­ra­tion fell.

Look­ing at some spe­cific months shows how noise over­whelms this slight long-term in­crease. For in­stance, in Septem­ber 2014 (the month of launch, where a few days were miss­ing since those were pre-launch), there were 26,256 pageviews. The cor­re­spond­ing num­bers in Septem­ber 2015 and Septem­ber 2016 were 30,467 (+16.0% year-over-year) and 26,347 (-14.5% year-over-year). For Oc­to­ber, the num­bers in 2014, 2015, and 2016 were re­spec­tively 25,237, 26,259 (+4.0% year-over-year), and 32,880 (+25.2% year-over-year).

With that said, there is clearer ev­i­dence of year-over-year growth for Novem­ber and De­cem­ber. As a re­sult, the gap be­tween Giv­ing Sea­son and other months has been in­creas­ing from 2014 to 2015 to 2016.

Data is very limited but one ten­ta­tive the­ory that fits the ev­i­dence so far is that the growth of the EA move­ment has been strongest among read­ers whose in­ter­est in EAF con­tent is high only dur­ing Giv­ing Sea­son. Another is that con­tent cre­ators are fo­cus­ing their con­tent cre­ation and pub­lish­ing efforts more dur­ing Giv­ing Sea­son and less dur­ing other months, and this is in­creas­ing the traf­fic differ­ence be­tween Giv­ing Sea­son and other months.

One in­ter­est­ing trend be­tween late 2014 and late 2015: al­though the in­crease in num­ber of pageviews was small, ses­sions in­creased quite a bit. Pages/​ses­sion dropped to com­pen­sate. This could be be­cause peo­ple be­gan us­ing EAF more on their phones (thus in­creas­ing the num­ber of ses­sions and users recorded by GA). It could be be­cause the num­ber of peo­ple ac­cess­ing EAF in­creased, but the in­ter­est and ded­i­ca­tion of the early adopters de­clined.

I also com­pared this growth rate against that of other web­sites that cater to similar au­di­ences. I was limited to the ones that have re­leased some pub­lic data, namely, 80,000 Hours and GiveWell. Here’s the com­par­i­son:

  • 80,000 Hours does not re­lease pageview data but it does re­lease an es­ti­mate of unique vis­i­tors to the site. You can get the num­bers here. The num­ber grew 244% year-over-year from 149,164 in 2014 to 513,697 in 2015, and then again by 62% to 834,310 in 2016. The num­ber for 2016 is an un­der­es­ti­mate since it only in­cludes data till Novem­ber, when the blog post was pre­pared. The cor­re­spond­ing in­crease in vis­i­tors from 2015 to 2016 for the Effec­tive Altru­ism Fo­rum was 15.6% from 66,384 to 76,712. Thus, 80,000 Hours had more unique vis­i­tors and a faster growth rate.

  • GiveWell’s web traf­fic num­bers. The lat­est num­bers re­leased by GiveWell are in this spread­sheet and go up to June 2016. You can also see the PDF re­port that goes up to the end of 2015 (in­clud­ing Jan­uary 2016, since GiveWell takes the year as start­ing in Fe­bru­ary and end­ing in Jan­uary). The re­port shows an year-over-year in­crease of 17% from 737,592 in 2014 to 863,883 in 2015. Those are about ten times the num­bers we see for EAF, but a similar growth rate.

Im­pact of EA events

The main story re­gard­ing EA events is the non-story: there is no in­crease in traf­fic to EAF in or around EA Global or ma­jor events. I got a list of events from Effects of ma­jor events on EA ac­tivity by Eric Yu. His post didn’t dis­cuss effects on EAF traf­fic, al­though the origi­nal post by Peter Hur­ford that his post an­swered had asked for that data. So I thought this would be a good set of events to look at:

  • GWWC pledge drive 2014–2015: a few weeks around 1/​1/​2015: Traf­fic was lower in Jan­uary than De­cem­ber, but it’s hard to iso­late the effects of the GWWC Pledge Drive from the gen­eral Giv­ing Sea­son cy­cle.

  • GWWC pledge drive 2015–2016: a few weeks around 1/​1/​2016: Same com­ment as for the pre­vi­ous pledge drive.

  • EA Global 2015: 7/​31-8/​2, 2015 (SF), 8/​14-8/​16 (Melbourne), 8/​28-8/​30 (Oxford): Traf­fic was lower dur­ing the days of the sum­mits than the sur­round­ing days. This could be due to it be­ing a week­end. It could be be­cause the in-per­son event sub­sti­tuted for web­site vis­its and re­duced the amount of con­tent be­ing posted. Over­all, traf­fic in Au­gust was the same as July, so it doesn’t seem like EA Global in­creased traf­fic to EAF in the sur­round­ing days. More­over, most of the pages that were vis­ited dur­ing these events were qual­i­ta­tively similar to the pages that were vis­ited be­fore and af­ter them. The main ex­cep­tion was the Au­gust open thread, that got a to­tal of 304 views, most of them in the days right af­ter the event ended. The view count for this was too small to mean­ingfully af­fect over­all traf­fic.

  • Do­ing Good Bet­ter pub­lished: 7/​28/​2015 (around the same time as EA Global): Traf­fic was higher on this day but the qual­i­ta­tive pat­tern of posts viewed wasn’t differ­ent from other days, sug­gest­ing that the pub­li­ca­tion of the book wasn’t the driv­ing fac­tor for the traf­fic in­crease.

  • The Most Good You Can Do pub­lished: 3/​15/​2015 (pub­lished on Ama­zon 4/​7/​2015): Traf­fic moved up and down around this time and there wasn’t a clear effect on over­all traf­fic. The qual­i­ta­tive na­ture of ar­ti­cles vis­ited also didn’t change. Note that this is in con­trast with the Wikipe­dia page for effec­tive al­tru­ism, where the pub­li­ca­tion of the book ap­prox­i­mately marked a per­ma­nent in­crease in the traf­fic level to the ar­ti­cle.

  • GiveWell top char­i­ties an­nounced (2015–2016): 11/​20/​2015 (cor­rected date from source): Also hard to iso­late the effect of this rel­a­tive to Giv­ing Sea­son changes. Look­ing at traf­fic to posts with GiveWell in the ti­tle, we see spikes at var­i­ous points in time re­lated to posts about GiveWell, but these don’t cor­re­spond to the timing of GiveWell’s recom­men­da­tion re­lease.

Post traf­fic life­cy­cle: norms and exceptions

The vast ma­jor­ity of EAF posts get the vast ma­jor­ity of their traf­fic within a few days of pub­li­ca­tion, and all the as­so­ci­ated en­gage­ment ac­tivity (post up­votes and com­ments) also oc­curs within that time. This time can vary from a day to a lit­tle over a week. Typ­i­cal ex­am­ples:

The fact that most posts have short lifes­pans is part of the ex­pla­na­tion for why, de­spite the cu­mu­la­tive con­tent on EAF be­ing much much more now than it was dur­ing launch in Septem­ber 2014, to­tal pageviews aren’t up that much. If the posts had longer lifes­pans, and con­tinued to get steady trick­les of views, these steady trick­les would add up to a lot of to­tal pageviews, and that num­ber would in­crease as the to­tal num­ber of pub­lished posts went up. How­ever, since posts typ­i­cally die quickly af­ter be­ing pub­lished, we don’t see the steady ac­cu­mu­la­tion of pageviews. So pageviews in a given month are de­ter­mined more by new con­tent cre­ated at the mar­gin rather than the to­tal cor­pus of con­tent so far.

How­ever, the top viewed items of all time dis­play a differ­ent pat­tern. Like the other posts, they have a spike around the time of ini­tial pub­li­ca­tion (or ini­tial im­port into EAF), and much of the en­gage­ment ac­tivity (up­votes, com­ments) oc­curs within that time. How­ever, un­like other posts, the top viewed items con­tinue to get a steady trickle of views later in life, or ad­di­tional later spikes in traf­fic, long af­ter ac­tive en­gage­ment with them (in the form of com­ments and up­votes) has died down.

Another note­wor­thy fea­ture of the top posts is that many of them were im­ported into EAF, or cross-posted to EAF, from other places (such as LessWrong, or the old Effec­tive Altru­ism blog that was hosted on the same do­main).

The situ­a­tion with Face­book likes + com­ments + shares seems similar to long-term traf­fic (and there­fore differ­ent from com­ments and up­votes on EAF): posts that con­tinue to get a steady trickle of views also con­tinue to get a steady trickle of Face­book likes, com­ments, and shares. How­ever, I don’t have good time se­ries data on Face­book likes, com­ments, and shares, so this is just a rough es­ti­mate based on what i do have.

Here are the top five posts:

  • In­tro­duc­tion to Effec­tive Altru­ism (20 net up­votes, 5 com­ments, 292 Face­book likes + com­ments + shares), av­er­aged about 50 pageviews a day from mid-2015 to Au­gust 2016, and then dropped to about 10 pageviews a day ((prob­a­bly due to com­pe­ti­tion from In­tro­duc­tion to Effec­tive Altru­ism on effec­tivealtru­ism.org).

  • The Drown­ing Child and the Ex­pand­ing Cir­cle (8 net up­votes, 8 com­ments, 138 Face­book likes + com­ments + shares) sees traf­fic in waves, with sud­den spikes af­ter some sort of me­dia pub­lic­ity and then grad­ual de­cay over time, till the next spike.

  • What is Effec­tive Altru­ism? (7 net up­votes, 7 com­ments, 16 Face­book likes + com­ments + shares) had a traf­fic spike in July 2015, af­ter which it saw 20 pageviews a day for a while. The num­ber has de­cayed grad­u­ally and is now about 5 a day. Search query com­pe­ti­tion could ex­plain the de­cline.

  • Effi­cient char­ity: do unto oth­ers… (13 net up­votes, 5 com­ments, 87 Face­book likes + com­ments + shares) spiked in late Jan­uary 2016, and has since fluc­tu­ated, av­er­ag­ing be­tween 2 and 15 pageviews a day.

  • Should you donate to the Wiki­me­dia Foun­da­tion? (10 net up­votes, 3 com­ments, 126 Face­book likes + com­ments + shares) had three spikes: one at the time of ini­tial writ­ing (in late March 2015), and two more dur­ing the Giv­ing Sea­sons of 2015 and 2016, co­in­cid­ing with the Wiki­me­dia Foun­da­tion fundraisers through Wikipe­dia ban­ners. Dis­clo­sure: I wrote this post, and have spent some time un­der­stand­ing its view statis­tics.

Beyond the top five, the pic­ture be­comes more mixed. The sixth and sev­enth ar­ti­cle, namely Philo­soph­i­cal Cri­tiques of Effec­tive Altru­ism by Prof Jeff McMa­han and Cheer­fully, had only one spike at time of pub­li­ca­tion, but do still get a very slow but still steady trickle of traf­fic. In eighth place is the wide-in­ter­est ar­ti­cle Six Ways To Get Along With Peo­ple Who Are To­tally Wrong*. This has had three big spikes and a much much smaller steady stream of views.

The most-com­mented ar­ti­cle I could find was Con­cerns with In­ten­tional In­sights, with 39 net up­votes and 182 com­ments, and 60 Face­book likes + com­ments + shares. This placed ninth in terms of life­time pageviews. The traf­fic to this ar­ti­cle had a sin­gle, re­ally huge spike at the time of launch, and has since been pretty low.

Another way of try­ing to gauge the rel­a­tive im­por­tance of steady views to top-perform­ing con­tent ver­sus views of new con­tent is to com­pare to­tal pageviews against pageviews of the top five ar­ti­cles of all time over time. I have in­cluded a graph be­low with this data. All time, the top five ar­ti­cles ac­count for 52,872 pageviews out of 467,380. In De­cem­ber 2016 (af­ter any of them was pub­lished, so it misses the ini­tial peak of all five) the top five ac­counted for 2,703 out of 28,284 views. Over­all, then, the top five of all time ac­count for some­thing like 10% of pageviews in a given month.

Spec­u­la­tive thoughts on implications

The slow in­crease in to­tal traf­fic to EAF sug­gests some tem­per­ing of the wide­spread nar­ra­tive of rapid EA move­ment growth. It would be in­ter­est­ing to un­der­stand why EAF traf­fic lags other widely cited in­di­ca­tors of move­ment growth, such as GWWC pledge sign-ups and GiveWell money moved. A num­ber of plau­si­ble the­o­ries can be pro­posed. For in­stance, it could be that EAF ap­peals only to a seg­ment of EAs (and EA-penum­bra peo­ple) and this seg­ment is not rep­re­sented well in new re­cruits. On the other hand, it could be that EAF has huge churn, so that even though the EA move­ment is grow­ing, EAF mem­ber­ship at any given time isn’t grow­ing that fast be­cause peo­ple leave quickly. It’s also pos­si­ble that the prob­lems are on the sup­ply side (not enough qual­ity con­tent on EAF, or more com­pe­ti­tion for read­ers’ at­ten­tion).

More in­ter­est­ing (to me) than the over­all traf­fic trend is the ques­tion of what the pageviews and other met­rics for spe­cific posts can tell us about the im­pact of those posts. Con­tent cre­ators get feed­back on posts based on ini­tial trac­tion in the form of com­ments and up­votes, which cor­re­late well with ini­tial pageviews. But for the best-perform­ing posts in the long term, a lot of the pageviews hap­pen over time, and the long-term pageview trend is not as cor­re­lated with the ini­tial spike in pageviews as we might think.

One pos­si­ble im­pli­ca­tion is that con­tent cre­ators on EAF may be over­pro­duc­ing con­tent that gets a lot of ini­tial trac­tion, at the ex­pense of cre­at­ing ev­er­green con­tent with long-term value. I mean this more in the sense that peo­ple are re­luc­tant to pro­duce ev­er­green con­tent be­cause they don’t see it get­ting that much trac­tion, rather than that they are cyn­i­cally gen­er­at­ing con­tent of short-term value to in­crease their karma. Of course, this im­pli­ca­tion is far from proven, as it’s not clear that long-term pageviews are a good in­di­ca­tor of long-term value. But even if it’s not, analo­gous rea­son­ing might ap­ply for long-term value.

Take a look at the top five posts. Four of them are generic, ev­er­green con­tent whose rele­vance de­clines very slowly over time. The fifth one (on donat­ing to the Wiki­me­dia Foun­da­tion) is more spe­cific, but the topic it touches on, namely dona­tions to Wikipe­dia, is some­thing that a siz­able frac­tion of the In­ter­net-us­ing pop­u­la­tion comes across dur­ing Giv­ing Sea­son. Con­trast this with a post like Con­cerns with In­ten­tional In­sights that gets a lot of in­ter­est when it is pub­lished but does not ad­dress a re­cur­ring, long-term hu­man need, and there­fore dies down quickly. Note that it’s still pos­si­ble (and many have ar­gued) that the In­ten­tional In­sights post did have a huge in­di­rect long-term im­pact not re­flected in re­cur­ring pageviews, by chang­ing norms around fundrais­ing and ac­countabil­ity that would serve as cau­tion­ary les­sons for fu­ture donors and or­ga­ni­za­tions. My point of us­ing it as an ex­am­ple was mostly to sim­ply high­light it as a case where di­rect in­ter­est (as mea­sured by pageviews) dropped sharply af­ter an ini­tial pe­riod. There is also a pos­si­bil­ity that the post will see an­other traf­fic surge when similar is­sues crop up in eval­u­at­ing other or­ga­ni­za­tions, just like the Wiki­me­dia Foun­da­tion post has had traf­fic surges trig­gered by Wikipe­dia ban­ner fundrais­ing.

Another im­pli­ca­tion is around the way we link to older ma­te­rial. A lot of con­tent is high-qual­ity, but fails to get vis­ited af­ter a huge amount of traf­fic in the first week. Reg­u­lars for­get about the con­tent, new­com­ers never hear of it. Bet­ter, stronger link­ing prac­tices to older con­tent, and more high-qual­ity fol­lowup posts that help re­vive older con­tent, could help ad­dress this. Dis­cus­sions and de­bates ad­vance when peo­ple build on top of past ma­te­rial, rather than re­gen­er­ate or re­hash it, so a strong mem­ory of past con­tent, re­in­forced through ex­ten­sive link­ing to it, can help.

Miscellanea

Prior to the Effec­tive Altru­ism Fo­rum, the same do­main name hosted an in­vite-only blog that had a hand­ful of posts. Th­ese ports were ported over to EAF when it started (along with some more seed con­tent). Below is the Google An­a­lyt­ics for the pre-EAF site. You can see it was sig­nifi­cantly less than traf­fic to EAF. Data is available only start­ing Oc­to­ber 2013. The drop in Septem­ber 2014 co­in­cides with the launch of EAF dur­ing the month.

Below, I paste the pageview data for the new Effec­tive Altru­ism Fo­rum at daily and weekly gran­u­lar­ity re­spec­tively for the pe­riod Septem­ber 1, 2014 to De­cem­ber 31, 2016, and then the hourly pageview data for De­cem­ber 2016. By look­ing at this data you can ver­ify that there is a clear daily cy­cle in traf­fic, but no weekly cy­cle.

Data exports

Although I have Read&An­a­lyze ac­cess to the EAF’s Google An­a­lyt­ics, I don’t have the au­thor­ity to grant oth­ers ac­cess. You can con­tact Ryan Carey to give you Read&An­a­lyze ac­cess if you are in­ter­ested.

If there are spe­cific EAF posts that you think are im­por­tant enough to have their traf­fic pat­terns cov­ered in this post, please list them in the com­ments and I’ll add their traf­fic pat­terns to the post when I get time.

Re­lated read­ing on move­ment size and growth met­rics for EAF and re­lated web­sites and com­mu­ni­ties:

Public data on Effec­tive Altru­ism Fo­rum web traf­fic:

  • Alexa: Whereas Google An­a­lyt­ics uses di­rect, first-party mea­sure­ment of traf­fic, Alexa sim­ply pro­vides an es­ti­mate based on traf­fic that it is able to track for a sub­set of users. There­fore it is less ac­cu­rate.

  • SimilarWeb Like Alex, this uses in­di­rect, third-party mea­sure­ment by track­ing a sub­set of users. There­fore it is less ac­cu­rate. Un­like Alexa, it shows es­ti­mates of ac­tual ses­sion and pageview counts rather than just a rank es­ti­mate.

  • HypeS­tat: This com­bines data from var­i­ous sources such as Alexa. It is fairly in­ac­cu­rate and pro­vides only a loose bal­l­park.

More nor­ma­tive dis­cus­sions of EA move­ment growth:

Re­lated read­ing on web­site traf­fic vari­a­tion and un­der­stand­ing web­site au­di­ences:

  • Un­der­stand Your Web­site Traf­fic Vari­a­tion with Time, on wik­iHow. This ex­plains daily, weekly, and an­nual cy­cles in web traf­fic, as well as long-term trends and how to un­der­stand them. There is a sec­tion on the an­nual cy­cle for con­tent re­lated to philan­thropy and char­ity. I’ve con­tributed most of the con­tent for this ar­ti­cle.

  • Un­der­stand Your Web­site Au­di­ence Pro­file on wik­iHow. This dis­cusses gen­eral facts about gen­der, age, and lo­ca­tion dis­tri­bu­tion for web­site au­di­ences. I did not cover the as­so­ci­ated anal­y­sis in this post, as I wanted to fo­cus on con­tent and to­tal traf­fic rather than the de­mo­graphic char­ac­ter­is­tics of the au­di­ence. I might do a fol­lowup post ap­ply­ing the ideas there to EAF, if there’s suffi­cient in­ter­est in and pos­i­tive feed­back for this post.

  • Get Traf­fic An­a­lyt­ics for a Web­site You Don’t Own on wik­iHow. This ex­plains ways to es­ti­mate traf­fic num­bers for a web­site that’s not owned by you. It might help you get more up-to-date data for the Effec­tive Altru­ism Fo­rum and for other sites.

UPDATE June 17, 2017: Start­ing around late April 2017, I no longer have ac­cess to the Effec­tive Altru­ism Fo­rum Google An­a­lyt­ics data. Thus, I will not be able to up­date the ex­ist­ing ex­ports or pub­lish fol­low-up posts on the Fo­rum’s an­a­lyt­ics.