EA Giving Tuesday Donation Matching Initiative 2018 Retrospective

Avi Norow­itz, William Kiely, Anisha Zaveri, and Arushi Gupta, were on the or­ga­nizer team for this pro­ject, and many oth­ers also pro­vided valuable con­tri­bu­tions. See “Pro­ject con­trib­u­tors” at the end of this post for more in­for­ma­tion.


  • In 2018, Face­book and PayPal an­nounced plans to match up to $7 mil­lion in dona­tions on Giv­ing Tues­day. This rep­re­sented an un­usual op­por­tu­nity to get EA dona­tions coun­ter­fac­tu­ally matched.

  • We made a co­or­di­nated effort to di­rect match­ing funds to EA-al­igned or­ga­ni­za­tions, with a fo­cus on donat­ing fast and start­ing prepa­ra­tion work early. Our efforts were suc­cess­ful: We had $469k (65%) of our $719k dona­tions matched, rep­re­sent­ing a 10x in­crease from our co­or­di­nated efforts in 2017.

  • We’re cur­rently in­volved in fol­low-up work with or­ga­ni­za­tions to help them en­sure they re­ceive the ex­pected amounts, and that the amounts are al­lo­cated cor­rectly.

  • We also con­ducted a fol­low-up sur­vey with donors, which pro­vided some in­sight on pay­ment prob­lems and the ex­pe­rience of donors out­side the US. The re­sponses also sug­gest that $85k (12%) of dona­tions may have been coun­ter­fac­tu­ally caused by our ini­ti­a­tive, though this es­ti­mate is highly un­cer­tain.

  • I dis­cuss how we im­proved since 2017, in­clud­ing: fo­cus­ing on donat­ing fast, start­ing early, fo­cus­ing on pay­ment prob­lems, im­prov­ing out­reach, im­prov­ing co­or­di­na­tion with or­ga­ni­za­tions, and im­prov­ing data col­lec­tion and an­a­lyt­ics.

  • I also dis­cuss ar­eas we could im­prove on in 2019, in­clud­ing: im­prov­ing com­mu­ni­ca­tion with donors, prepar­ing in­struc­tions ear­lier, helping or­ga­ni­za­tions fundraise, in­vest­ing more in tech­nol­ogy, and in­cor­po­rat­ing dona­tion sur­vey re­sults.

  • The ex­pe­rienced fur­ther con­firms to me that EAs can fol­low in­struc­tions and work well as a com­mu­nity. It also fur­ther con­firms to me that effec­tive co­or­di­na­tion with key or­ga­ni­za­tions can yield large benefits.

  • I dis­cuss some un­ex­pected prob­lems we dis­cov­ered that could have led to sub­stan­tially less match­ing funds, and how we avoided these prob­lems.

  • I dis­cuss a few ques­tions for next year, in­clud­ing: how much time we’ll have be­fore the match runs out, what prob­lems we might face, and how open we should be about our work.

  • I dis­cuss our pro­ject costs, which con­sisted of an es­ti­mated 330 hours con­tributed among our or­ga­nizer team, and an un­known num­ber of hours by other con­trib­u­tors.

  • I con­clude with a list of pro­ject con­trib­u­tors and some of their con­tri­bu­tions.

Face­book’s Giv­ing Tues­day match­ing program

In early Oc­to­ber 2018, Face­book an­nounced that Face­book and PayPal would match dona­tions made on Face­book on Giv­ing Tues­day (Novem­ber 27, 2018) up to $7 mil­lion. Face­book later pro­vided the fol­low­ing de­tails on the match:

In sup­port of #Giv­ingTues­day 2018, we’re part­ner­ing with PayPal to match dona­tions made on Face­book to non­prof­its up to a to­tal of $7 mil­lion. Start a fundraiser for Giv­ing Tues­day or donate to a non­profit.
About the match
- Face­book and PayPal will start match­ing dona­tions at 8:00am ET (5:00am PT) on Tues­day, Novem­ber 27, 2018. The match­ing offer will ex­pire when $7 mil­lion in dona­tions is reached or at 11:59pm PT on Novem­ber 27.
- Dona­tions will be matched dol­lar for dol­lar on a first-come, first-served ba­sis.

- Match­ing is available for any el­i­gible US-based 501(c)(3) non­profit that can re­ceive dona­tions on Face­book.

- Dona­tions up to $250k per non­profit and $20k per donor are el­i­gible to be matched.

In ad­di­tion to the limits de­scribed in the an­nounce­ment, we de­ter­mined that the max­i­mum amount per­mit­ted per dona­tion in the US was $2,500, so 8 dona­tions were re­quired to reach $20,000.

Face­book ran similar match­ing pro­grams in 2016 for $900,000, and in 2017 for $2 mil­lion. In 2017 we at­tempted to di­rect match­ing funds to EA-al­igned or­ga­ni­za­tions. Un­for­tu­nately, be­cause we had less than a week to pre­pare, and be­cause the match­ing funds ran out much faster than we had planned for (at 86 sec­onds), only 13% of our to­tal dona­tion amount was matched.

To learn more about the Face­book match­ing pro­gram and how 2018 com­pared to pre­vi­ous years, see our FAQ.

The op­por­tu­nity for EAs

Face­book’s an­nual dona­tion match on Giv­ing Tues­day rep­re­sents a rare op­por­tu­nity to get dona­tions to EA-al­igned or­ga­ni­za­tions coun­ter­fac­tu­ally matched, similar to US em­ployer match­ing pro­grams that match em­ployee dona­tions to any US reg­istered non­profit or­ga­ni­za­tion. This op­por­tu­nity is differ­ent than most match­ing op­por­tu­ni­ties in the fol­low­ing re­spects:

  • The match­ing funds are available to nearly any US non­profit or­ga­ni­za­tion reg­istered as a 501(c)(3). This per­mits dona­tions to EA-al­igned or­ga­ni­za­tions, and re­duces the risk that dona­tion de­ci­sions are dis­torted to­wards less effec­tive or­ga­ni­za­tions be­cause of the match.

  • Match­ing funds that are not di­rected to EA-al­igned or­ga­ni­za­tions oth­er­wise go to or­ga­ni­za­tions of more av­er­age effec­tive­ness. This con­trasts with most dona­tion match­ing op­por­tu­ni­ties, where it’s likely that the match­ing part­ner would have made the same dona­tion to the same or­ga­ni­za­tion any­way.

We see this dona­tion-match­ing op­por­tu­nity as one that avoids most of the pit­falls de­scribed by this GiveWell blog post on dona­tion match­ing.

To learn more about the op­por­tu­nity for EAs, see our FAQ.

Our strat­egy to di­rect the match­ing funds

In 2017, the match­ing funds ran out in 86 sec­onds, which was much faster than most of us had an­ti­ci­pated. As a re­sult, we es­ti­mate that only 13% ($48k) of our to­tal dona­tion amount ($379k) was matched. We learned from this ex­pe­rience that suc­cess in 2018 would likely re­quire us to donate fast. As such, we em­pha­sized that donors should try to donate within as few sec­onds as pos­si­ble af­ter the match be­gins. We also pro­vided ba­sic in­struc­tions, full in­struc­tions, and demon­stra­tion videos on donat­ing fast.

We also wanted to give donors an easy way to donate to a large va­ri­ety of EA-al­igned or­ga­ni­za­tions, and we wanted to be able to col­lect data on dona­tions and matches to these or­ga­ni­za­tions. To fa­cil­i­tate this, we cre­ated one Face­book fundraiser for each or­ga­ni­za­tion and added them to a spread­sheet.

Be­cause only US reg­istered 501(c)(3)s non­profit or­ga­ni­za­tions were el­i­gible, we faced the risk that a large num­ber of im­por­tant or­ga­ni­za­tions, pro­grams, and funds would re­main out of reach for donors. This could have re­duced the amount of donors el­i­gible to par­ti­ci­pate, or could have in­fluenced donors to donate less effec­tively for the match. Thank­fully, we were able to work with CEA, ACE, and GiveWell to provide donors the op­por­tu­nity to donate to the ma­jor­ity of these difficult-to-reach or­ga­ni­za­tions, pro­grams, and funds through Face­book on Giv­ing Tues­day.

In 2017, we be­came aware of the op­por­tu­nity less than a week in ad­vance, which didn’t give us much time to pre­pare. In 2018 we started our prepa­ra­tion around 2 months in ad­vance. This gave us the op­por­tu­nity to care­fully plan our strat­egy, give or­ga­ni­za­tions a heads up, ar­range a plan to get dona­tions and matches to difficult-to-reach or­ga­ni­za­tions, pre­pare good ma­te­ri­als, and work on donor out­reach.

To learn more about our strat­egy, see our FAQ.

Pledge and dona­tion amounts

By the Satur­day morn­ing be­fore Giv­ing Tues­day, we had $200k in pledges. By the time the match be­gan on Tues­day morn­ing, this had in­creased to $614k in pledges from 160 in­di­vi­d­u­als.

After our fundraisers ended at mid­night on Wed­nes­day, we had a to­tal of $719k in dona­tions. The vast ma­jor­ity of these dona­tions were made by 209 donors within the first 10 min­utes af­ter the match be­gan.

Within 2 min­utes af­ter the match be­gan, Face­book be­gan re­port­ing on fundraiser pages that the match­ing funds had run out.

You can see a more com­plete break­down of pledges and dona­tions on our web­site.

Or­ga­ni­za­tions on Face­book Pay­ments have likely re­ceived dona­tion amounts by mid-De­cem­ber. Or­ga­ni­za­tions not on Face­book Pay­ments should re­ceive dona­tions in mid-Jan­uary as a check or an elec­tronic funds trans­fer from Net­work For Good.

Match amounts

Face­book had re­ported that they would com­mu­ni­cate match amounts to donors, fundraiser cre­ators, and or­ga­ni­za­tions on or af­ter the Thurs­day, Novem­ber 29. Un­for­tu­nately, Face­book did not fol­low through as it said it would. Nev­er­the­less, by mid-De­cem­ber, we were able to ac­cu­mu­late enough data to es­ti­mate the match end time and the match amounts. We es­ti­mate that the match ran out within 15 sec­onds, and that dona­tions made be­tween sec­onds 0 and 14 were matched. Even though the match ran out as fast as it did, we es­ti­mate that $469k (65%) in dona­tions to our fundraisers were made be­fore the match ended and were there­fore matched.

Our es­ti­mates are sup­ported by mul­ti­ple lines of (in­com­plete) data that donors, fundraiser cre­ators, and or­ga­ni­za­tions have been able to ob­tain from Face­book. At this point, we be­lieve our es­ti­mates are ac­cu­rate within a few per­centage points.

You can find a more com­plete break­down of dona­tions and matches on our web­site.

In ad­di­tion to the fundraisers we cre­ated, other peo­ple cre­ated Face­book fundraisers for EA-al­igned or­ga­ni­za­tions as well. A few of these re­ceived a large amount of match­ing funds from Face­book and PayPal. Speci­fi­cally, I es­ti­mate that this fundraiser for An­i­mal Char­ity Eval­u­a­tors re­ceived $20k in match­ing funds, and that this fundraiser for the Ma­chine In­tel­li­gence Re­search In­sti­tute re­ceived around $25k in match­ing funds. If we were to in­clude these in our to­tal amount matched, this would in­crease the amount to $514k.

Or­ga­ni­za­tions should re­ceive match amounts in mid-Jan­uary as a check or an elec­tronic funds trans­fer from Net­work For Good.

Fol­low-up with organizations

We’re cur­rently in the pro­cess of fol­low­ing up with or­ga­ni­za­tions that we ex­pect to re­ceive a non­triv­ial amount of dona­tions and/​or matches through our ini­ti­a­tive. We’re pro­vid­ing them our es­ti­mated to­tal dona­tion and match amounts, as well as the times­tamps, donor names, dona­tion amounts, and match amounts for each dona­tion. Since or­ga­ni­za­tions won’t be able to see con­tact in­for­ma­tion for their donors, we’re also pro­vid­ing con­tact in­for­ma­tion for pledgers who gave us per­mis­sion to do so.

We’ve also asked or­ga­ni­za­tions to provide us with dona­tion and match amounts they’ve re­ceived on Face­book on Giv­ing Tues­day, and we’ll up­date our web­site with these amounts when we have them available. We ex­pect to re­ceive this in­for­ma­tion from only some or­ga­ni­za­tions.

So far, the dona­tion and match amounts re­ported by or­ga­ni­za­tions have been con­sis­tent with our ex­pec­ta­tions. That be­ing said, as far as we know, Face­book has not pro­vided any in­for­ma­tion to or­ga­ni­za­tions not reg­istered for Face­book Pay­ments, so we don’t have in­for­ma­tion from any of these or­ga­ni­za­tions yet. This in­cludes dona­tions and matches via CEA and GiveWell, which com­bined make up around half of all dona­tions and matches.

We’re also fol­low­ing up with each re­grant­ing or­ga­ni­za­tion—CEA, GiveWell, and ACE—to provide in­struc­tions on how to al­lo­cate the dona­tions and matches they’ll be re­ceiv­ing to the ap­pro­pri­ate or­ga­ni­za­tions, pro­grams, and funds. Th­ese in­struc­tions will be based on the dona­tion and match amounts we es­ti­mate from our fundraisers.

Fol­low-up with donors

At around 20 min­utes af­ter the match be­gan, we sent a fol­low-up sur­vey to donors. We performed an anal­y­sis on the sur­vey re­sults, which you can find here. We seem to have cap­tured the vast ma­jor­ity of donors in our sur­vey, weighted by ei­ther dona­tion amounts or num­ber of donors. In this sec­tion, I re­view some of the sur­vey find­ings.

Pay­ment problems

In 2017, pay­ment prob­lems caused some dona­tions to get de­layed. We wanted to learn about pay­ment prob­lems that donors ex­pe­rienced in 2018 so we could help miti­gate those prob­lems in 2019. We found the fol­low­ing:

  • We had sug­gested that donors call their pay­ment is­suers to in­form them about their plans. Our sur­vey re­sults fail to sup­port the view that this was helpful: Cal­ling pay­ment is­suers was un­ex­pect­edly as­so­ci­ated with more pay­ment prob­lems.

  • As ex­pected, pay­ment prob­lems were com­mon among large donors: 47% of US donors plan­ning to donate >= $5k re­ported pay­ment prob­lems.

  • We wanted to know what pay­ment meth­ods were as­so­ci­ated with the fewest pay­ment prob­lems. In gen­eral, we found that Chase Visa cards fared well in this area.

Donors out­side the US

25% of donors were out­side the US, but be­cause of Face­book dona­tion amount limits and as­so­ci­ated pay­ment prob­lems, their dona­tions rep­re­sented only 4% of the to­tal dona­tion amount.

Were EAs mo­ti­vated to donate more?

We be­lieve the $469k in match­ing funds are coun­ter­fac­tu­ally valid. That is, with­out the efforts of the EA com­mu­nity, we be­lieve the match­ing funds would have oth­er­wise gone to or­ga­ni­za­tions of av­er­age effec­tive­ness.

Some peo­ple have sug­gested that some dona­tions might be coun­ter­fac­tu­ally valid as well: It’s pos­si­ble that with­out the EA Giv­ing Tues­day dona­tion match­ing ini­ti­a­tive, some peo­ple would not have donated, or would have donated less. To provide some in­sight into this, we asked donors whether the match­ing op­por­tu­nity mo­ti­vated them to donate more, not in­clud­ing dona­tions they would have oth­er­wise made in 2018 or 2019. Donors re­ported that 12% of their dona­tions ($85k) would have not oth­er­wise been made in 2018 or 2019. We con­sider this amount to be highly un­cer­tain, given ques­tions about the ac­cu­racy of self-re­ports, the time pe­riod con­sid­ered, and the ques­tion phras­ing.

How we im­proved since 2017

Our es­ti­mated match amount in­creased from $48k in 2017 to $469k in 2018, while our es­ti­mated match per­cent in­creased from 13% to 65%. I be­lieve much of this in­crease was caused by im­prove­ments in our work in 2018.

Fo­cus on donat­ing fast

In 2017, the key to get­ting dona­tions matched was donat­ing within the first 86 sec­onds. We did not know in ad­vance when the match­ing funds would run out in 2018, but had rea­sons to be­lieve it may run out much faster. We sug­gested that peo­ple make their dona­tions within as few sec­onds as pos­si­ble. This turned out to be good ad­vice, be­cause the match ended within 15 sec­onds.

Start­ing early

In 2017, a num­ber of us dis­cov­ered the match op­por­tu­nity about a week in ad­vance, giv­ing us just a week to pre­pare. In 2018, we be­gan prepa­ra­tion work at the end of Septem­ber, around 2 months in ad­vance. Most of the im­prove­ments de­scribed in this sec­tion were only pos­si­ble be­cause we had started early, giv­ing us enough time to im­ple­ment them.

Fo­cus on miti­gat­ing pay­ment problems

In 2017, many large donors had pay­ments de­clined by their pay­ment is­suers, caus­ing de­lays in their dona­tions. We were also un­cer­tain about the max­i­mum amounts per­mit­ted per dona­tion. Face­book said they would match up to $20k in dona­tions per donor, but would Face­book per­mit sin­gle dona­tions as large as $20k?

We spent a con­sid­er­able amount of time in­ves­ti­gat­ing pay­ment limits and de­ter­min­ing on the op­ti­mal amount per dona­tion to recom­mend to large donors. In the end, it’s un­clear how much of this work was helpful. Since pay­ment prob­lems could have been very costly, how­ever, I think the ex­pected value of this work still made it worth­while. The in­for­ma­tion we learned from this may also help us in 2019.

Im­proved out­reach and marketing

We spent a con­sid­er­able amount of time on out­reach and mar­ket­ing to donors, in­clud­ing the fol­low­ing:

  • At EA Global Lon­don, William Kiely spoke to key per­son­nel at CEA to help get them on­board. This was in­stru­men­tal in get­ting EA Giv­ing Tues­day pro­moted in the EA newslet­ter, the GWWC newslet­ter, and on so­cial me­dia.

  • We also spent a con­sid­er­able amount of time on 1-to-1 out­reach to large donors, in­clud­ing talk­ing to donors in per­son and send­ing per­son­al­ized e-mails to donors offer­ing to an­swer any ques­tions.

  • We shared in­struc­tions through e-mail and Face­book, with fol­low-up re­minders lead­ing up to Giv­ing Tues­day.

  • We cre­ated a web­site with in­struc­tions, real-time pledge charts and data, and fol­low-up charts on dona­tions and matches.

Im­proved com­mu­ni­ca­tions with organizations

Given our time con­straints in 2017, we were un­able to no­tify or­ga­ni­za­tions in ad­vance of our plans. This re­sulted in a num­ber of prob­lems:

  • Some or­ga­ni­za­tions were un­able to re­ceive dona­tions be­cause they were not their own 501(c)(3)s, and in most cases we were not able to help them ar­range a way to re­ceive dona­tions through an­other or­ga­ni­za­tion.

  • We didn’t give or­ga­ni­za­tions the op­por­tu­nity to help mo­ti­vate their donors to take ad­van­tage of the Face­book dona­tion match­ing op­por­tu­nity.

  • If any or­ga­ni­za­tions wanted to ask us any ques­tions or ex­press any con­cerns, they were not given the op­por­tu­nity.

In 2018 we made an effort to ad­dress these prob­lems by start­ing a di­a­log with most or­ga­ni­za­tions 1-2 months in ad­vance of Giv­ing Tues­day.

Co­or­di­na­tion with re­grant­ing organizations

In 2017, some or­ga­ni­za­tions, pro­grams, and funds were un­able to re­ceive dona­tions be­cause they were not their own 501(c)(3)s. This could have re­sulted in lost op­por­tu­ni­ties to get dona­tions matched. This could have also in­fluenced some peo­ple to donate less effec­tively just for the match. In ad­di­tion, fol­low-up re­stric­tion re­quests to CEA and GiveWell cre­ated in­con­ve­niences for these or­ga­ni­za­tions.

To miti­gate these con­cerns, we be­gan a di­a­log early with CEA, GiveWell, and ACE to work on ar­range­ments so they could ac­cept dona­tions and matches for or­ga­ni­za­tions, pro­grams, and funds that could not di­rectly re­ceive them through Face­book. We pro­posed ar­range­ments that would min­i­mize the amount of work these re­grant­ing or­ga­ni­za­tions would need to un­der­take. So, for in­stance to avoid the need for re­grant­ing or­ga­ni­za­tions to com­mu­ni­cate with donors in­di­vi­d­u­ally about fol­low-ups, we ar­ranged to in­form the re­grant­ing or­ga­ni­za­tions of the to­tal amounts donated and matched to each or­ga­ni­za­tion, pro­gram, and fund.

This co­or­di­na­tion turned out to be highly valuable, since 40% of dona­tions and 45% of matches were from re­stricted dona­tions go­ing through CEA, GiveWell, and ACE.

Data col­lec­tion and analytics

We an­ti­ci­pated that high-qual­ity data and an­a­lyt­ics would be im­por­tant for a num­ber of rea­sons:

  • We had planned to take on the re­spon­si­bil­ity of com­mu­ni­cat­ing donated and matched amounts to re­grant­ing or­ga­ni­za­tions so they could al­lo­cate the funds ac­cord­ingly. It was un­clear whether matched amounts would be clearly pro­vided to us by Face­book, and we wanted to en­sure we could ac­cu­rately es­ti­mate matched amounts given the data we had available.

  • We wanted to en­sure that we could ver­ify that or­ga­ni­za­tions were re­ceiv­ing donated and matched amounts that were con­sis­tent with our ex­pec­ta­tions. We also wanted to be able to in­ves­ti­gate any dis­crep­an­cies that might arise.

  • We wanted to be able to share in­for­ma­tion on dona­tions to or­ga­ni­za­tions.

  • We wanted to provide reg­u­lar up­dates on pledges to keep prospec­tive donors mo­ti­vated.

  • We wanted to be able to ac­cu­rately es­ti­mate and re­port our im­pact in 2018.

  • We wanted to be able to make bet­ter de­ci­sions in 2019.

For these rea­sons, we took a num­ber of ap­proaches to fa­cil­i­tate data col­lec­tion and anal­y­sis:

  • We had sign-up forms, a pledge form, and a dona­tion fol­low-up sur­vey to cap­ture data from peo­ple donat­ing or con­sid­er­ing donat­ing through our ini­ti­a­tive.

  • We used Google Forms, Google Sheets, and Google Sites to pre­pare a web page with real-time charts and data on pledges over the week lead­ing up to Giv­ing Tues­day. We also pro­vided daily up­dates on Face­book.

  • We struc­tured fundraisers in a way to make data or­ga­nized and eas­ily ac­cessible to us, with­out com­pro­mis­ing donor pri­vacy. In short, each or­ga­ni­za­tion, pro­gram, and fund had its own Face­book fundraiser, cre­ated by one of the or­ga­niz­ers. This per­mit­ted us to see donor names, dona­tion amounts, and times­tamps while main­tain­ing donor pri­vacy.

  • We had some ap­pli­ca­tion de­vel­op­ment work done in ad­vance to per­mit us to ex­port dona­tion data from Face­book fundraiser pages into spread­sheets. This per­mit­ted us to de­ter­mine which dona­tions were matched and in­ves­ti­gate in­di­vi­d­ual dona­tions as nec­es­sary.

  • Within hours af­ter the match ended, we put to­gether a web page with charts and data on dona­tions. After we had enough data to es­ti­mate matched amounts, we added charts and data on matches to the page.

  • We crowd­sourced data from donors on which dona­tions were matched. This data was es­sen­tial in helping us de­ter­mine when the match ended and how much was matched per or­ga­ni­za­tion, pro­gram, or fund.

Areas for im­prove­ment for 2019

If the Face­book dona­tion match­ing op­por­tu­nity is available again in 2019, we could likely im­prove in a num­ber of ar­eas.

Im­proved com­mu­ni­ca­tion with donors

Even though donors out­side the US rep­re­sented 25% of donors, Face­book dona­tion limits and as­so­ci­ated pay­ment prob­lems pre­vented them from donat­ing more than 4%. I think we could have com­mu­ni­cated bet­ter that donors out­side the US may face pay­ment prob­lems and prob­a­bly won’t be able to donate large amounts.

We also re­ceive some feed­back from a large donor that they were not aware that ad­vance prepa­ra­tion on their side was re­quired. I think we could have com­mu­ni­cated this bet­ter as well.

Ear­lier prepa­ra­tion of instructions

Although we had an FAQ ready about 5 weeks in ad­vance, we only be­gan work­ing on in­struc­tions (in­clud­ing demon­stra­tion videos) about 2.5 weeks in ad­vance. The in­struc­tions took longer than ex­pected to fi­nal­ize, and we were only ready to share them about 1.5 weeks in ad­vance. This meant that we were barely able to have in­struc­tions ready on time for GWWC to share us in their newslet­ter. This also meant that we didn’t give or­ga­ni­za­tions much time to adapt the in­struc­tions for their own fundrais­ing pur­poses.

Helping or­ga­ni­za­tions fundraise

Since or­ga­ni­za­tions have ex­per­tise in fundrais­ing, they may have been well-po­si­tioned to help mo­ti­vate their donor base to get their dona­tions matched through Face­book. We in­formed or­ga­ni­za­tions they could use our in­struc­tions to help with their fundrais­ing efforts, but I think we prob­a­bly did not spend as much effort work­ing with them on this as we could have. A few or­ga­ni­za­tions ran effec­tive fundrais­ing efforts to help get dona­tions matched through Face­book, but there may have been more op­por­tu­nity to get more or­ga­ni­za­tions on­board.

In­vest­ing more in technology

We in­creased our us­age of tech­nol­ogy in a num­ber of ar­eas, in­clud­ing mak­ing bet­ter use of Google Forms and Google Sheets, reg­is­ter­ing do­main names (GoDaddy), cre­at­ing a web­site (Google Sites), and mak­ing use of a URL short­en­ing ser­vice (Re­brandly). Be­cause of time con­straints and a prefer­ence not to spend money, how­ever, I think we didn’t in­vest as much as we should have.

  • We tried to stan­dard­ize on us­ing ea­giv­ingtues­day.org for in­di­vi­d­ual e-mail com­mu­ni­ca­tions, but be­cause we had tech­ni­cal prob­lems us­ing ea­giv­ingtues­day.org as the sender, this ended up not work­ing too well. In 2019, I think we should con­sider sign­ing up for a proper paid web host­ing provider such as Blue­host to fa­cil­i­tate e-mail de­liv­ery.

  • We used Gmail to send mul­ti­ple mass e-mails to a few hun­dred re­cip­i­ents who had in­di­cated in­ter­est. Our fol­low-up sur­vey e-mail got de­layed for 2 days by Gmail, prob­a­bly be­cause we had sent too many e-mails re­cently. In 2019, I think we would benefit from look­ing at al­ter­na­tive meth­ods of send­ing mass e-mails, such as Mailchimp or a paid web host­ing provider.

  • We used Face­book for out­reach, but I think we could have made bet­ter use of it. As an ex­am­ple, we pointed peo­ple on Face­book to our web­site, but our web­site pre­view image did not work re­li­ably. We also didn’t have our own Face­book page, even though we may have benefited from one.

In­cor­po­rat­ing dona­tion sur­vey results

If the Face­book dona­tion match­ing op­por­tu­nity is available again in 2019, I think we could likely use our dona­tion sur­vey re­sults to help us do bet­ter. For in­stance:

  • We may want to stop recom­mend­ing that peo­ple call their pay­ment is­suers in ad­vance, since the sur­vey re­sults don’t sup­port the view that it helped.

  • We may want to recom­mend that large donors use Chase Visa cards if they have them.

  • We may want to de-em­pha­size tar­get­ing donors out­side the US in our out­reach. In 2018, they were not able to donate much be­cause of dona­tion amount limits and as­so­ci­ated pay­ment prob­lems.

Les­sons learned

We learned a num­ber of les­sons from our work on this, some of them spe­cific to the Face­book dona­tion match­ing and oth­ers that might be more gen­er­al­iz­able to other EA pro­jects.

EAs can fol­low in­struc­tions and work well as a community

In our 2017 ret­ro­spec­tive, I noted that I was im­pressed with how well EAs worked well as a com­mu­nity. Although our match per­cent was only 13%, this was be­cause we did not provide in­struc­tions to donate within as few sec­onds as pos­si­ble. With re­gard to the in­struc­tions we did provide, EAs fol­lowed those in­struc­tions im­pres­sively well.

In 2018, I an­ti­ci­pated that a non­triv­ial amount of prepa­ra­tion would be nec­es­sary for donors to donate fast enough to get matched, es­pe­cially large donors mak­ing mul­ti­ple dona­tions. We had pro­vided in­struc­tions and demo videos, but I had con­cerns that peo­ple might not read them, and that we might end up with a low match per­cent once again. In the end, these prob­lems did not ma­te­ri­al­ize: 65% of the to­tal dona­tion amounts by EAs were made within 15 sec­onds and were there­fore matched. This sug­gests that, in gen­eral, EAs read the nec­es­sary in­struc­tions and did the nec­es­sary prepa­ra­tion.

Co­or­di­na­tion with a small num­ber of or­ga­ni­za­tions yielded large benefits

I saw it as es­pe­cially im­por­tant for peo­ple to have the abil­ity to donate to a large va­ri­ety of EA-al­igned or­ga­ni­za­tions, pro­grams, and funds. We spent some time ar­rang­ing ways for some or­ga­ni­za­tions to ac­cept dona­tions for oth­ers, with­out cre­at­ing too much ad­minis­tra­tive over­head. In the end, we were able to make ar­range­ments for CEA, GiveWell, and ACE to ac­cept dona­tions for a large num­ber of or­ga­ni­za­tions, pro­grams, and funds in a way that min­i­mized ad­minis­tra­tive over­head for the or­ga­ni­za­tions in­volved. This co­or­di­na­tion yielded large benefits, since 40% of dona­tions and 45% of matches were from re­stricted dona­tions go­ing through CEA, GiveWell, and ACE.

Most EAs waited un­til the day be­fore to get involved

In 2017, the vast ma­jor­ity of pledges, fundraiser cre­ations, and other ac­tivity by the EA com­mu­nity oc­curred within the 24 hours lead­ing up to the match. I saw this as prob­le­matic, be­cause it left lit­tle time for peo­ple to pre­pare, or tell their EA friends about the op­por­tu­nity.

In 2018, we tried to in­crease early in­volve­ment by hav­ing a full plan in place with FAQs, in­struc­tions, and videos available more than a week in ad­vance. We also be­gan pro­mot­ing the op­por­tu­nity a month in ad­vance. Even so, most pledges and other ac­tivity oc­curred within the 24 hours lead­ing up to match.

In 2019, I’d like to con­tinue mak­ing an effort to get peo­ple in­volved ear­lier. But it might be the case that no mat­ter what we do, most peo­ple will wait un­til the 24 hours lead­ing up to the match be­fore get­ting in­volved.

Face­book Dona­tions Sup­port re­sponses were mostly unreliable

In 2018, I opened a large num­ber of sup­port re­quests with Face­book Dona­tions Sup­port to re­quest clar­ifi­ca­tions about donat­ing in gen­eral and Giv­ing Tues­day in par­tic­u­lar. The replies I re­ceived were largely un­re­li­able, and of­ten con­tained con­firmed in­ac­cu­ra­cies. For the most part, we had to make ed­u­cated guesses and find cre­ative ways to get the in­for­ma­tion we needed.

On the other hand, we later found that au­to­matic replies from Face­book Dona­tions Sup­port in­di­cat­ing whether par­tic­u­lar fundraisers or dona­tions were matched were ac­cu­rate.

In 2019, it might still be worth­while to open sup­port re­quests with Face­book Dona­tions Sup­port, but we should start with the as­sump­tion that an­swers writ­ten by hu­mans are un­re­li­able un­less we can in­de­pen­dently con­firm them.

Near misses

We dis­cov­ered a few un­ex­pected prob­lems along the way that could have sub­stan­tially re­duced the amount of dona­tions that were matched. This sec­tion de­scribes those prob­lems and how we were able to avoid them.

Or­ga­ni­za­tions could have be­come ineligible

In our ini­tial com­mu­ni­ca­tions with or­ga­ni­za­tions who were not already signed up for Face­book Pay­ments, we had sug­gested they sign up so they could re­ceive dona­tions ear­lier, and also see the names of their donors. Shortly af­ter, Arushi Gupta dis­cov­ered that sign­ing up could cause or­ga­ni­za­tions to be­come tem­porar­ily in­el­i­gible to re­ceive dona­tions while they un­dergo a vet­ting pro­cess. After in­ves­ti­gat­ing the is­sue fur­ther, we re­versed our recom­men­da­tions. If we had not no­ticed this prob­lem, this could have led to a large num­ber of or­ga­ni­za­tions be­com­ing in­el­i­gible to re­ceive dona­tions on Giv­ing Tues­day.

We saw this prob­lem ma­te­ri­al­ize with one or­ga­ni­za­tion that had signed up for Face­book Pay­ments in prepa­ra­tion for Giv­ing Tues­day. The or­ga­ni­za­tion be­came in­el­i­gible to re­ceive dona­tions, and be­came el­i­gible again just two days be­fore Giv­ing Tues­day.

It’s un­likely we could have pre­dicted this be­hav­ior, but I think this illus­trates the im­por­tance of care­fully read­ing through po­ten­tially rele­vant doc­u­men­ta­tion.

Peo­ple could have donated be­fore the match began

We ob­served that times­tamps on Face­book fundraiser pages were de­layed by around 2 sec­onds, and sug­gested in our demon­stra­tion videos that donors may be able to donate 1-2 sec­onds be­fore the match be­gan. The night be­fore, William Kiely re­al­ized this was a mis­take: Face­book be­gan print­ing e-mail times­tamps in dona­tion re­ceipts, and these times­tamps did not re­flect this 2 sec­ond de­lay. We quickly re­versed course and com­mu­ni­cated that peo­ple should not donate early. If we had not no­ticed this, we could have lost as much as $100k in matches, de­pend­ing on how many peo­ple had planned to donate early.

Ques­tions for 2019

If the Face­book dona­tion match­ing op­por­tu­nity is available again in 2019, our ex­pe­rience in 2018 raises some ques­tions for 2019.

How much time will we have?

In 2016, Face­book re­ported that the match ran “within hours,” al­though we don’t have in­de­pen­dent ev­i­dence of this. In 2017 the match ended at 86 sec­onds, and in 2015 the match ended at 15 sec­onds. If trends con­tinue in 2019, the match could run out even faster. Our strat­egy will likely vary de­pend­ing on how fast we ex­pect the match­ing funds to run out.

What un­ex­pected prob­lems could we face?

In 2017, most of us failed to an­ti­ci­pate the im­por­tance of donat­ing fast, lead­ing to a match per­cent of only 13%. In 2018, we avoided this and other prob­lems, and achieved a match per­cent of around 65%. But as de­scribed pre­vi­ously, we had a few near misses in 2018 that could have led to much fewer dona­tions get­ting matched.

Given that we now know about these par­tic­u­lar prob­lems, we’ll make sure to avoid them in 2019. But I think we should make sure to be vigilant about other un­ex­pected prob­lems similar to these that could oc­cur.

How open should we be about our work?

In 2018 we strug­gled with ten­sion be­tween want­ing to pro­mote the op­por­tu­nity broadly within the EA com­mu­nity, but not want­ing to pro­mote it widely out­side of it. Although we wanted to get as many dona­tions to EA-al­igned or­ga­ni­za­tions matched as pos­si­ble, we an­ti­ci­pated a num­ber of risks in pro­mot­ing it out­side of EA:

  • Other fundraisers and donors out­side of EA could have used our in­struc­tions to get their dona­tions matched, caus­ing the match­ing funds to run out much faster. The in­struc­tions could have been widely shared on so­cial me­dia or in the fundrais­ing ma­te­ri­als of non-EA or­ga­ni­za­tions. This could have re­duced our abil­ity to get dona­tions matched in 2018 and in fu­ture years.

  • We were con­cerned that a com­bi­na­tion of (a) the match run­ning out fast and (b) me­dia cov­er­age about our in­volve­ment could have had nega­tive con­se­quences. For in­stance, we could have been per­ceived by Face­book as the rea­son the match­ing funds ran out as fast as it did. This could have caused Face­book to in­ves­ti­gate our 2018 strat­egy, and op­ti­mize their rules to make the strat­egy more difficult in fu­ture years. Similarly, this could have caused oth­ers to see us as the rea­son the match­ing funds ran out as fast as it did, which could have led to nega­tive per­cep­tions of EA.

    (It’s no­table though that we were fea­tured as a case study in Face­book’s white pa­per on Giv­ing Tues­day 2017, which means Face­book already knew about us. It’s also no­table that me­dia cov­er­age about the match run­ning out in sec­onds in 2018 was pre­sented in a pos­i­tive light.)

For these rea­sons, we tried to avoid gen­er­at­ing at­ten­tion out­side the EA com­mu­nity. For in­stance, we had two op­por­tu­ni­ties to get pos­i­tive me­dia cov­er­age of our efforts, but de­clined them. We also took steps such as avoid­ing men­tion of “Giv­ing Tues­day” in our Face­book group name and de­scrip­tion. This made us less eas­ily search­able by peo­ple out­side of the EA com­mu­nity, but one EA said it made it difficult for them to find us.

It’s not clear whether or not we struck the right bal­ance, and this is an is­sue we should con­tinue to think about in 2019.

Pro­ject costs

Most of our pro­ject costs were in the form of hours worked by our or­ga­nizer team and oth­ers. We also had a small amount of fi­nan­cial costs, which we self-funded.

Hours worked

I es­ti­mate that our or­ga­nizer team worked a to­tal of 330 hours. In ad­di­tion to this, many other EAs were in­volved (see the “Pro­ject con­trib­u­tors”), and com­bined this might have added hun­dreds of more hours of work on the pro­ject. Given the large match amount ($469k), I be­lieve the num­ber of hours worked was worth­while. The main reser­va­tion I have to this is that we did in­volve some highly valuable in­di­vi­d­u­als at the EA or­ga­ni­za­tions we co­or­di­nated with, and the coun­ter­fac­tual use of their time could have been highly valuable.

Fi­nan­cial costs

Our fi­nan­cial costs were self-funded among our or­ga­nizer team. I es­ti­mate we spent $30 on tech­nol­ogy, $75 in pre­paid debit cards for test­ing pay­ment limits, and $650 in test dona­tions.

Most of the pre­paid debit cards were later re­deemed for per­sonal use.

The vast ma­jor­ity of test dona­tions were made by William Kiely to CEA, and will be pre­sum­ably put to good use.

Pro­ject contributors

A large num­ber of in­di­vi­d­u­als from the EA com­mu­nity con­tributed to the EA Giv­ing Tues­day dona­tion match­ing ini­ti­a­tive. This sec­tion notes many of these con­trib­u­tors, but is not com­pre­hen­sive. I apol­o­gize to any con­trib­u­tors I ne­glected to men­tion be­low.

  • Avi Norow­itz, who wrote most of this doc­u­ment (but not this bul­let point), led the en­tire EA Giv­ing Tues­day ini­ti­a­tive, and took re­spon­si­bil­ity for start­ing the or­ga­niz­ing and prepa­ra­tion work early in Oc­to­ber. Avi put in the most time and work into the ini­ti­a­tive and was es­sen­tial to mak­ing it go as well as it did, not just in terms of the match per­centage achieved, but also in terms of how the ini­ti­a­tive was per­ceived by non­prof­its, donors, and the en­tire EA com­mu­nity.

  • William Kiely, who pro­vided valuable work on: out­reach, co­or­di­nat­ing with CEA, fo­cus­ing on the need to donate fast, test­ing dona­tion strate­gies, cre­at­ing demo videos, fore­cast­ing, cre­at­ing fundraisers, col­lect­ing data, iden­ti­fy­ing and work­ing to miti­gate a crit­i­cal flaw in our plans, and pro­vid­ing ideas and feed­back.

  • Anisha Zaveri, who pro­vided valuable work on: prepar­ing real-time pledge charts, post­ing up­dates on so­cial me­dia, helping crowd­source dona­tion match data, test­ing pay­ment limits, and pro­vid­ing ideas and feed­back.

  • Arushi Gupta, who iden­ti­fied a crit­i­cal flaw in our early plans, and pro­vided other ideas and feed­back.

  • Ben Forgy, who did the ap­pli­ca­tion de­vel­op­ment work nec­es­sary to al­low us to effi­ciently ex­port dona­tion data into Google Sheets. This helped us pre­pare charts, de­ter­mine the amounts matched, and perform other data anal­y­sis.

  • Tay­mon Beal, who cre­ated a JavaScript book­mark to help us see when dona­tions were made to the sec­ond.

  • Caitlin Eli­zondo at CEA, Erin Wolff at GiveWell, and Gina Stuessy at ACE, who worked with us to ar­range plans for their re­spec­tive or­ga­ni­za­tions to ac­cept dona­tions for a wide va­ri­ety of difficult-to-reach or­ga­ni­za­tions, pro­grams, and funds.

  • Colm Ó Ri­ain at MIRI and Brit­tany Cham­bers at GFI who pro­moted the op­por­tu­nity to their re­spec­tive or­ga­ni­za­tions’ donors.

  • Aaron Gertler at CEA, Ju­lia Wise at GWWC, and Rob Wiblin at 80,000 Hours for helping us pro­mote the op­por­tu­nity in their re­spec­tive or­ga­ni­za­tions’ newslet­ters and/​or on so­cial me­dia.

  • Aaron Gertler (again) for pro­vid­ing valuable feed­back on this post and other as­pects of our work.

  • Scott Weathers, Cul­lyn O’Keefe, Luke Fried­man, Jeff Kauf­man, and oth­ers who helped pro­mote the op­por­tu­nity on so­cial me­dia and el­se­where.

  • Rochelle Har­ris, Luke Fried­man, Alfredo Parra, Vic­tor Veitch, and Jess Sor­rell who helped test dona­tion amount limits.

  • David Chudz­icki, for pro­vid­ing in­sight on how he made his dona­tions fast enough to get matched in 2017.

  • An­gelina Li, who pro­vided ideas and feed­back.

  • Che Green, for pro­vid­ing valuable ex­per­tise on sur­vey de­sign.

  • The 209 donors who, in a co­or­di­nated effort, made most of their dona­tions within 15 sec­onds af­ter the match be­gan, di­rect­ing $469k in match­ing funds to EA-al­igned or­ga­ni­za­tions.

  • The nu­mer­ous in­di­vi­d­u­als at EA-al­igned or­ga­ni­za­tions who co­or­di­nated with us to get dona­tions matched, and pro­vided us valuable in­for­ma­tion on dona­tions and matches.