Page MenuHomePhabricator

Dedupe suggestion: track down OG contact records for "new" donors
Open, MediumPublic

Description

Our analysts recently ran some great reports on our France campaign, one of which was that we took in 7,900 "new" donors from our email 1-3 appeals. These are likely new duplicate records in our system, since these emails only go out to past donors.

Could there be a dedupe trigger set up for these new records to see if they can be merged on another record with similar criteria even if email address does not match?

EDIT - Clarification: I was thinking we could look for donations where utm_medium=email and utm_source NOT LIKE 'RML%', assuming email generally does not bring in new donors

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptNov 12 2019, 9:52 PM
CCogdill_WMF triaged this task as Medium priority.Nov 12 2019, 9:52 PM

Seems like we can resolve this on the email side by adding the contact_hash, along with contactID, to all urls. Will discuss this with the email team next Tuesday and confirm when we implement

MBeat33 added a subscriber: SHust.May 27 2020, 9:35 PM

@MNoorWMF will be taking the lead on this! We will need to make a change to all urls in all templates which will take a bit of time. We are committing to wrapping this by EOQ.

Hey there! I met with CCogdill to test this out using contact_hash. Unfortunately, a new contact record was made instead of merging my new donation with my old contact record.

The URL I believe I used was:
https://donate.wikimedia.org/w/index.php?title=Special:LandingPage&country=US&uselang=en&utm_medium=email&utm_source=spTEST&utm_campaign=C1920_Email3&appeal=Appeal-JimmyQuote&hpc=5.35&contact_id=9095559&link_id=2&pym_variant=monthlyConvert&optin=default&contact_hash=f870fede9337461747a96b125e150dfa

Existing civi record used in URL: CID 9095559
New civi record created by entering a new email address in the donation flow: CID 28020941

@KHaggard It looks like you have 2 donations about 45 minutes apart. They have different email addresses. Does that sound right? Did you do multiple tests today?

To address the description: there could be a few reasons why your email list generated new donors.

  • The donors could donate this latest time with new email addresses
  • There could be differences in how they typed their other information which is creating a new donor record and failing to dedupe

Also it depends on when and how you looked up this data. The dedupe script doesn't work on all incoming donations immediately. If Civi thinks it needs to, it will make a new record first and then dedupe it later.

Finally, if the donors are using the exact email address originally on record, there is a process to dedupe on export my email address only. So if you see new records in civi with exact email address matches, there are probably other reasons they aren't deduped... and it wont matter when the export runs and only gives you 1 email address in acoustic.

Could any of this be a factor? I'm happy to talk more.

KHaggard added a comment.EditedJun 10 2020, 10:49 PM

Thanks @DStrine - to answer your question, yes - I messed up the first donation by putting in my email associated with the URL's CID and contact_hash, when the purpose of the experiment was to use a new email address with the old CID and contact_hash in the URL. The second donation was the 'correct' one for the experiment. I made those donations last night, so maybe it could use some more time. I can check civi again tomorrow and update here.

KHaggard added a comment.EditedJun 16 2020, 5:31 PM

@DStrine Sorry for the delay in getting back to you. I checked civi again and my two contact records still haven't merged, even when my work email address donation was done on the URL with my old email contact ID and contact hash information.

@CCogdill_WMF Please let me know if I should give it another try or if there's something else I need to change before testing again.

I can't think of anything else! If fr-tech thinks this still works, can
y'all perform your own test or jump on a screenshare with Katers to
show her how to confirm it works? Seems like something might not be working
here.