Identifying by email only works if each event sends email

[tiina303]

The problem:

1. we make the decision to send the person or not based on if they have ever had email set

   customerio-plugin/index.ts

   Line 212 in 93205d2

   return customer.status.has('with_email')

2. however in the current event we might not have the email

   customerio-plugin/index.ts

   Line 185 in 93205d2

   const email = getEmailFromEvent(event)

   (

   customerio-plugin/index.ts

   Line 193 in 93205d2

   customerStatus.add('with_email')

   could have been set by a previous event)
3. then we never override the id, because we didn't have an email

   customerio-plugin/index.ts

   Line 244 in 93205d2

   id = customer.email

If I understand correctly, then we would always want to update the properties, i.e. send the identify when the person properties change

which would mean we'd need to instead of keeping the info that the person has been identified by email we'd need to keep the emails, which is a bit tricky (running arbitrary code, e.g. SQL injection that we'll need to guard against), in addition to increasing our storage demand.

We currently don't have a way to pass person info to the plugins, which would be the ideal solution for this problem

See thread for more info: https://posthog.slack.com/archives/C03PN99EJ4C/p1675946098284529

[Twixes]

For context, I exposed person to plugins in #10572, precisely to solve this class of problems (and Customer.io was the primary motivation too)… but that didn't go in, because of three complications:

1. When doing that, I discovered that our typing of the internal and external event formats was a real mess, and had to do a large refactor of those types to get all the ducks in a row (at least for the purpose of passing person along). The refactor solved the problems blocking person passing, but it resulted in big merge conflicts.
2. The merge conflicts would be manageable, but that PR also added the ability for person to be modified in processEvent, which was a much bigger change.
3. Well, at the same time I switched back to team Product Analytics as planned.

So this should probably be done by someone who isn't switching teams in a week. 😄 I'd also drop processEvent from the scope – it would be really nice, but I'm pretty sure it's not a must.
Considering the above, it should be pretty simple to pass person in exportEvents. That data already must be known at that stage, and it can't be mutated from that point.

A plugin-scaffold typing update will be needed though. PluginEvent currently contains person as an optional property, but that's not real in practice, so it's a bad type (and I introduced it – it was just transitional for #10572, but after switching teams, I didn't get to cleaning that up_. PluginEvent shouldn't have person.
Meanwhile, ProcessedPluginEvent (the exportEvents and such version) definitely should have person as required – currently it's optional and also not real in practice AFAIK.

[MarconLP]

Following example:
we send two events:

posthog.capture(
    'test_user_1',
    event='event_name',
    properties={
        '$set': {'name': 'Max Hedgehog'},
        '$set_once': {'initial_url': '/blog'}
    }
)

and later we decide to set the email:

posthog.capture(
    'test_user_1',
    event='event_name',
    properties={
        '$set': {'name': 'Max Hedgehog', 'email': 'max@posthog.com'},
        '$set_once': {'initial_url': '/blog'}
    }
)

Inside of PostHog there will be one person:

But inside Customer.io there will be two people:

[oagix]

@tiina303 @Twixes - any ETA plans or workaround on how to fix this? It's causing major issues with our CIO instance via Posthog and it's pretty much the final piece of the puzzle before a reliable CRM setup. Thank you!

[tiina303]

Unfortunately this isn't on our immediate roadmap, there are some more important scaling improvements we need to do first before we come back to this.

[oagix]

Unfortunately this isn't on our immediate roadmap, there are some more important scaling improvements we need to do first before we come back to this.

Any pointers on how to fix it or workarounds? This is completely breaking the PostHog integration, doubles billing on CIO, and segregates the data completely.

I'm sure there must be a tweak that can be done - I remember the integration used to work quite well a while ago.

We'd really appreciate a workaround or fix... as it's something we use on more than one PostHog project.

Thank you

[tiina303]

As the title suggest the workaround is to send email with every event (you send to PostHog)

[oagix]

Hey @tiina303 , just to make sure we're on the same page, here's the issue I'm referring to. Our team was sent to this GitHub issue as a reference by PostHog support for that duplicate email problem.

We are able to identify PostHog users in CIO, but for some reason it double-identifies them with both the email and the ID...

[MarconLP]

Hi @altechzilla, as mentioned in this comment #26 (comment), the issue is, that when sending events without the email in the properties object, they will not be grouped to the other events of the same user (events where the email is in the properties object).

As @tiina303 already mentioned, a workaround to that issue is to always send the email in the event properties to posthog itself.

[andreacavagna01]

This bug is blocking me from using Customer.io at the moment. How I can overcome that or include the email in each event in the best way possible?

[MarconLP]

+1 https://posthoghelp.zendesk.com/agent/tickets/9111

[andreacavagna01]

Even if email is added as an attribute, the identifier is not recognized in customer.io.

How do I have to format each event to be identified by email?

[MarconLP]

Even if email is added as an attribute, the identifier is not recognized in customer.io.

Could you link such an event?

How do I have to format each event to be identified by email?

You will need to include the event in each event either by including it manually:

posthog.capture('user signed up', { $set: { email: 'marcus.h@posthog.com' } })

You can also set that property as a Super Property, which will include them in all future events:

posthog.register({ $set: { email: 'marcus.h@posthog.com' } })

[MarconLP]

+1 https://posthoghelp.zendesk.com/agent/tickets/10285