Zendesk dbt model innacurate

[guillaume]

Hi,

I started using Fivetran about 2 months ago mainly for Zendesk data.

We implemented the model as defined here: https://hub.getdbt.com/fivetran/zendesk/latest/ and everything is working fine.

We have been a bit deeper in the data though and as we have been building some reports, we discovered some issues with the data when compared to the reports built directly within Zendesk:

- what is defined as "agent replies" in Zendesk explore is closer to "count_public_agent_comments" value than "total_agent_replies" in zendesk__ticket_metrics table. But even when using "count_public_agent_comments", I can't reach the same value for all records.

- merged tickets are not defined the same way as in Zendesk explore. If a ticket is merged to another one, it can show with no last_solved_at or a last_solved_at value is not taking in account the merge and close (so the last_solved_at is keeping the old value). Also, the full_resolution_business_minutes is sometimes null following that which is incorrect.

Kind regards,

Guillaume

[avinash-kunnath]

Hi @guillaume , thanks for diving deep into the Zendesk package and identifying these issues!

-Regarding "agent replies", it does appear we use `count_agent_comments` as the base model for calculating agent replies rather than `count_public_agent_comments`, so that would be the reason these numbers are a bit off.

Would you be willing to submit a bug issue on dbt_zendesk? That way we will try and tackle fixing this issue in a coming sprint.

- I'm curious as to how you're seeing merged tickets defined in our package? We haven't really built out merged ticket functionality so we are unfamiliar with the overall logic. Could you show us an example of what discrepancies you are seeing (i.e. is it tickets you're seeing that are not being correctly defined in a model as merged/unmerged)? This will help us figure out the best way forward to address this particular issue.