How is Median First Response Time calculated?

  • 18 August 2022
  • 5 replies

Hi, we're trying to recreate the Intercom's Median First Response Time metric in our mission-critical company dashboard, and need an exact logic used for calculating this metric.


In particular:

  • Do you exclude conversations without responses from the calculations (in 99% of cases, those are spam messages)
  • If the first response is an auto-reply - does Intercom count the time before it, or do they only count the time before a real person’s response?
  • Reports in Intercom have time frames. Does Intercom count conversations that started before the chosen period but got their response within a selected period?
  • Is the median first response calculated only for the closed conversations?
  • Why is there a discrepancy in the time reports? (See below)

image (81) 

Advice on this would be highly appreciated.


Best answer by Aparna 22 August 2022, 18:28

View original

5 replies

Hey @datomni​ !


1. If an inbound conversation is closed without any initial response from an admin then this will not be considered in any of the median response time metrics. There needs to be a pair of user message + admin message to contribute towards to the median response time.

2. You can configure this. Just head to any of the reports with FRT charts, and select Bot Inbox times excluded in the top filters.

3. The median of the selected team or teammate's first response times to any new inbound conversation, from the first inbound message to the first reply, sorted by the date the conversation started. The charts show the First Response Time (FRT) grouped by conversation creation date. So, even though we only count the time within office hours (when 'Calculate your team’s response metrics within office hours only' has been enabled in Office Hours settings), this will be displayed on the chart according to the time the conversation was created.


For Example - A conversation is created at 4am GMT, but office hours are set to 9am - 6pm GMT. The conversation receives a first reply at 9:30am GMT. Reporting will show a 30min FRT for this particular conversation that was created at 4am, because it was answered 30mins after office hours started.


4.I think the median first response is calculated for all conversations(open/closed)

5.This is a unique count of all conversations replied to by teammates within the time period. Do you think that the count is wrong?

Hi @aparna​ thanks for the response.


Here's another issue. Is there any limit, filter, or transformation applied to the conversation data which is accessible via the Intercom API?


I'm asking because we're pulling data from Intercom API and we're seeing systematic issues in data quality that include missing, excessive or incorrect conversation data in our data warehouse. Please see the attachment below for details.


What could be an explanation for that?Screenshot 2022-08-24 at 14.34.12 

@datomni​ ! Apologies for the delay 🙏 I don't think there is any limit filter. Do you list/update large volumes of data? Can you open a conversation with our support team as this may need more investigation on our end?


Hi - I am also having a similar issue with trying to replicate the median first response time from the raw data to match the dashboard. I have tried using the fields author type and part type (from the API raw data) to select only those conversations which have a user comment and an admin comment. That is, conversations which have a data entry with author type = admin and part type = comment as well as author type = user and part type = comment. Can you confirm exactly what the logic is which decides whether a conversation is included in the median first response time calculation. Many thanks - Sam


Would love to hear if there is any update on Sam’s question!