Time to First Response (definition please)

(Bcguy) #1

Just updated to current version of sw.

What exactly does “Time to First Response” mean, how is it calculated, and what is the discourse team thinking is the value of this metric?


(Mittineague) #2

My first thought was “Time to First Reply” might be better.
Meaning you started this topic ~10 minutes ago so this post would result in a value of ~10 minutes.

But thinking more, maybe it averages in Messages and emails too? (I doubt it, but maybe)

As for value, I’m thinking it reflects “forum health” i.e. Higher indicates topics are going unanswered longer, Lower means the activity and participation are better.

(Michael Downey) #3

Where are you seeing phrase this exactly?

(Mittineague) #4

I’ve seen it on the Admin Dashboard page - a Report


(Régis Hanol) #5

It’s the average time (in hours) between the creation of the topic and the date of the first reply.

Using the following SQL query

It’s from one of our customers who wants to know how long it takes (on average) for topics to get a reply on a per-category basis (not sure this stat has a value when considered globally). They have moderators for every categories and wants to know how well they are doing.

(Jeff Atwood) #6

The query should exclude self-answers though. E.g. the owner of the topic posting the first reply, should not be considered a reply.

(Régis Hanol) #7

That’s done in this where clause :wink:

(Lisa Wess) #8

Oh this is awesome for us as well! Can’t wait to get updated now. Thank you :slight_smile:

(Bill Snider) #9

This gives you an idea of long new topics sit before being responded to. This is a very important metric when looking at the health of your community. In a healthy community, new topics don’t sit more than a day without being responded to. Also, you can see which boards are most active, both in post counts, page views, and times to response.

Furthermore, in an enterprise environment, management may require that all posts are being responded to. This metric lets them see if that is happening.

(Scott Trager) #10

It’s time to Response instead of Reply because it excludes self-replies (a lot of users reply to their own posts with updates etc. or to bump their post if it’s been sitting for awhile and this way it doesn’t stop the clock in these instances).

(Michael Downey) #11

Do response times really follow a normal distribution? Aren’t some topics more likely to be neglected or ignored due to being more edge cases for a community? (EDIT: i.e., outliers?) If so, is mean really better to use than median? After all, it seems the value of this measure is to find out how many people are getting frustrated due to lack of replies.

Or maybe I’m thinking about this metric wrong…

(Mittineague) #12

In the lab we used Standard Deviation to account for outliers that might skew results

But forum stats are not medical stats so I think getting deep into statistics here might be a bit over-kill

(Michael Downey) #13

Excluding self-replies is good because it will take out the 0 values. But it seems to me that the “infinity” (well, hours since first post) values (that is, topics which have yet to receive a response) are going to throw the actual average way off. Or did I miss something in the calculation that excludes topics with zero replies from others?

(Régis Hanol) #14

This value only take into account topics that have at least one reply from a different user.

(Bcguy) #15

Yes - we have people post “news” which is more of an “FYI” and not necessarily a discussion topic - and since people don’t respond, it throws off the average dramatically. I tend to agree - I think the Median would make more sense for many of us. Rather than encouraging Mods to post something just for the sake of posting something to lower the average.

(Scott Trager) #16

Those posts don’t get counted… as long as there is no response it’s not factored into the average (unless there is a bug).

(Wes Osborn) #17

I noticed that our numbers seemed a bit off and I dug into it deeper and think I know why.

This currently appears to count any post types, including moderator actions as a reply. Is there a chance that it could be limited to counting only those replies with a post_type of 1 (regular)?

Only count posts types of regular when calculating time to first response for admin stats