Support metrics
How long we actually take to answer.
Every shop-management vendor says support is a priority. The complaint that made us build this one was the opposite: hold queues, weeks without a reply, and no way for a customer to check the claim before they signed. So these are our numbers, in public, computed from the real support conversations in the product rather than from a survey.
They are measured two ways, per quarter and per severity: how long until a human on our side replied, and how long until the ticket was resolved. Both are medians and 90th percentiles, not averages — an average response time is the number that hides the ticket that waited three days, and that ticket is the one you want to know about.
How to read this
What we withhold, and why we withhold it.
The uncomfortable part of publishing support statistics for a small company is that a statistic over a small group is not anonymous. These are the rules we hold ourselves to, written down so you can check that the empty cells are a policy rather than a bad quarter being quietly dropped.
- Each figure needs five tickets from three different shops.
- We test the tickets that actually reached each milestone, not merely every ticket in the row. There must be at least five from three shops, and no shop may supply half or more; otherwise the figure reads “too few to publish” instead of turning one shop’s experience into a statistic.
- Ticket counts are bands, never exact.
- If we published exact counts everywhere and a total, you could recover a withheld cell by subtraction, which would defeat the point of withholding it. So volume is always a range and there is no total on this page for anyone to subtract from.
- The 90th percentile waits for twenty tickets.
- Nearest-rank p90 has three observations in its upper tail only once the sample reaches twenty, and the rounded tail must belong to at least three shops. Every duration is also rounded to the nearest hour inside a day and the nearest six hours beyond one, so no exact clock is published.
- Nothing on this page identifies a shop.
- No shop names, no ticket subjects, no tenant identifiers, and nothing that could be joined back to one. The page is served an aggregate that was already an aggregate before it left our platform, and the code serving it has no way to reach an individual ticket.
- Demo and test data is excluded.
- Our seeded demo shop and our automated end-to-end tests both file support tickets. Every one of them is stamped as test data and filtered out twice — once in the query and once again in the code that computes the figures — because a seeded row that reached a published number would be a number we made up.
- A quarter is a UTC quarter.
- Shops run on their own local time and the product shows them their own local time, but a figure that spans every shop at once cannot use any one of their midnights without a ticket changing quarters depending on who filed it. So the boundary is the UTC instant, and the quarter still running is labelled as partial rather than published as if it were finished.
One thing these numbers cannot tell you: a ticket is measured from when it was opened to when one of us replied, and a reply is a human reply — the earliest message on the ticket written by our side. A status change, an automated acknowledgement, or the ticket being touched for some other reason does not count as an answer, and the clock does not stop for it.