Gorgias Stats / Reports Showing Wrong Numbers: Fixes

Start where the scope is: filters, date range, and channel

Before you distrust the math, check what the dashboard is actually counting. The single most common reason a Gorgias number "looks wrong" is that it's answering a narrower or wider question than you think. Every analytics view carries filters — channel, agent, and date range — and a stray selection silently reshapes every KPI on the page.

The Analytics Live overview with the "All channels" and "All agents" filter dropdowns top-right — the scope controls that most often explain a "wrong" number. Confirm channel, agent, and date range match the question you're asking before you dig deeper. (Live overview shown here; this view reflects current state, not a reproduced error.)

Reset the date range to the exact period you're comparing against, set channel and agent back to "All," and re-read the figure. If it now matches, you were comparing two different scopes — done. If it still looks wrong, move down the list.

The common causes, in order

Here's the ranked checklist. Each is a documented Gorgias behaviour, not a defect, and each is quick to confirm.

1. You're reading a median, not an average. This is the biggest "why doesn't this match my spreadsheet" trap. Gorgias reports first response time and resolution time as the median, not the mean — deliberately, to minimize the impact of extreme values. If you export raw ticket timestamps and average them, your number will almost never equal the dashboard's, because one 40-hour outlier drags a mean but barely moves a median. The dashboard isn't wrong; you're computing a different statistic. Compare median to median.
2. The automation numbers are on an up-to-72-hour delay. Gorgias only counts an interaction as automated after the ticket is resolved by AI Agent (or another automation feature) and 72 hours pass with no help from a human agent. So AI Agent and automation metrics lag reality by up to three days — a launch-day spike won't show its true automated share until the window clears. If your automation rate looks low right after go-live, wait out the 72 hours before drawing conclusions.
3. Two reports disagree with each other. A specific, documented case: the Performance by Feature report won't match the Automate / AI Agent Overview report. Performance by Feature runs on legacy data infrastructure with no 72-hour holdup, while the Overview honours the 72-hour attribution delay above. Same underlying activity, two counting rules — so the two dashboards should differ. Pick one report as your source of truth for a given metric and stick to it rather than reconciling across both.
4. Spam, trash, and merged tickets are excluded — as designed. Per Statistics 101, Gorgias metrics exclude deleted (merged), trash, and spam tickets. Spam emails caught by your rules land in the Spam view and are never counted. That's usually what you want — but if your volume looks lower than your raw inbox, aggressive spam auto-close or a broad merge rule is the likely reason. Conversely, spam that escapes your filters and reaches the main queue will count and inflate volume. Reconcile against your Spam view before assuming under-reporting.
5. Reopened and snoozed tickets skew closed-ticket counts. A reopened ticket has no handle time until it's closed again, and a snoozed ticket counts toward the closed count until it's reopened. So a queue full of reopens and snoozes can make "resolved" figures wobble between report refreshes even though nothing changed operationally. If closed counts look unstable day-to-day, check how many tickets are bouncing between closed, snoozed, and reopened.
6. After-hours time is baked into your response averages. Gorgias states plainly that outside business hours are included in first response time. If your team is 9–5 but tickets arrive at midnight, the overnight gap is counted against your FRT unless you're reading a business-hours-aware view. A response time that looks alarming is often just weekend and overnight dead time — not slow agents. Segment by business hours before you act on it.
7. Phone/voice tickets aren't in the counters you expect. Phone tickets are handled differently: their handle time is the sum of call duration and ticket-viewing time, and voice activity is surfaced in the dedicated Voice / Live reports rather than folded into standard email/chat response-time stats. So if you run a voice channel and your "response time" report seems to ignore it, that's expected — check the Voice report for call metrics like average wait time.
8. You're outside the retained window. Gorgias reporting works on a rolling recent window (commonly the last 90 days for standard views), with a maximum date aggregation range of one year and per-cluster cutoffs for older historical data. If a metric reads zero or a comparison looks impossibly clean, confirm the period you selected is still inside the retained range — older data may simply no longer be queryable. Export what you need on a schedule if you require a longer history than Gorgias keeps live.

A quick symptom → cause table

Once you've reset filters, this table points you at the likely cause fast.

The honest limits — and what actually needs a higher tier

Most of the fixes above cost nothing but attention: reset a filter, compare median to median, wait out the 72 hours. Be clear-eyed about the two things that do have prerequisites. Business-hours-aware reporting depends on having your schedule configured, and Voice metrics only exist if you run the Gorgias phone channel — no connected store or custom domain is required to reconcile the email/chat numbers this post is about, and a trial account shows the same statistics engine. If you want the conceptual tour of every panel, see Gorgias stats and reporting explained; for how satisfaction specifically is measured, Gorgias CSAT explained covers the survey side, and if you're newer to the platform, what is Gorgias sets the context.

Where an AI layer changes the picture

Here's the part the dashboard can't fix for you: a lot of "wrong numbers" anxiety is really volume anxiety. Response-time medians balloon because too many repetitive tickets pile up overnight; automation rates disappoint because rules only deflect the tickets whose exact wording you anticipated. That's the seam where an AI agent for customer service helps — not by editing your reports, but by removing the ticket volume that made them look bad. Macha is one such layer. It runs on top of the Gorgias you already use — it doesn't replace your help desk, it connects to it and reads and writes the same tickets, replies, and tags your team works. It resolves the repetitive, intent-heavy tickets before they age into your response-time median, so the numbers improve because the underlying reality did, not because you reconciled a spreadsheet. For how the reasoning works, see the Gorgias AI agent explained.

Because an agent acts on meaning rather than a fixed keyword, you can't just eyeball it. Macha lets you grade a candidate agent against your real historical Gorgias tickets before it touches a live queue, and you can extend what it does through a custom tool that turns any REST API into an action the agent can call — checking order or shipping status mid-conversation, the kind of lookup that turns a WISMO ticket into an instant resolution instead of another line in your volume chart. See pricing for the specifics — credits are charged per AI action, not per resolution, so the cost tracks the work done.

FAQ

Why don't my Gorgias reports match my own exported numbers? Almost always because Gorgias reports first response time and resolution time as the median, not the average. If you export ticket timestamps and take a mean, a single long outlier will pull your number away from the dashboard's median. Compare like for like — median to median — and the gap usually disappears.

Why is my automation / AI Agent rate lower than it should be? Gorgias counts an interaction as automated only after the ticket is resolved and 72 hours pass with no human agent involvement, so automation stats lag by up to three days. Right after a launch, wait out the 72-hour window before judging the number. Note too that the Performance-by-Feature report skips this delay, which is why it won't match the Automate Overview.

Why is my ticket volume lower than the emails I actually received? Gorgias statistics exclude deleted (merged), trash, and spam tickets by design — spam caught by your rules goes to the Spam view and is never counted. If volume looks too low, check whether spam auto-close or a broad merge rule is removing more than you expected.

Why does my response time look so high? Because outside business hours are included in first response time. Overnight and weekend gaps get counted against your FRT unless you segment by business hours. The number often reflects dead time, not slow agents.

How far back does Gorgias reporting go? Standard views work on a rolling recent window (commonly the last 90 days), with a maximum aggregation range of one year and per-cluster cutoffs for older data. If you need longer history, export on a schedule before it ages out of the live reports.

Ready to stop fighting your response-time median and shrink the volume behind it? Start a free trial and grade a Macha agent against your own Gorgias history before it goes live.