REPORT · METHOD FIRST
How fast do WhatsApp sales teams actually reply?
Nobody has measured this properly, including us, not yet. This page fixes the method in public first, so the numbers mean something when they arrive.
- Method published 2026-09-01
- Figures publish with the first cohort
n/aNo cell has cleared the cohort rule yet, so no cell carries a figure.
First response, measured
timestamps at captureLaura Bennett and Chris Delgado · one workspace, Hudson Cotton Co.
- 14:22inbound, opens the conversation
Hi, I'd like to order the blue dress. Size M please.
- 2m 0selapsed. This is the whole measurement.
- 14:24first human reply, from anyone on the team
Of course! Blue dress, size M, $89. Want me to send a payment link?
- automated greeting
- team-opened thread
- history-sync import
- group conversation
the method is fixed, the figures are not
This page defines how WAOversight measures WhatsApp sales response benchmarks: median first response, coverage and deletion rates, from anonymised aggregates across customer workspaces. No figures are published yet, because the cohort has not cleared the minimum size that makes anonymisation honest. The method is public now; the first numbers follow quarterly once it does.
01 — THE THRESHOLD
What a figure has to clear before it is printed
Two counts and one judgement. Until a cell clears all three, it stays empty and says why.
Publication threshold
rule fixed 2026-09-01A cell below the line is suppressed, not estimated. So is any cell that would let a reader work out which workspace it came from.
- 10Workspaces before a cell publishes
- 25Linked numbers before a cell publishes
- 0Figures this report has published so far
A benchmark computed from a handful of teams is not anonymous and it is not representative. It is a number with a company showing through it. So the rule is a count, fixed in public before the first figure exists: 10 workspaces and 25 linked numbers behind anything we print.
The cohort row is the one that matters today. Nothing has reached the line, so every cell in the report reads n/a, and it keeps reading n/a until it can be filled honestly rather than filled early.
Splits obey the same count. A team-size band or an industry cut publishes only where enough workspaces sit inside it, which is why the first report will be narrower than the one after it.
02 — MEDIAN AND P90
The typical customer and the unlucky tenth are two questions
The report will answer both, on every metric, because either one on its own flatters or frightens.
First response, one workspace
demonstration dataset, not a benchmark- The typical customer waited8m 12shalf of conversations were answered inside this
- The slowest tenth waited at least41m 3sand the mean would have sat between the two, describing neither
- 8m 12sMedian, the wait a typical customer had
- 41m 3sp90, the wait the unlucky tenth had
- 5.0×The second figure over the first, one week
A mean would sit between these two bars and describe neither. One long silence in a week of fast replies drags it past an hour, and a manager reading it goes looking for a problem that is not where the number pointed.
The median is what a typical customer got. p90 is what the unlucky tenth got. The distance between them is the part a manager can act on, so the report carries both side by side and never a single headline figure.
What is drawn here is one workspace's week from the demonstration dataset, shown so you can see the shape the report will take. It is not a benchmark and it will not become one. A benchmark is this shape pooled across a cohort that has cleared the rule, which has not happened yet.
What this page is
Every page ranking for WhatsApp response benchmarks recycles the same 2 cross-channel studies, measured on phone and web forms, relabelled for WhatsApp, methodology absent. We will not add to that pile or invent a median to have something to publish. What we can publish now is the measurement itself: what is counted, on what data, what is excluded, and the threshold the cohort must clear before a single figure appears. The figures then arrive quarterly, dated, and reproducible from this method.
The numbers
This is the shape of the report. It fills in once the cohort clears the threshold below, quarterly from then on.
| Metric | Median | p25 | p75 | p90 |
|---|---|---|---|---|
| First response time, business hours | n/a | n/a | n/a | n/a |
| Coverage rate, 24 hours | n/a | n/a | n/a | n/a |
| Deletion rate per 1,000 messages | n/a | n/a | n/a | n/a |
| Edit rate per 1,000 messages | n/a | n/a | n/a | n/a |
Splits by team-size band and by industry follow the same rule: a split publishes only where enough workspaces sit in it to anonymise, and not before.
The method
What is measured
- First response time. From a new inbound customer message that opens or reopens a conversation to the first human reply from the team, using timestamps recorded at capture, not read later from phones. Automated greetings do not count as a reply.
- Coverage rate. The share of new inbound conversations that received any reply within 24 hours.
- Median and p90, never an average. One 3-day silence in a week of 5-minute replies drags a mean past an hour and tells you nothing. The median is what a typical customer experienced; p90 is what the unlucky ones did, and the gap between them is the management signal.
- Deletion and edit rates. Messages deleted or edited by the sender per 1,000 outbound messages, counted from the capture record, which keeps the original.
Cohort definition
- Company-owned numbers linked to active, paying workspaces. No trials, no demo data, and none of our own workspaces.
- A workspace enters the quarter's cohort only if its numbers were linked for the full quarter.
- The tenant has consented to anonymised aggregation. Consent is opt-in, and leaving removes the workspace from the next quarter's cohort.
Thresholds
- No figure is published from fewer than 10 workspaces or 25 linked numbers.
- No team-size or industry split is published from fewer than 10 workspaces in that cell.
- Any cell that would let a reader isolate a single workspace is suppressed, even above those counts.
Exclusions
- Group conversations, where response ownership is ambiguous and would flatter the numbers.
- Messages imported by history sync, whose timestamps predate capture and cannot be verified the same way.
- Conversations opened by the team rather than the customer, for the response metrics, because there is nothing to respond to.
- Days a number was unlinked or a session was down, rather than counting silence as slowness.
Anonymisation
Figures are medians and percentiles pooled across the cohort. Nothing is attributable to a workspace, number or person, no tenant is named, and the thresholds are checked before every release. Durations round to the minute and rates to one decimal, removing the precision that could fingerprint a small team. If a cell cannot honestly be published in a quarter, it stays empty and says why, which is the reason this page exists before the numbers do.
Measure your own team
You do not need the report to know where you stand. The metrics guide defines all 47 metrics worth tracking, and the 4 that matter most are the ones this report benchmarks. To see them on your own numbers, pricing is per monitored number and setup is 1 QR scan each.
03 — ABOUT THE REPORT
Questions about the method
Why are there no numbers on this page yet?
Because the customer base has not cleared the minimum cohort we set for anonymised publication. A benchmark computed from a handful of identifiable teams is neither anonymous nor representative, and publishing it would be the kind of unverifiable claim this site exists to be the opposite of. The method is fixed and public now; the figures follow once the threshold is met.
What counts as a first response?
The elapsed time from a new inbound customer message that opens or reopens a conversation to the first reply from anyone on the team, measured from timestamps recorded at capture. Automated greetings do not count as a reply. Messages that arrive outside the workspace's stated business hours have the out-of-hours portion reported separately, not hidden.
How will teams be anonymised?
Only aggregates are published: medians and percentiles pooled across workspaces, never a figure attributable to one team. No cell is published below the minimum cohort, splits that would isolate a workspace are suppressed, and participation requires the tenant's consent to anonymised aggregation. No workspace, number or person is ever named.
Your numbers, before the industry's.
Fifteen minutes, your data, no slides. The dashboard measures your team exactly the way this report will measure everyone's.
WAOversight is an independent product and is not affiliated with or endorsed by WhatsApp or Meta.