The Unanswered Lead Study: 38.9% of People Who Message a Business Never Get a Reply

38.9% of People Who Message a Business Never Get a Reply

We looked at every message that passed through our platform in the last twelve months. 1.85 million of them, across 86 agencies and their client accounts.

Then we asked one question nobody seems to publish an answer to: when a real person writes to a business, how often does the business write back at all?

Not how fast. Whether.

The answer is that 38.9% of the time, nobody ever replies.

The short version

• 33,020 of 84,823 conversations that received an inbound message got zero replies. Ever. Not a slow reply. Not an automated one. Nothing.
• 1,129 people sent five or more messages and were never answered once.
• Spam does not explain it. Excluding conversations with any spam-flagged message makes the figure worse, 40.9%.
• Among individual inbound messages, 12.7% never received a subsequent reply, and 47.1% were answered in under a minute. The distribution is not a bell curve, it is two separate populations.
• The gap between agencies is enormous. The best answers everything. The median leaves 28.6% of inbound conversations unanswered. The worst leaves 100%.
• We could not show that turning on automation reduces this, and we explain below why we are not publishing the number we got.

What we measured

Every message row between 9 September 2025 and 9 September 2026, in both directions, across 86 agency accounts and their clients. That window holds 707,743 inbound messages across 179,563 conversations.

"Inbound" means a message the business received. "Answered" means at least one outbound message exists in the same conversation afterwards. Both are structural facts in the data rather than judgements, which is the point: we are not scoring reply quality, only presence.

Of the 179,563 conversations, 84,823 contain at least one inbound message. Those are the ones a human being started, and they are the entire subject of this article.

The headline: 33,020 conversations, zero replies

Read the second row against the third and fourth. Of every conversation a person started by writing in, more went permanently unanswered than were answered.

Those 33,020 conversations contain 52,448 messages. Fifty-two thousand messages sent to a business over a year, none of which produced a single character in response.

They kept trying

The obvious defence is that these are drive-by messages, a single "hi" that nobody could reasonably act on. Some are. Most are, in fact. But not all of them, and the tail is the uncomfortable part.

One thousand one hundred and twenty-nine people wrote to a business at least five separate times and never received one reply.

Whatever the first message was, by the fifth attempt intent is not in question. Those are not tyre-kickers. That is 1,129 instances, in a single year, of somebody trying hard to give a business money and being ignored.

"It is just spam"

This is the first objection everyone raises, so we tested it directly rather than arguing about it.

Of 707,743 inbound messages, 1,796 carry a spam flag. That is 0.25%.

Exclude every conversation containing any spam-flagged message and the unanswered rate does not fall. It rises, from 38.9% to 40.9%, because the spam-flagged conversations were disproportionately ones somebody had already dealt with.

Spam is not the explanation. It is not close to being the explanation.

For the ones that do get answered

Restricting to individual inbound messages rather than whole conversations, here is the full distribution of time-to-first-reply:

The median is 54 seconds. The 90th percentile is 11.4 hours.

A median of under a minute alongside a p90 of eleven hours is not one distribution with a long tail. It is two populations sharing a chart. Roughly six messages in ten are handled almost instantly by something automated, and the rest wait for a human who may or may not arrive.

If you report an average response time to a client, you are averaging across those two populations and describing neither.

The speed gap between automated and human replies

A subset of outbound messages records what produced them, which lets us compare directly.

Median 30 seconds against median five hours.

Read that comparison carefully, because it is not a fair fight and we are not going to pretend it is. AI agents fire on configured triggers, which are chosen precisely because they are predictable. Messages that reach a human are disproportionately the ones nobody automated, which usually means harder. The right conclusion is not "AI is 600 times better than your team". It is that the two are doing different jobs, and the messages left to humans are the ones that wait.

The spread between agencies is enormous

Across the 32 agencies with at least 100 inbound conversations in the window:

One agency answered every single person who wrote in. Another answered nobody, across more than a hundred conversations.

This is the most useful number in the article, because it is the one that tells you the 38.9% is not a law of nature. A quarter of these agencies keep it under 9.4%. Whatever they are doing is available to everyone else.

When messages arrive, and when they get answered

Hours are UTC, and our sample skews toward US and European accounts, so treat the clock as indicative rather than exact.

Inbound volume peaks at 21:00 and bottoms out at 04:00, which is what you would expect. The unanswered rate does something less expected.

The hours with the most messages have the fewest unanswered ones. The worst hour of the day is 09:00 UTC at 17.3% unanswered; the best is 21:00 at 9.5%, which is also the single busiest hour.

By weekday, Sunday is worst at 15.5% and Thursday best at 10.8%.

The most plausible reading is that peak hours are the ones people have deliberately covered, by staffing or by automation, and the quiet hours are the ones nobody thought to cover because they feel low stakes. A message arriving in a quiet hour is worth exactly as much as one arriving at peak.

What we deliberately did not conclude

We ran the comparison you would expect a CRM company to run: do agencies using AI or workflow replies leave fewer conversations unanswered?

The raw answer came out backwards. Agencies with automation showed a higher median unanswered rate than those without, 57.5% against 23.6%.

We are not publishing that as a finding, in either direction, because it is confounded beyond rescue at this sample size. The ten agencies with automation handle 64,358 conversations between them; the twenty-two without handle 18,952. High-volume accounts turn automation on precisely because they are drowning, so the causation plausibly runs backwards. When we split by volume band to control for it, the pattern did not hold up: median unanswered was 39.2% for the smallest band, 22.2% for the middle and 32.0% for the largest.

With 32 agencies, that is noise. The honest position is that our data does not currently show whether automation reduces unanswered leads, and we would rather say so than pick the framing that suits us.

What to do about it

Four things, in order of cost.

Count your own number first. Take last month's conversations and work out what share received no outbound message at all. It is one query or one afternoon in a spreadsheet. Most people are wrong about their own figure, and they are wrong in the same direction.

Cover the quiet hours before you optimise the busy ones. The data says the gaps are at 04:00 to 11:00 UTC and on Sundays, not at peak. Peak is already covered because it feels urgent.

Answer the persistent ones today. If somebody has sent three or more messages and received nothing, that is the single highest-intent list you own and it takes minutes to pull.

Make capture structural rather than diligent. Every process that relies on a person remembering to check a channel works for two weeks. The 100%-unanswered agency in our sample was not lazy; it had a channel nobody owned.

For the mechanics of routing and ownership across multiple client accounts, our comparison of the best CRM for marketing agencies covers how the main platforms handle it.

How this relates to our cold DM research

We previously published cold DM benchmarks built on 63,286 outbound openers, which includes a speed-to-lead gradient showing how response time relates to whether a conversation continues.

That study and this one answer different questions and use different populations. That one measures conversations the business started and asks whether replying quickly keeps them alive. This one measures conversations the customer started and asks whether anyone replied at all. Where the two overlap, the cold DM figures are scoped to conversations seeded by an opener and should be read as the more specific of the two.

Method and limitations

Sample. All message rows from 9 September 2025 to 9 September 2026 inclusive: 707,743 inbound messages across 179,563 conversations and 86 agency accounts. Per-agency figures use only the 32 agencies with 100 or more inbound conversations.

Definitions. Inbound is any message the business received. A conversation counts as answered if any outbound message exists in it after an inbound one. Response time is measured from the inbound message to the next outbound message in the same conversation.

Limitations, stated plainly.

• 86 agencies is a modest sample, weighted toward businesses that chose our platform and toward Instagram and Meta channels specifically. It is not a census of small business.
• A reply sent outside the platform is invisible to us. If somebody read a DM and phoned the person back, our data records that as unanswered. This is the single largest source of overstatement in the 38.9% and we cannot size it.
• Some unanswered messages should be unanswered: reactions, one-word acknowledgements, wrong numbers. The spam flag catches only 0.25% and is clearly not catching all of it.
• Response times are observational. Faster accounts may differ from slower ones in ways we cannot see.
• The automation comparison is confounded, as described above, and no causal claim is made anywhere in this article.

No individual person, account, agency or customer is identified. Every figure is aggregated across multiple accounts, and no message content was read or published.

FAQ

Is 38.9% unusually bad?
We have nothing to compare it against, which is why we published it. If you know of another dataset of this size measuring whether inbound messages get answered at all rather than how fast, we would genuinely like to see it.

Does this include cold outreach the business sent?
No. Conversations where the business wrote first and got no reply are a separate row in the first table, 52.8% of all conversations. The 38.9% covers only conversations a person started by writing in.

Why measure whether, instead of how fast?
Because response-time averages hide it completely. A business answering 60% of its messages in thirty seconds and ignoring the rest reports an excellent median and is losing four leads in ten.

What is the single number to copy from this?
Your own. The spread between the best and worst agency here runs from 0% to 100%, which means the population figure tells you almost nothing about your business, and one query tells you everything.