Omnichannel Messaging Statistics 2026: The Numbers That Survive a Source Check

Every messaging statistic you are about to paste into a deck has one of three problems: it is a survey answer wearing the costume of a behavioural measurement, it is quoted in a unit that does not match the number sitting next to it, or it is from 2023 and nobody told you. We went looking for the primary sources behind the numbers people cite about customer conversations in 2026, and opened every document rather than trusting the citation on top of it. Several of the most repeated figures in business messaging turned out to have no primary source at all.

What follows is what survived. Every figure here is attributed to the place it actually came from, and linked wherever a stable primary URL exists: an SEC filing, an earnings call transcript, a platform's own developer documentation, or a published dataset with a stated methodology. Where a number is real but fragile, we say so. Where we could not verify something, we left it out and listed it at the end so you know we looked.

The four rules we used, and why they eliminated so much

Benchmark posts are usually a chain of citations with no origin. Post A cites Post B, which cites Post C, which cites a 2019 slide from a webinar that is now a 404. We used four rules to break the chain.

Rule one: the source must be the party that measured it. Meta is the only entity that can count WhatsApp users. If a number about WhatsApp does not trace back to Meta, an SEC filing, or a company with server-side access, it is someone's guess with a chart around it.

Rule two: the unit must be stated. "1.3 billion users" is not a fact until you know whether that counts people who opened an app this month, accounts that were ever registered, or sessions. These are different numbers by a factor of three or more, and they get printed side by side constantly.

Rule three: the date must be attached forever. A 2025 figure is fine. A 2025 figure presented as "current" in mid-2026 is not. Messaging platforms move fast enough that a sixteen-month-old user count is a historical artifact, not a benchmark.

Rule four: if the publisher disclaims the metric, we report the disclaimer. This one removed more material than the other three combined. Several of the most-cited numbers in email marketing come with warnings from the companies that published them, and almost nobody quotes the warning.

The uncomfortable summary: the messaging benchmark market has a supply problem. There are perhaps a dozen genuinely primary numbers about business messaging in 2026, and there are thousands of blog posts. The arithmetic guarantees most of what you read is derived from very little.

Monthly actives by channel, and the unit problem that makes the table a lie

Here is the table everyone wants. Read the third column before the second, because the third column is the part that matters and the part that never gets printed.

Now look at what you just read. Four different units are stacked in one table: monthly actives, trailing-twelve-month actives, daily actives, and cumulative registrations. Two rows have no unit at all because the question does not apply. Anyone who ranks these ten rows by the middle column has produced a ranking of measurement conventions, not of reach.

The LinkedIn row is the clearest offender, and it is not LinkedIn's fault. LinkedIn reports "more than 1.3 billion members" and has always said members. Members means accounts that exist. A dormant account from 2013 belonging to someone who has not logged in since is a member. When a comparison chart puts that 1.3 billion next to WhatsApp's 3 billion monthly actives and calls both "users", it is comparing a cemetery census to a turnstile count.

The practical consequence is that channel-size tables should never drive your channel strategy. They are the least decision-relevant numbers in this entire post, and they are the ones that get screenshotted. Where your customers actually are is a question about your specific market and your specific customer list, which is why the Telegram versus WhatsApp comparison comes down to geography rather than features.

The 550 million figure is not X, and the filing says so twice

This is the most widely mis-stated number in social media right now, and it became mis-stated within about 48 hours of becoming available.

In May 2026, SpaceX filed an S-1 with the SEC ahead of its IPO. Because SpaceX had absorbed xAI, which had absorbed X in 2025, the filing contains the first management-disclosed X user metrics to appear in an SEC document since Twitter last reported in 2022. The number travelled fast. Almost every write-up rendered it as "X has 550 million monthly active users."

That is not what the filing says. The S-1 says:

"Our integrated AI platforms across Grok and X have over 1.3 billion supported accounts active in the last twelve months ended March 31, 2026, including approximately 550 million MAUs, up from over 1.1 billion supported accounts and approximately 520 million MAUs as of December 31, 2025. Of our MAUs, we had approximately 117 million MAUs that used Grok's AI features as of March 31, 2026."

The 550 million is Grok and X combined. The filing's own definitions section removes any ambiguity: MAU "refers to the total number of users who have interacted with Grok or X", and "in presenting combined MAUs across the two platforms, we seek to identify and account for users who access both Grok and X based on sign-in traffic so that such users are not double-counted." It also notes that "only users who have registered for an X or Grok account are included."

Read that carefully and the conclusion is unavoidable: because the two platforms are deduplicated into a single 550 million, X standalone must be smaller than 550 million. The filing never publishes an X-only figure. Anyone quoting 550 million as X's user count is quoting a ceiling as if it were a measurement.

Two more details from the same document that almost nobody carried. The first is sitting inside the quote above: only about 117 million of those monthly actives used Grok's AI features, roughly one in five. The AI half of the "integrated AI platforms" framing accounts for a fifth of the combined number, which tells you where the other four fifths come from. The second is that SpaceX distances itself from the metric in its own words:

"While MAUs provide an estimated measure of the size and engagement of our user base, we are focused on revenue and operating margin, and manage our business with the objective of driving sustainable revenue growth and profitability rather than with the primary objective of growing or maintaining MAU levels."

That is a company telling its future shareholders not to weight the number you are currently pasting into a slide. It is also worth noting what the filing quietly retires: the frequently repeated "600 million" figure never appeared in any filing. The audited-adjacent number, covering two products rather than one, came in below it.

None of this makes X a bad channel. It makes X a channel whose size you should describe carefully. If you run outreach or support there, the number that governs your day is your own DM volume and your API access tier, not a headline count. That is the practical layer we cover in the X CRM breakdown.

What Meta actually discloses about business messaging, quarter by quarter

Meta is the only company at this scale that discusses business messaging in enough detail to build a trend line from. The figures below all come from earnings call transcripts on Meta's investor relations site, which means they were said by named executives to investors under securities law rather than written by a content team.

The headline adoption number comes from Mark Zuckerberg on the Q3 2025 call:

"Every day, people have more than 1 billion active threads with business accounts across our messaging platforms ranging from product questions to customer support."

A billion active business threads per day is the single most important number in this post. It is the one that establishes that business messaging is not an emerging behaviour, it is the default behaviour, and it is measured server-side by the company that owns the servers.

The revenue trend underneath it is public too. On the Q4 2025 call, CFO Susan Li said paid messaging within WhatsApp was "crossing a $2 billion annual run rate in Q4", and that "click-to-message ads revenue growth accelerated in Q4 with the US up more than 50% year over year". A quarter earlier she had put click-to-WhatsApp ads at 60% year-over-year revenue growth.

The steepest curve is business AI. Track it across two calls:

Tenfold in one quarter. On the Q1 2026 call Li also reported Family of Apps Other Revenue at "$885 million, up 74%, driven primarily by WhatsApp paid messaging and subscriptions revenue."

Then the sentence that should worry anyone building a business case on it. Li noted that business AIs are "currently free for most businesses on our messaging apps", but that "as we make more progress, we expect that we will also work towards establishing a longer-term monetization model". Free today, priced later, on a platform where you do not control the rate card. Keep that in mind when you read the cost section below.

Open rate is a broken metric, and the company that publishes the benchmark says so

The most-quoted email table in the industry is Mailchimp's benchmarks page. It reports an all-industry average open rate of 35.63%, a click rate of 2.62%, and an unsubscribe rate of 0.22%, drawn from billions of delivered emails across campaigns with at least 1,000 subscribers.

Two facts about that table almost never travel with it.

The first is the date. The underlying data is from December 2023. It is being quoted in mid-2026 as the current state of email, which makes it about two and a half years stale in a period that included the Gmail and Yahoo bulk sender requirements and a general collapse in unauthenticated delivery. Nothing on the page claims to be current. Everyone quoting it supplies that claim for free.

The second is that Mailchimp disclaims its own headline metric on the same page: "The accuracy of email open rates may be impacted by Apple's privacy changes and their Mail Privacy Protection (MPP) feature, and this should be considered as you interpret open rate data."

Their support documentation is blunter:

"If a contact enables Apple MPP, Apple Mail will preload pixels, even if your contact hasn't opened the email, resulting in unreliable open metrics." And: emails in Apple Mail "are reported as 'opened,' regardless of the contact's activity, resulting in inflated and inaccurate open rates."

The mechanism is simple. Apple's proxy servers fetch the tracking pixel on the recipient's behalf, before and regardless of any human looking at anything. For any contact with MPP on, your reported open rate for that contact is effectively 100% forever. Mail Privacy Protection shipped in late 2021, which means the December 2023 data was already contaminated when it was collected. The 35.63% was never a measurement of humans opening email. It was a measurement of humans opening email plus Apple's servers pre-fetching images, blended at an unknown ratio that varies entirely with how many of your subscribers use Apple Mail.

Mailchimp's own recommendation is the correct one: "Clicks and purchases are stronger signs of engagement than opens, and aren't impacted by Apple MPP."

So here is the honest position on email open rates in 2026. There is no trustworthy open rate benchmark, there cannot be one while pixel pre-fetching exists, and the number you should compare against your peers is click rate or click-to-open on a list whose Apple share you know. If your board deck has an open rate line, it is measuring your subscribers' choice of mail client. Our own email inbox has no open tracking in it at all, which is less a principled stand than an admission that the number would not mean anything: what it shows you is the thread and where it stands, and you judge the relationship from whether people write back.

Why reply rate benchmarks disagree by a factor of twenty

Ask five vendors what a normal cold email reply rate is and you will get answers between 0.45% and 10%. That is not measurement noise. That is a twenty-fold spread on a single question, and it happens because they are quietly answering different questions.

Start with the cleanest dataset we found. Belkins published a study covering 7,530,489 emails sent between January and December 2025, producing 34,393 tracked replies. Their definition is stated explicitly: unique replies divided by emails sent, excluding auto-replies and bounce notifications. Their characterisation of the traffic is stated too: strict cold outreach to net-new contacts.

Their answer is 0.45%.

They also disabled open tracking for the year, which is a methodologically serious decision, because it removes the temptation to compute reply rate against a pixel-inflated denominator. The segment detail is where it gets useful:

Read the top and bottom rows together: a ten-person company replies at roughly 3.3 times the rate of a ten-thousand-person company. Company size moves your reply rate more than any subject line ever will. So does seniority, but in the opposite direction from the one most playbooks assume: founders reply more than VPs, because founders are the company and VPs have gatekeepers and 400 unread.

Now the arithmetic that explains the twenty-fold spread. Belkins recorded 34,393 replies against 7,530,489 sends. To report an 8.5% reply rate from that same reply count, you would need to divide by roughly 404,600 instead: a denominator about 5.4% the size of the real one.

Nobody is lying. They are dividing by something else. Common denominators in circulation include emails delivered, emails opened (pixel-inflated, see above), contacts in the campaign rather than messages sent, or a filtered subset described as "cold" that includes warm intros and prior touches. Instantly publishes ranges an order of magnitude higher, calling 5% to 10% solid for B2B and 10% to 15% excellent, while openly conceding why the published ranges conflict: "'Cold' sometimes includes warm intros or prior touches. List quality and verification differ by study and sender." That concession is the whole story. Two vendors can both be honest, count the same event, and land twenty-fold apart, because one is dividing by every address it touched and the other by a filtered, verified, warmed subset.

The rule that follows is short. A reply rate without a stated denominator is not a number. Before you accept any benchmark, including ours, ask what was on the bottom of the fraction. If the answer is not available, the top of the fraction does not matter. The tactical version of this argument is in what actually gets a reply in a cold DM.

The cross-channel response table we are willing to sign

This is the table this post exists to publish. The last column is the point: it tells you how much weight the row can carry. We would rather hand you six defensible rows and four honest blanks than ten confident inventions.

Three of those rows need their footnotes read out loud.

The LinkedIn 13% is a policy, not an average. LinkedIn's Recruiter documentation states that recruiters "must keep their InMail response rate at or above 13% on 100 or more InMail messages sent within every 14-day assessment period", and that falling below it lands you in an InMail Improvement Period where bulk InMail is disabled for two weeks. That is not LinkedIn telling you what normal looks like. It is LinkedIn telling you what it considers bad enough to switch you off. It is still the most useful LinkedIn number in public, because it reveals where the platform draws the line, and it implies competent senders clear it comfortably. Note the shape of the incentive: LinkedIn is policing response rate because response rate is the thing that decays when a channel gets flooded.

The WhatsApp 98% should be retired. It is the most repeated statistic in business messaging and it traces back to early marketing copy with no published methodology, no dataset size, and no definition of "open". Nobody who repeats it can tell you what the denominator was, which by the rule above means it is not a number. It is also unnecessary, because unlike email, WhatsApp read data is genuinely observable: the Cloud API fires separate webhooks for sent, delivered, and read on every message you send. You can compute your own delivered-to-read ratio from your own logs today, for free, with a real denominator. Your number will land below 100% partly because recipients can switch read receipts off entirely, and it will be worth more than the 98% ever was because it will be yours.

The DM row is the honest one. There is no primary, methodologically stated reply rate benchmark for Telegram, Instagram, or X direct messages. Not a stale one, not a bad one. None. The platforms do not publish it and the vendors who could will not. Every DM reply rate you have ever read was either someone's private campaign data presented as an industry average, or invented. This is a real gap in public knowledge and we are not going to fill it with a guess.

The regional split: WhatsApp wins 70 of 100 countries, and that is the boring part

Global platform totals conceal the only thing that matters, which is that messaging is not a global market. It is roughly 100 national markets that happen to share app store infrastructure.

Similarweb's March 2025 study of Android app data across 100 countries found WhatsApp ranked first in 70 of them, with 1.18 billion yearly downloads and installation on 84.02% of devices in its markets. It also reports 1.26 billion daily returning users and users opening the app roughly 20 times a day. WhatsApp's dominance is not narrow: it is the top messenger across most of Latin America, Europe, Africa, and South Asia.

The interesting part is the other 30.

Telegram's map is the one worth studying, because it explains why the app looks enormous to some teams and invisible to others. Telegram is not a smaller WhatsApp spread evenly across the world. It is highly concentrated, and its strongholds cluster in Eastern Europe and Central Asia. If you sell in Kazakhstan, Telegram is not a channel to consider, it is the channel. If you sell in Brazil, Telegram is a rounding error next to WhatsApp no matter what the global billion-user number says.

This is why "which channel should we be on" has no general answer and why the global MAU table at the top of this post is close to useless for the decision. The correct method is to look at where your existing customers already are, which you can read directly off your own contact list. Telegram also has a second concentration that does not show up in country data at all: it is disproportionately the messenger of crypto, trading, gaming, and developer communities everywhere, including in countries where its national share is trivial. Group and channel culture drives that, not geography. That is the reason a Telegram CRM makes sense for a Berlin trading community and no sense for a Berlin dentist.

One caveat on the Similarweb data worth stating plainly: it is Android-only, and it is from March 2025. Android-only means it systematically understates iMessage, which is the actual default messenger in the United States among iPhone users and appears nowhere in the ranking because it cannot be measured this way. Any messaging map that shows WhatsApp winning the US is measuring the Android half of the country.

Cost per conversation: one channel is metered and one is free, and it is structural

Channel economics get discussed as if the differences were small and negotiable. They are neither. Two of the largest messaging channels on earth have opposite billing models, and that difference will shape your strategy more than any benchmark in this post.

Telegram's own bot documentation states the position without qualification: "By default, bots are able to message their users at no cost", with the only caveat being limits "on the number of messages they can broadcast in a single interval". There is no rate card. There is no per-message fee. There is no conversation window. The constraints are rate limits, not invoices:

• In a single chat, no more than about one message per second.
• In a group, no more than 20 messages per minute.
• For bulk notifications, roughly 30 messages per second, unless paid broadcasts are enabled.

Thirty messages per second, free, is 108,000 messages an hour. Most companies reading this will never touch that ceiling. If you do, paid broadcasts cost 0.1 Telegram Stars per message above the free 30 per second and raise the limit to 1,000 per second, but the qualification bar is high: a bot needs at least 100,000 Stars on its balance and at least 100,000 monthly active users. In other words, Telegram only starts charging you at a scale where you are unmistakably a large broadcaster.

WhatsApp is the opposite by design. Per Meta's pricing documentation, "effective July 1, 2025, Meta charges on a per-message basis", replacing the older conversation-based model. Charges land on delivery, not send, and only template messages are billable. Rates vary by template category and by the recipient's country calling code, published in per-market rate cards. We are deliberately not quoting a rate here: they differ by market by more than an order of magnitude and they change, so quoting one number would make this post wrong somewhere and stale everywhere. Go read the rate card for the countries you actually sell into.

The structure, which does not change, matters far more than the rate:

Note the asymmetry inside WhatsApp's own model. Utility and authentication messages qualify for volume-based discounts as you send more. Marketing messages do not: every one bills at full rate, forever. Meta has priced its network so that transactional messaging gets cheaper with scale and promotional messaging never does. That is a pricing sheet expressing a worldview, and the worldview is that you should stop sending marketing blasts.

The free windows are where your messaging bill is actually decided

This is the part of WhatsApp economics that most teams never model, and it inverts the usual assumption that cost scales with volume.

Meta's documentation defines two windows. A 24-hour customer service window opens when a user messages your business, during which you can send non-template messages at no charge. Separately, a 72-hour free entry point window opens when a user reaches you through a Click to WhatsApp ad or a call-to-action button and you respond within 24 hours. Inside that window, per Meta's own wording, "you can send any type of message to the user at no charge."

Work the consequence through with round numbers. Say you generate 1,000 conversations a month from click-to-WhatsApp ads.

• You reply inside the window. All 1,000 conversations, including any follow-ups within 72 hours, cost zero in messaging fees. Your entire WhatsApp bill for the month is the ad spend you already budgeted.
• You reply on day four. The free entry point window has closed. Every one of those 1,000 conversations now needs a paid template to reopen, at marketing rates, with no volume discount available.

Same leads, same headcount, same ad spend. The difference between a zero messaging bill and a four-figure one is response time. Not copy, not targeting, not tooling. Response time.

On WhatsApp, speed to lead is not a conversion tactic. It is a billing mechanism. Meta has made slow replies literally more expensive than fast ones, and almost nobody has this line in their model.

That is a rare case of platform incentives pointing the same direction as good practice, and it stacks on top of the conversion effect, which is the subject of why the first five minutes decide the deal. The same reply that wins the deal also happens to be the free one.

It also reframes what automation is for. The usual argument for an autoresponder is customer experience. The WhatsApp-specific argument is that an instant first reply opens a window in which everything else you send is free. Any first reply does this, whether a person types it at 2am or an AI agent answers the inbound DM in seconds, and if a human takes the conversation over an hour later they are still inside the window the first reply opened. The billing does not care who was fast, only that somebody was.

One honest caveat on Telegram, since we sell a Telegram product and it would be convenient to leave this out. "Free" applies to the Bot API. If you operate through user accounts over MTProto rather than a bot, you are subject to a different and much less forgiving set of limits, where aggressive sending triggers flood waits and, past a point, account restrictions. The message cost is still zero. The risk is not. That tradeoff, and how to pace around it, is covered in avoiding Telegram bans.

The case against being on every channel, from a company that sells every channel

The obvious conclusion from a post full of billion-user numbers is that you should be everywhere. We sell software for being everywhere, so that conclusion is commercially convenient for us. It is also wrong for most teams, and the numbers in this post are what make it wrong.

Start with the staffing arithmetic. Tidio's data puts the average operator at 29 conversations per day and the average first response at 1 minute 35 seconds. Those two numbers are linked. A person sustains that response time because they are watching one queue. Give the same person six queues and you have not created six times the capacity. You have created five extra places for a conversation to sit unanswered while they are looking somewhere else.

Now add the window mechanics. On WhatsApp, a channel you check twice a week is not merely a slow channel, it is a channel that generates a bill, because the 24-hour service window closes and reopening costs a paid template. A neglected channel has negative unit economics, not neutral ones. On Telegram, neglect costs nothing in fees but produces exactly the same silence.

Then add customer expectations, which are moving against you. Zendesk's CX Trends 2026, based on responses from more than 11,000 consumers and CX leaders across 22 countries, reports that "88% of customers expect faster response times than they did just a year ago" and that "74% of consumers now expect customer service to be available 24/7". Two caveats you should carry with those numbers: the fieldwork was done in June 2025, so a report labelled 2026 is describing what people said a year ago, and survey data measures stated preference rather than behaviour. Nobody has ever told a researcher they are happy to wait. Treat the exact percentages as directional. The direction is not in doubt.

Put those together and the conclusion reverses. Every channel you add without staffing it lowers your average response time, raises your costs, and adds a surface where customers are ignored in public. Two channels answered in ninety seconds beat six channels answered in a day, and it is not close.

The honest version of the recommendation:

• Pick channels from your contact list, not from a MAU table. Export your customers. Count where they already message you. That is your channel strategy, and it is already written.
• Add a channel only when you can answer it inside its window. If you cannot commit to a 24-hour WhatsApp response, do not open WhatsApp.
• One inbox is a staffing fix, not a strategy. Consolidating six queues into one screen genuinely helps a small team hold a response time, which is the actual argument for a unified inbox. It does not conjure attention out of nothing.

And the part we have a commercial interest in not writing. If your customers are US consumers who want to text a phone number, we are the wrong product: we have no SMS channel, no phone channel, and no iMessage. If your customers are enterprise buyers who live in Outlook and have never sent a DM in their lives, a messaging-first CRM is solving a problem you do not have, and you should buy something built for email and calendars. We would rather tell you that here than after you have migrated.

What we could not verify, and what nobody publishes

The gaps are as useful as the figures, because they tell you which confident claims in your feed are unsupported. Everything below is something we actively looked for and did not find.

That last row deserves a moment. Of every number in this post, exactly one carries the liability of a securities filing behind it, and it is the one about the smallest platform. WhatsApp's 3 billion and Instagram's 3 billion were said out loud by executives on earnings calls. Telegram's billion was a post by its founder. These are probably all roughly true. "Probably roughly true, asserted by an interested party, with no methodology" is nonetheless a different evidence class from a number a company has to defend in a registration statement, and the gap between those classes is invisible once the numbers are sitting in the same bar chart.

There is a structural reason for the void, and it is worth naming. Every party who could measure DM reply rates has a reason not to publish them. Platforms would be publishing a number that advertisers would use against them in negotiations. Vendors would be publishing a number that is either unimpressive or would invite the methodology question they cannot survive. So the space fills with claims that sound like data and are not, and they propagate because a benchmark post needs a table and a table needs cells.

The benchmark that beats every number in this post is your own

Here is the turn. You have just read several thousand words of sourced industry data, and the correct use of it is to stop relying on industry data.

Every public benchmark suffers from the same defect: it is an average over a population you are not in. Belkins' 0.45% is agency-run B2B outbound. Tidio's 1 minute 35 seconds is SMB live chat. Mailchimp's click rate is opt-in bulk email from December 2023. None of those populations is your customer list. Meanwhile you are sitting on a dataset with perfect coverage of exactly the population you care about, and it costs nothing to compute.

Four measurements, each with a denominator you control:

That last row is the one nobody tracks and everybody should. It is where customer experience and cost per conversation turn out to be the same number viewed from two angles.

The rest is routing. Once you know which channels your customers actually use, connect those and leave the rest closed. Route each channel into the pipeline that owns it, so an inbound Telegram message and an inbound email do not land in the same undifferentiated pile. Use contact records to hold channel history in one place, because the same person messaging you on two channels is one lead, and counting them twice is how a pipeline starts lying. Then read your own numbers rather than someone else's in a blog post, including this one.

Do this for one quarter and you will have something no benchmark post can give you: a set of numbers about your own customers, with denominators you wrote down yourself.

Frequently asked questions

What is the single most reliable business messaging statistic in 2026?

Mark Zuckerberg's statement on Meta's Q3 2025 earnings call that "every day, people have more than 1 billion active threads with business accounts across our messaging platforms." It is measured server-side by the company that owns the servers, and it was said by a named executive to investors. That combination is rare. Meta does not break it down by app, so treat it as a portfolio figure.

Does X really have 550 million monthly active users?

No. The SpaceX S-1 filed with the SEC in May 2026 reports approximately 550 million MAUs for X and Grok combined, deduplicated by sign-in traffic, counting registered accounts only, as of 31 March 2026. Because the two products are merged into that single figure, X standalone is necessarily lower. No X-only number has been published, and the widely repeated 600 million never appeared in a filing.

Is WhatsApp's 98% open rate real?

There is no published methodology behind it, no stated dataset, and no definition of "open". It traces back to early marketing copy and has been repeated ever since. You do not need it: the WhatsApp Cloud API fires separate sent, delivered, and read webhooks for every message, so you can compute your own read rate from your own logs with a denominator you can defend.

What is a good cold email reply rate in 2026?

Ask what the denominator is before accepting any answer. Belkins measured 0.45% across 7,530,489 emails sent in 2025, defined as unique replies divided by sends, excluding auto-replies, on strict cold outreach. Figures near 8% typically divide by something much smaller, such as opens or a filtered subset. Both can be honest. They are not the same metric.

Why is Telegram free to message on and WhatsApp is not?

Different business models. Telegram's Bot API documentation states bots can message their users at no cost, constrained by rate limits rather than fees, roughly 30 messages per second for bulk sends. WhatsApp has charged per message since 1 July 2025, priced by template category and recipient country, with free service windows. Telegram monetises Premium subscriptions and ads. Meta monetises the business messaging itself.

Which channel should my business actually be on?

The one your customers already message you on, which is a question about your contact list rather than about global user counts. Export your contacts and count the channels. Telegram is dominant in Kazakhstan and a rounding error in Brazil, and no worldwide MAU table will tell you which of those you live in. Add a channel only when you can answer it inside its response window.

Where to start

Pick the one channel your customers use most, measure your median first response time on it for two weeks, and write the number down. It will probably be worse than you expect, and it will be more useful than any benchmark in this post, because it will be about you rather than about an average of strangers.

If you want that measurement to happen across Telegram, WhatsApp, Instagram, X, email and live chat without stitching six dashboards together, that is what a unified inbox is for. There is a free plan, so you can measure your own response times before deciding whether any of this is worth paying for: see the plans.