The 98% has no study behind it. The best measured number is 79%, and even that is a floor.

Nobody knows, and that includes us. The best-documented figure we could find is 79%, from 13.9 million campaign messages across 382 brands, and even that number is a floor rather than a measurement, for a reason built into how WhatsApp reports reads.

The 98% you have seen everywhere has no study behind it at all. We looked.

01 / Where it comes fromWhere 98% comes from

Nowhere retrievable.

We traced it across every chain we could find. Every trail ends at a company that sells WhatsApp messaging software asserting the figure in its own marketing material. There is no report. No sample size. No date. No country. No method. No definition of what counts as an open.

Our working theory is that it is a transplanted SMS statistic, because a similar number circulates for SMS. That one is untraceable too.

This matters more than it looks. A statistic repeated by ten vendor blogs that all cite each other is unsourced, no matter how familiar it feels. Familiarity is not provenance.

02 / The best numberThe best number anyone has published

One vendor did the honest thing and published a measured figure with the method attached:

• 13.9 million campaign messages
• 382 brands
• 2026-03 to 2026-05
• 79.0% volume-weighted open rate

The same company states publicly that they REMOVED their own 98% claim, because no real account they could see was achieving it.

That is a usable number and we cite it. Nineteen points below the folklore, and still extremely good compared to email.

03 / Every figure is a floorThe part nobody says: every one of these numbers is a floor

Here is the mechanic, and it is the reason we do not treat any WhatsApp open rate as a measurement.

A WhatsApp "open" is not observed. It is reported. The platform fires a read webhook when the recipient's app confirms the message was read... and that webhook does not fire at all when the recipient has read receipts switched off.

Someone with read receipts disabled can read your message, act on it, walk into your shop, and never register as an open.

So every published open rate is:

(people who read it AND have read receipts on) ÷ (people it was delivered to)

The numerator is missing a whole category of person. The true read rate is higher than any figure you have ever been shown, and by an unknown amount.

And here is the gap we cannot close. We do not know what share of Malaysian WhatsApp users have read receipts switched off. Neither, as far as we can find, does anybody else... there is no published figure for it in this market. Anyone quoting an open rate to a Malaysian business is quoting a number whose largest error term has never been measured here.

We would rather say that than pretend the 79% is precise.

04 / The wrong metricWhy the metric is the wrong one anyway

Suppose the number were exact. It would still be close to useless as a business measure, for the same reason email open rate stopped being useful.

An open tells you a message was seen. It does not tell you:

• whether the person could reply without leaving what they were doing
• whether anyone on your side saw the reply
• whether the conversation reached a human, a booking, or a sale

We have watched all three of those fail on accounts where the open rate looked excellent. A message that is opened by 79% of recipients and answered by nobody who can act on it is a 79% open rate and a broken system.

05 / Measure this insteadWhat to measure instead

Three numbers, all of which you can read off your own account inside 30 days, and none of which require you to trust a vendor:

1. Reply rate, per message type. Not opens. Replies. This is the only number that proves a human on the other end did something.
2. Time-to-first-human-reply, on your side. If your bot answers in six seconds and a person answers in nine hours, your bot's performance is decorative.
3. The share of your traffic that lands inside the free 24-hour window. This one is money, not engagement. A reply inside the customer's own window costs nothing to send; starting a conversation from cold costs a template fee every time.

Every one of those is specific to your business, measurable without asking anyone's permission, and impossible for a competitor to have already published.

06 / What we doWhat we do with this

We will quote the 79%, with the sample and the window in the same sentence, because that is what a defensible number looks like.

We will not quote 98%, and if you see it on an agency's landing page it is worth asking them for the study. Most of the statistics used to sell chat automation trace back to a single vendor blog post with no sample size attached. That is a fair question to put to anyone, and it is a fair question to put to us.

The uncomfortable version: the reason the 98% survives is that it is more persuasive than the truth, and the people repeating it have never once been asked for the source.