Coverage runs from 21.4% on funding to 3.2% on product

Every comms team has a folk theory about timing. Send it Tuesday morning. Never send on a Friday. Get in before the news desk fills up. We had our own version of it, and we had three years of our own data sitting there to check it against, so we checked it.

The answer surprised us, mostly because of where the effect turned out to live. What kind of news you are announcing moves the coverage rate by a factor of nearly seven. The hour you press send is worth a few percentage points on top, in a handful of specific cases. Both are useful. They are not the same size.

What we counted

We read 301 announcements, sent as 3,856 individual pitches to journalists between September 2023 and July 2026, and we labelled each one by the kind of information it actually carried. Three definitions do most of the work here.

An announcement is a release that leads with the company's own news. Op-eds, comment offers, journalist callbacks and cold introductions are a different job and sit outside this analysis. 213 pitches were reclassified out on the read.

One announcement is one row, not one pitch. A single release goes to as many as 112 journalists inside about fifteen minutes. Counting each of those as separate evidence turns one editorial decision into 112 observations and inflates a send-time window ninety-fold. Every confidence figure below is computed on announcements. The pitch count is only ever the volume of work behind them.

We read the release, not the label. The type recorded against a release when it was filed did not survive a read of the release itself. "Funding announcement" was the stored label on malware discoveries, product launches and an April Fools stunt. Every type below was assigned from the copy.

One workspace is held out of every figure. It runs on a different distribution model and logs a 27% open rate against a 70% average, so its 621 pitches would distort the numbers rather than add to them.

Funding earns coverage nearly seven times more often than product

A funding announcement earns coverage 21.4% of the time. A product launch, 3.2%. That is the same agency, often the same journalists, and a difference of nearly seven times - the widest single effect anywhere in this data.

The order holds inside individual client accounts as well as across the whole book. In every workspace that ran both a funding release and a product release, funding out-covered product. That is what makes this look like a property of the news rather than of one company's media relationships.

Product launches are the largest single block of work in the book - 784 pitches - and the weakest converter. That combination is the finding most likely to change how a team plans its year. Either find the funding, partnership or people angle to lead the release with, or cut the list hard and treat the launch as a briefing round rather than a broadcast.

Open rate does not predict coverage rate

Product announcements are opened more than any other type - 79% of the time - and turn into coverage in 4.0% of those opens. Funding is opened slightly less, at 73%, and converts 29.1% of them. Across the seven types with enough volume to read, the correlation between open rate and coverage rate is 0.04. There is effectively no relationship.

This is worth sitting with if open rate is on your reporting dashboard, because it flatters exactly the type that performs worst. A product launch that gets opened by four fifths of the list looks like a good week. It is not predicting anything about the coverage that follows.

Product announcements still do real work. They keep journalists current on what a company ships, and they are the reason a reporter recognises the name when the funding round lands. Send them. Just do not budget coverage against them.

Three in four replies produce coverage

The metric that does predict coverage is the reply. Once a journalist replies to an announcement, coverage follows 74.7% of the time - 549 replies, 410 pieces of coverage. And unlike opens, reply rate tracks what happens next: the types that earn the most replies also convert the most of them, correlating at 0.85.

Reading the table left to right shows where each type actually leaks. Funding barely leaks at all: 73.3% opened, 35.0% of those replied, 83.1% of those replied to produced coverage. Product opens best of anything in the book and then loses almost everyone at the reply - 7.8%.

Research reports leak at both stages, converting 12.4% of opens into replies and only 42.9% of those replies into coverage. That is the one row where a change of approach has two places to earn, and it is the type we are rethinking hardest as a result.

The practical version: report reply rate per open alongside coverage on every announcement, and treat a thread with a reply and no coverage as an open task rather than a closed one. A quarter of them end that way.

Wednesday before 09:00 CET is the strongest slot, at 29.4%

With every type pooled together, the week does have shape. The best slots are Wednesday before 09:00 at 29.4% and Monday midday at 18.9%. The weakest are Thursday and Friday afternoon, at 6.5% and 6.8%, against the 10.6% average.

One caveat to carry into the grid below: the before-09:00 row rests on four or five announcements per cell, so those figures will move the most as more announcements land. From 09:00 onward every cell holds between 14 and 39.

Before acting on any of that, it is worth knowing how settled each pattern is. We shuffled the send day and hour across announcements 5,000 times and asked how often chance alone produced a spread as wide as the real one. A low p value means the pattern is hard to get by luck.

Pooled across every type, day of week comes out at p = 0.83 and time of day at p = 0.77. In other words, the shape of that grid is the kind of shape chance produces. Part of the reason is composition rather than timing: the before-09:00 band is 39% funding and 32% partnership, the two strongest types in the book. Look within funding alone and the early send is 17.8% against 22.6% later in the day.

Two types are the exception, and both have a mechanism you can describe out loud.

Two send-slot effects with a mechanism behind them

Both effects sit at the edge of what this sample can resolve, and the threat research figures come from a single company's disclosures, which brings its own bias. What makes them worth acting on anyway is that the slot is free. Choosing Monday over Friday costs nothing, so the bar is expected value rather than statistical significance.

The two rules we have adopted: no threat research after Wednesday, and partnership news lands between noon and 2pm. Everything else goes when it is ready.

Funding, partnership and threat research all peak early in the week

Split the week by type and the picture sharpens. Funding sent on Monday earns coverage 27.7% of the time against 20.6% on other days - 34% more coverage from the same release. Partnership roughly doubles, at 21.9% against 10.8%. Product runs the other way and does best on Wednesday and Friday.

What the day choice is worth

Taking each type's best day and applying it to that type's whole year gives the raw numbers below. Read the two notes underneath before using them for anything.

How we are using this

Five things changed on our side as a result of reading this data, and all five are things any comms team can do.

1. Ask what kind of news this is before asking when it should go. The gap between the strongest and weakest type is nearly seven times. The day choice, once the winner's-curse correction is applied, is worth roughly 21 extra pieces of coverage a year on top of that.
2. Lead the release with the strongest angle it honestly has. If a product launch also carries a named customer, a funding round or a senior hire, that is the story. The launch is the second paragraph.
3. Report reply rate per open, not open rate. Opens correlate with coverage at 0.04. Replies correlate at 0.85 and convert at 74.7%. A thread with a reply and no coverage is an open task.
4. Default the four news-led types to Monday and Tuesday - funding, partnership, threat research and research report - and default product to Wednesday or Friday. Where an embargo fixes the date, use the table to decide which day to press for.
5. Aggregate to the announcement before making any timing claim. One release going to 112 journalists is one editorial decision, not 112 pieces of evidence. This is the single change that most altered what we thought we knew.

What this data does not tell you

It is one agency's book, weighted towards B2B technology, fintech, security and venture capital, with clients across the UK and Europe. The types with the fewest announcements - event, milestone, award, acquisition, people - carry two to nine announcements each, which is not enough to read a rate from, and we have marked them as such rather than quietly dropping them.

Coverage rates are observational. We did not randomise which day a release went out on, so a day-of-week figure carries whatever else travels with that day. The composition problem in the pooled grid is a live example of exactly that, and it is why the type-by-type view is the one to trust.

We are running the two timing rules as a tracked test over the next quarter, logging the send day and hour on every announcement, so the next edition can test these cells against a second year of data rather than a first.