Monday 20 July to Sunday 26 July 2026. The quietest sending week of the engagement, and deliberately so. This was the week the targeting was rebuilt from scratch: a new lookalike list was built from the sixteen people who had actually replied positively, three new audience theories were tested at small scale, and the hiring-signal rail agreed on the 20 July call was documented and automated. One reply came back. The week that followed produced thirty two.
Published late. This report covers 20 to 26 July and should have gone up on 27 July. It is being published on 3 August alongside the Week 3 report, so the record on this portal is complete. Every number in it was pulled live from the sending platforms on 3 August, not written from memory.
This is the week that looks worst on a chart and mattered most in practice. 888 emails, 547 new people, one reply. There is no way to present that as a strong sending week and this report is not going to try.
What was actually happening: the June-sourced lists had been shown the previous week to be the source of the bounce problem, and the replies coming back were not concentrated in any audience we could name. So rather than pour more volume through targeting we could not defend, the week was spent rebuilding it. The new list was seeded from the sixteen people who had actually replied positively to Ripple outreach, pulled live from the pipeline rather than from anyone's memory, and then expanded to companies that look like them.
Alongside that, three new audience theories were put into market at small scale on 24 July to find out which of them was worth real volume. The answer came back quickly and it was mostly negative, which is exactly what a cheap test is for.
Five campaigns went live and the entire targeting method was rebuilt. None of this shows up in a sending chart, and all of it is what the following week ran on.
The call on 20 July produced a specific targeting idea: companies advertising for customs, trade compliance or classification staff are, by definition, telling the market they have more of that work than they can handle. That signal was written up and automated into a repeatable sweep rather than left as a note. It became segment C in the week that followed.
Instead of guessing at an audience, the new list was built backwards from evidence. Every positive reply Ripple outreach had ever produced was pulled live from the pipeline, sixteen named people with titles and companies, and the shape they shared became the filter. Names in that seed set include Jeff Spengler at Lennox, Adrienne Braumiller of Braumiller Law Group, Hardy Haenisch at ARC Global, Cindy DeLeon at DeLeon Trade and Deepak Yadav at Sharp Electronics USA.
That method is the reason the following week's main segment reached 2,687 people who looked like proven repliers instead of 2,687 people who merely matched an industry code.
| Test campaign | People reached | Replies | What it answered |
|---|---|---|---|
| International forwarders, no brokerage | 290 | 1 | Weak. Forwarders who do not clear customs in house have nothing to attach the pitch to. |
| Forwarders with in-house brokerage | 234 | 0 | Silent at this volume. |
| US customs brokers | 23 | 0 | Too small to read. Worked properly later, with a different message. |
Total cost of finding that out: 547 emails. The alternative was committing several thousand sends and a fortnight to an audience that does not do the work we automate.
Two fresh LinkedIn campaigns went live off the new list, one for Adrian's account and one for Jonny's, filtered to people who had actually posted recently so connection requests were not spent on dormant profiles.
Two merge-field bugs were also found and fixed the same day, one on each platform. Left alone, both would have sent messages with broken or empty personalisation to real prospects.
| Metric | Result | Read |
|---|---|---|
| Emails sent | 888 | The lowest of the engagement, by design |
| New people reached | 547 | All three test campaigns launched 24 July |
| Replies received | 1 | 0.18%. The honest low point, and the test result that redirected everything |
| Bounces | 6 | 1.10%, the cleanest week to date and the first sign the new lists were better |
| Metric | Result | Read |
|---|---|---|
| Connection requests sent | 36 | The new campaigns started mid-week |
| Connections accepted | 8 | Requests sent late in the week were still maturing |
| Messages sent | 5 | To those who had accepted |
| Replies | 3 | Three times email's reply count off a fraction of the touches |
One reply from 547 people is a bad number in isolation and a cheap answer in context. It told us that freight forwarders without in-house brokerage are not a buyer for this product, because classification is not work they do. That single finding removed a whole audience from the plan before any real money or sending reputation went into it, and it pointed the volume at the people who own classification instead. The bounce rate falling to 1.10% in the same week was the first evidence the new list source was cleaner than the old one.
A forwarder who outsources customs work to a broker has no classification problem to solve, because it is not their problem. 290 of them produced one reply. This is the finding that stopped the engine spending August on the wrong audience.
Sixteen real positive repliers, pulled live from the pipeline, turned out to be a far better description of the buyer than any industry filter. The segment built on that method reached 2,687 people the following week and produced 32 replies, against 1 from 547 the week before.
547 emails bought a definitive answer on three audience theories in five days. Running those theories at full volume would have cost weeks and put the sending domains at risk to learn the same thing. The low number in this report is the price of that answer, and it was worth paying.
Everything built in this week went live on Monday 27 July. The lookalike list became the classifier decision maker segment, the hiring-signal rail became the job-signal segment, and the internal champion angle was added as a third test.
The result is in the Week 5 report: 3,097 new people reached, 32 replies, a call booked with the CEO of Known and a direct call request from the CEO of NTMK Logistics. Read against that, this week is the reason it happened rather than a gap in the record.