Week 7 Performance Report

The most important thing that happened did not happen in a campaign dashboard. Adrian's call with John Moran at Rock Solid Logistics converted into commercial motion. In Adrian's own words: "no real interest in customs in the short term, but were interested in invoice processing and possibly customs next year. I have got them to agree to a short proof of value on invoice processing. NDA is with them to sign then we should deploy over the next month." That is the engagement's first deal in motion, and it came out of cold outbound.

One new call was booked. Keith Shavers is a licensed customs broker (LCB) at UMAX Shipping, and he came in through the thread we opened with UMAX's president. He told us plainly that all their entries currently run through third party brokers and that UMAX is not yet licensed to act as a customs house broker itself. That is a buyer describing his own bottleneck. His call sits this morning.

Now the part that matters most for the next three months. Two campaigns ran side by side this week carrying the same email. One was built from the official register of licensed US customs brokers. The other was built from a broader scraped list of brokers and forwarders. The registry list has produced every booked call this engagement has generated. The scraped list has produced zero opportunities from 3,219 people.

That is not a small difference in performance. It is the difference between the engine working and the engine not working, and it is now measured rather than assumed. The next section shows exactly how we know.

The broker and forwarder campaign ran four different first emails, split evenly. One of those four is word for word the same email that the registry campaign runs as its opener. Same sender, same week, same sending infrastructure. Only the list underneath is different.

The copy is not the variable. When the same words land in front of licensed customs brokers they get answered, and when they land in front of the scraped list they do not.

All 3,044 people on that campaign are tagged internally as customs brokers and forwarders. Their job titles say otherwise. There are 1,890 different job titles on a list that should have one kind of job title on it.

Real examples of people who received Ripple's customs classification pitch this week: an index equity portfolio manager, a director of wealth technology, a vice president of tax insurance, three vice presidents of talent acquisition, and a vice president of software engineering. None of them will ever file a customs entry.

The registry campaign was built differently. It comes from the CBP Permitted Customs Broker Listing and the FMC ocean freight forwarder and NVOCC register. Everyone on it is legally licensed to file customs entries. That is the entire difference between the two campaigns.

Two things fall straight out of this table. Brokers reply roughly four times better than importers, across every campaign we have run, which settles a question that has been open since June. And the bounce rate and the reply rate move together. A list that bounces is a list that does not reply, because both are symptoms of the same thing: the data is not who it claims to be.

Named, so there is no ambiguity about which lane produces pipeline.

There is a second lesson hiding in the middle row. The classifier campaign leads with an offer to send information, and it produces the best reply rate of any campaign we run at 1.49%. It has produced no calls. Asking someone if they want information gets you a reply. Asking for the call gets you the call. The registry campaign asks for fifteen minutes, and that is the one filling the diary.

Open tracking stays switched off on all your campaigns, because open-tracking pixels hurt deliverability, so there is no open rate here and there will not be while we send this way.

Cumulative since outbound began: 21,442 emails to 11,922 people, producing 100 replies and four booked calls. This week added a quarter of all emails ever sent on this engagement and a sixth of all replies.

Two LinkedIn campaigns went live on Friday 14 August, one from Adrian's profile and one from Jonny's, covering 96 people between them.

Two days is not a sample. The first real read on LinkedIn lands in next week's report, against a cleaned list.

Every one of those four names came from the licensed broker registry list. That is the clearest argument in this report for what the next batch should be built from.

LinkedIn URLs are only shown where the platform we ran the campaign on actually holds one. We never guess a profile link, because a wrong link on a live call is worse than no link.

Twenty one messages came back this week. Five were automatic out of office notes from two people and are not counted as replies. The honest split of the remaining sixteen, from twelve people, is below.

On Friday, the restarted LinkedIn campaign sent a message from Adrian's own profile to a VP at KlearNow.AI, an AI customs clearance platform. She replied politely to say KlearNow has built similar features into their own product and that they are not an importer. Adrian screenshotted it on Saturday morning with two lines: wrong audience, this is a competitor.

He was right, and the correct response was not to remove one contact. It was to ask how many others were in there.

Every company website still in that campaign, 348 of them, was then read and scored against your own buyer definition. Only 43% came back as a confirmed fit. That check now runs before any list is loaded rather than after messages have gone out.

Every campaign so far changed the list and the copy at the same time. When both move together, a change in results cannot be attributed to either one. That is why six campaigns of history had not settled a question as basic as which audience to target.

The only reason this week produced an answer at all is that one of the four test emails happened to be a duplicate of the proven one, which created a clean comparison by accident. We are not relying on luck again.

From here, one variable changes at a time.

We take the single email that produced three booked calls and change nothing about it. No new angles, no improvements. It runs against three different slices of the licensed register, in the same week, from the same inboxes.

Minimum 1,000 people per arm. Below that the result is noise, and every test on this account that has ignored that floor has produced an unreadable answer.

Whichever segment wins gets held constant, and three copy angles run against it at a thousand people each. Only then is a copy result meaningful, because the audience underneath it is no longer moving.

One honest caveat. At roughly one opportunity per 880 people, a single week cannot crown a winner on booked calls alone. The design accumulates over two to three weeks, and we will say so rather than declare a winner off one thin arm.

Reply rate stops being the headline number, because we now have proof it can be high while the campaign produces nothing. From next week these are the numbers reported:

Every reply is already tagged interested or not interested at the point it arrives, so this costs nothing to start reporting. It should have been the headline months ago.

1. Rebuild entirely from the licensed registers. We hold a database of 7,148 firms drawn from the CBP licensed broker listing and the FMC ocean forwarder and NVOCC register. Only 384 have ever been worked. 3,074 firms with a working website have never been contacted, and 4,977 people are already named on them. That is the supply the next several months come out of.

2. Let the scraped campaign drain, then retire the list. About 1,040 people are still mid-sequence with one follow-up left, roughly a day of sending. Those addresses already accepted the first email so the remaining sends carry almost no bounce risk, and stopping early recovers nothing. The action that matters is that this list is never loaded again.

3. Screen every list before it sends. This is now enforced in software rather than left to judgment. Before any list can be loaded, each company's website is read and scored against your buyer definition, and any list where fewer than 70% of companies come back a confirmed fit is blocked from loading at all. Run against this week's campaign it would have stopped it before a single email went out. It cannot be skipped or overridden without changing the rule itself.

4. Run the who test described above, with the proven copy frozen, three arms, a thousand people minimum per arm.

5. Let LinkedIn run a full week on the cleaned list. Both profiles keep sending, and next week's report carries the first honest LinkedIn numbers since the restart.

• Adrian: hold the Karan Virmani slot on Tuesday. Two reschedules already. He is the strongest pure customs fit in the pipeline and the third invitation needs to be the one that sticks.
• Adrian: tell us how the Keith Shavers call goes today. He has already described his own bottleneck in writing, so this is the cleanest qualified call the engine has produced.
• Adrian: confirm the Rock Solid proof of value once the NDA is signed. It is the first commercial outcome of the engagement and it should be on the record here, with dates.
• Adrian and Jonny: is there any reason NOT to test ocean forwarders and NVOCCs? That is 5,110 licensed firms we have never touched, and arm B of next week's test points straight at them. If you already know that audience does not buy, say so now and we will spend the volume somewhere better.
• Adrian and Jonny: send us any other competitor names you want blocked. Seventeen are blocked now. You know the market better than any list does, and a name from you takes about a minute to enforce permanently.
• Decision needed from Adrian: non US prospects. Seventy four people currently being emailed sit outside the US, in the UK, UAE, Nigeria, Mexico and Turkey. Your buyer definition is US import and export activity. Keep them or cut them, and we will action it the same day.
• Decision needed from Adrian: SGS and DFDS. Seventy one contacts across the two. Our check says neither has a customs classification function, but SGS does run a UK e-customs arm, so we would rather you called it than we guessed.