Nearly every piece of advice on personalising cold email was written for someone sending at a volume you're nowhere near. Merge fields, city mentions, a note about company size dropped into a template: those are the ceiling for a team sending thousands a day, because at that volume a human being cannot look at each recipient individually. At thirty a day, a human being can. The mistake is not personalising too little at that volume. It's personalising the wrong parts, and skipping the one that actually works.
The cheap kind reads as fake precisely because it scales
A first-name field, a mention of the recipient's city, a line about their industry inserted from a spreadsheet column: these exist because they can be produced automatically for ten thousand people at once. Recipients have seen that pattern enough times now that most of them recognise it on sight, consciously or not, and reading past it costs them nothing. Reproducing that same shallow pattern by hand at thirty a day gets you zero credit for the one thing you actually have that a high-volume sender doesn't: time. You did the work a template does for free, and it reads exactly like the template did it.
Spend the time on the one line a template can't produce
What changes at low volume is that you can afford to find and state the actual, specific reason you're writing to this company today, something true of them that wouldn't be true of the next name on the list. Not their industry. Not their headcount. Something that happened: they posted a job for a role your product supports, they shipped a self-serve pricing page last month, they wrote a public post describing the exact problem you solve, in their own words, three weeks ago.
That single sentence is the entire personalisation budget worth spending at thirty a day, because it's the only part a mail-merge genuinely cannot fake. A recipient can tell the difference between "I noticed you're growing fast" and "I noticed you posted for a first support hire last Tuesday" without needing to think about why. The second sentence required a human to look at them specifically, today, and everyone can feel that, even people who couldn't articulate the rule.
Personalise the reason, not the offer
A useful structure at this volume: one sentence specific to the recipient, stating the real reason you're writing to them and not the next name on the list, followed by a paragraph that's identical for everyone in the batch, describing what you do and what you're asking for. Founders often do the opposite instinctively, rewriting the pitch itself for each person, which burns the time you have on something the recipient can't actually tell was customised, because the offer sounds roughly like an offer either way. The reason is what proves a human looked. The offer doesn't need to prove anything; it just needs to be clear.
Where personalisation stops paying, even at low volume
There's a failure mode specific to founders who do have the time: spending twenty minutes finding something to compliment. Their blog post, a nice bit of design on their site, a talk they gave at a small conference. That isn't personalisation about the buying decision. It's flattery, and most recipients read it as exactly that, because it has nothing to do with why you're writing or what you're offering. It doesn't cost you credibility so much as it costs you twenty minutes that could have gone toward confirming the one thing that actually matters: whether this person is still in the role, and whether the company still matches the segment you described. That check takes less time than finding a compliment and it's the check that keeps a list from going stale.
A before and after, so "the one true sentence" isn't abstract
Before: "Hi {{first_name}}, I saw you're a growing company in the logistics space and thought you might be interested in what we're building." That sentence is grammatically personalised and substantively generic. It would survive being sent, unedited, to four hundred companies, which is exactly the property that gives it away.
After: "Hi Dana, I saw the posting for a dispatch coordinator role you opened last week." Same length, same effort to read, and completely different in what it proves. The second version could only have been written by someone who looked at this specific company on this specific week, and that fact does more work than any adjective in the first version ever could.
Where the one true sentence actually comes from
At thirty a day you don't need a research process, you need a two-minute habit: check whether the company has a recent job posting, a recent product announcement, a recent changelog entry, or a public post that names the problem you solve. Any one of those is enough to write the specific sentence. If none of them turn up anything in two minutes, that's useful information too. It usually means the company hasn't hit whatever trigger makes them a live buyer yet, and the honest move is to send them fewer resources today and come back to them once something has actually happened, rather than manufacturing a sentence that isn't really true.
The volume you're at tells you which mistake you're at risk of
At thousands a day, the risk is a message that reads as templated because it structurally has to be. At thirty a day, the risk runs the other way: an hour spent polishing an offer nobody asked to be polished, on a message that never actually goes out because there's always one more detail to check. Low volume isn't a smaller version of the same problem high volume has. It's a different problem, with a different fix, and the fix is to write the one true sentence, keep the rest boringly consistent, and stop there.
This is also the shape a business built for thirty a day should be sending in, rather than pretending it's running a campaign at a scale it isn't. Rocketship's worker writes from your own inbox, one recipient at a time, which only works because someone, or something, actually looked at each name first. At thirty a day, that someone should be spending its attention on the one sentence that matters, not on the merge field nobody was fooled by anyway.