September 1, 2026 · 5 min read
What 369 HubSpot briefs actually asked for
I had been saying roughly 40% of the work out there is fixing what the last vendor built. So I counted. It came back lower — and something else in the data mattered more.
For about eighteen months I have been answering inbound briefs on a HubSpot services marketplace, and somewhere in there I started saying a number out loud. Roughly forty percent of the work available, I would tell people, is rescue work — fixing what a previous vendor already built and charged for. It is a good line. It matched what I was seeing. I had never actually counted it.
So I counted. Not the briefs I responded to, which would only describe my own judgement, but every brief the marketplace has shown me, including the two hundred and fifty-eight I passed on. Three hundred and sixty-nine in total.
What counting them involved
Of the 369, two hundred and seventy-four carry enough text to read properly. The rest are a title and a budget band, which is not nothing — these titles are unusually descriptive — but it is not a brief either, and I have kept the two apart everywhere it matters.
I classified them by hand and by pattern, and the pattern half was wrong twice before it was right. The first version put ninety-five percent of briefs in one category and ninety-three percent in another, because it was matching stray words rather than the subject: the word month alone was enough to classify a brief as ongoing work. The second version was better but still called something an acquisition sixty percent of the time it appeared, because in this corpus acquisition usually means paid acquisition, and merged usually means duplicate records.
What fixed it was writing down what I thought ten briefs meant before rewriting the classifier, then grading the classifier against that rather than adjusting it until the output looked sensible. And then reading every classification that mattered by hand, against the buyer's own words, and throwing out the ones that did not survive. Eight of forty did not.
The number I went looking for
Rescue work is the largest identifiable reason a buyer posts. Where a brief says why it exists at all — inherited a portal, a previous agency, a build that was never finished, a partner who left mid-project — that reason is rescue work about a third of the time. Thirty-four percent.
Which is close to what I had been saying, and the denominator is much narrower than the way I was saying it. Only a minority of briefs explain why they exist. Most describe what they want built and leave the cause out entirely. So the honest sentence is not "forty percent of the market is rescue work." It is "where a buyer says what went wrong, a third of the time what went wrong was the last vendor." That is a smaller claim, and it is one I can show you.
The finding I was not looking for
The briefs are sequentially numbered, so they sort by time without me having to guess at dates. Splitting them into four periods and asking a simpler question — is this buyer asking someone to build something, or to run something — produced the clearest signal in the whole dataset.
- Earliest period
- 76% asking for a build. 14% asking for someone to run it.
- Early-middle
- 68% build. 23% run.
- Middle
- 59% build. 32% run.
- Most recent
- 53% build. 42% run.
Requests for someone to operate the system rather than deliver it and leave have tripled. Over the same span, briefs describing an ongoing administrator — a person to hold the portal week to week — went from five percent to twenty-eight. And integration work sat at roughly sixty percent in every single period, which is the other half of the finding: it never moved, so it is not a trend, it is the floor. Everyone needs it and nobody is differentiated by it.
The thing I would be careful about is reading this as a market that only wants operators now. Over exactly the same period, briefs about data modelling and CRM architecture went from fourteen percent to thirty-five, and revenue operations work — quotes, invoicing, renewals, commissions — went from twenty-six to thirty-five. Both halves are growing. What is shrinking is the middle: the generic build, delivered and handed over, with nothing underneath it that needed designing and nobody staying to run it.
Where these numbers stop
What the data cannot tell you
The reason-for-posting counts are a floor, not a rate. The detector that finds them misses in one direction only, so more briefs state a cause than I counted. The time comparison uses posting order as a stand-in for dates, and older postings are likelier to be short, which biases the early periods. And the marketplace does not reveal who posted a brief until you respond to it, so this is a picture of demand with no people in it — I can tell you what was asked for and never who asked.
There are also numbers in here that look like findings and are not. Seventy-nine percent of briefs mention automation and workflows. Sixty-one percent mention integration. Those are not insights, they are the vocabulary of a HubSpot brief, and quoting them would make the analysis sound more conclusive than it is. The narrow ones carry the signal.
What I would take from it
If you are buying this kind of work, the shift from build to run is worth knowing before you write your brief, because the two need different people and the market is quietly rewarding one of them. And if a previous vendor is the reason you are looking, you are in the largest single group of buyers who bother to say why — which at least means the shape of your problem is well understood, even when it does not feel that way from inside it.
For my part, I went looking to confirm a number I liked and found it was a bit smaller and much more specific than I had been saying. I would rather have the smaller one.