In Q3 2025 we finished a 12-week engagement with a mid-market software company, delivered a diagnosis I still believe was correct, and lost the account 4 months later. The work was not wrong. Nothing we recommended has since been shown to be a bad idea. What we got wrong was that we optimised the whole engagement for being right, and being right is not the same as being adopted.
I have thought about this one more than any other piece of work we have done, because every quality signal we had was green throughout.
What the engagement looked like
The client's problem was that their pipeline had stalled and nobody could agree why. Sales blamed lead quality, marketing blamed follow-up, and the founder suspected the product had drifted from the segment it was built for.
We did the work properly. Over 12 weeks we sat in on 22 sales calls, read every closed-lost record in HubSpot for the previous 3 quarters, interviewed 9 customers including 4 who had churned, and built a picture that was uncomfortable and, I think, accurate. The founder was closest to right. The product had drifted, the messaging had followed it, and the leads were fine.
We presented it in a good deck. The client accepted the finding. Everyone in the room agreed. We were paid and thanked, and the engagement closed on schedule.
What happened next
Almost nothing. Over the following 4 months roughly 2 of our 11 recommendations were implemented, both of them the cheapest ones. The messaging work never started. When we followed up, the answers were the ones you would expect and all of them were true: a funding round took priority, the head of marketing left, a large customer needed attention.
Then they hired a different firm to do a project that overlapped substantially with what we had already told them.
I went back and checked whether this was a pattern or a single bad outcome. Across 11 engagements we had closed over 18 months, I scored each one on whether a specific recommendation had visibly shipped within 90 days of the final presentation. 4 had. The 4 that shipped had something in common that I had never treated as a variable: in every one of them, we had made a change to something live during the engagement itself rather than only at the end. The 7 that did not ship were, on average, the ones with the better decks.
That was the part that made me go back through everything. It is easy to read that as a client who did not listen. I no longer think that is what happened.
What we did not understand at the time
We had delivered a conclusion to a room, not a change to an organisation, and those require completely different work.
Our engagement had one moment of maximum energy, which was the final presentation, and it was positioned at the very end. Everything before it built toward being right. Nothing after it existed, because the engagement was over. We handed a correct diagnosis to a group of people who then had to go and do the hard part alone, with no external pressure, no sequencing, and no one whose job it was to make the first thing happen.
The firm they hired afterwards did something structurally different. I know this because the client told me later, without any apparent awareness that it was a comment on our work. That firm started with one small change, shipped it in the first fortnight, and used it to build the case for the larger one. They were less right than we were at the beginning and considerably more effective by the end.
There is a second thing, which is harder to admit. Our deck was persuasive to the room and useless to everyone outside it. When the head of marketing left, our finding left with her, because it lived in her understanding of a presentation rather than in any artefact the organisation could act on. A 40-page deck is not a durable object in a company. A rewritten page, a changed script, a new qualification question in the CRM, those are durable.
What we changed
Three things.
First, every engagement now ships something in the first 3 weeks. Not a finding, a change. Something live, small, and reversible. It is almost always less important than the eventual recommendation and that is fine, because its job is to establish that this engagement produces changes rather than documents.
Second, we sequence recommendations by who has to do them, not by impact. Our old decks ranked by expected value, which produced lists where the top item required 3 departments and a quarter. Now the top item is whatever the person in the room can start on Monday without asking permission.
Third, the deliverable is no longer the deck. The deck summarises. The deliverable is the changed artefact itself, in their systems, with their team, done together. That is slower and it is the only version that survives a personnel change.
What I would tell another consultant
The failure mode here is specific to people who are good at diagnosis. If you are confident you can find the real problem, the engagement naturally organises itself around demonstrating that you found it, and the demonstration becomes the product. Clients will reinforce this, because being told the truth about their business is genuinely valuable and they will say so warmly, right up until nothing changes.
The check I use now is simple and it is uncomfortable to apply. At the halfway point of any engagement, ask what has actually changed in the client's business as a result of the work so far. If the honest answer is that they understand something better, you are building a document, and there is a decent chance it will sit exactly where our 40 pages sat.
We still do the diagnostic work as rigorously. We just no longer treat it as the thing we are selling. How we structure engagements around that now is in our founder funnel approach.
