This morning I published two thousand words about how to evaluate an AI vendor, including four questions that are explicitly about money. It does not contain my prices. It never was going to, and I want to write down why, because "put your prices on the site" is advice I get given often enough that I should be able to defend not taking it.
The short version is that I have a standing rule in my own marketing system: no NexBDM price appears in any public post. It is enforced by a script, not by my judgement, and it blocks a publish rather than warning about one. That is deliberate. A rule I can talk myself out of at seven in the morning is not a rule.
The objection, stated properly
I am not going to argue against a strawman here, because the case for publishing prices is genuinely strong.
It filters. Someone who cannot afford you leaves before either of you spends an hour finding out. It builds trust, because hiding a number reads as being ashamed of it. It answers the question people actually typed into the search box. And it is what I would tell most product businesses to do without hesitating.
I agree with all of that. It is why this is a decision rather than a preference.
What actually breaks
The problem is that I do not sell a product. I sell an amount of somebody's mess being taken away, and the mess is different every time.
Two businesses walk in with what sounds like an identical request. One has three systems that already talk to each other and a person who knows where everything lives. The other has eleven years of history in a shared drive, two spreadsheets that disagree about the same number, and nobody who can say with confidence which one the invoices come from. Same words in the enquiry, wildly different job.
If I publish one number, one of those two is being quoted wrong. Publish a range wide enough to cover both and the range is so wide it has told nobody anything. Publish the low end and I am doing the thing I dislike most in this industry, which is advertising an entry price that is not the price.
There is a version of that last move that is close to fraud and a version that is just optimistic, and the distance between them is smaller than people think.
The part that is actually about the reader
Here is the thing I noticed while writing this morning's post, and it is the reason I am writing this one.
That article had one job: give somebody the questions to ask a vendor so they are not talked into something. Halfway down, I wrote four questions about cost. What is the unit and what makes it move. What happens at ten times the volume. What does leaving cost. What is in rands and what is in dollars.
Every one of those is a question I would have to answer honestly about my own work, in a conversation, before I could give a number. So publishing a number would contradict the article inside the article. It would be a page telling you not to accept a headline figure from a vendor, with a headline figure from a vendor on it.
I did not plan that. I noticed it at the point where I would normally have written "and here is what we charge", and the sentence would not go in.
What I do instead, and where it costs me
Instead of my price, that post carries other people's numbers, traced to primary documents. Stanford's inference cost figure, and what it actually measured rather than what it gets quoted as. The transparency index score falling from 58 to 40 in a year. Those took most of the morning to verify and they are checkable by anyone who wants to.
That is the trade. Research instead of a rate card. It is more work, it ranks worse for the exact phrase people search, and it means some percentage of readers leave to find someone who will just tell them.
I want to be straight that this costs me. It is not a clever growth tactic disguised as principle. Somebody reading this on a Monday morning with a budget in their head wants a number, does not get one here, and goes elsewhere. That happens, I cannot measure how often, and I have decided it is worth it anyway.
Where the number does live
It is not a secret. It is in the proposal, in writing, with what it covers and what it does not, after a conversation where I have seen the actual work. Everyone I have ever quoted has a number. What they do not have is a number I invented before I knew anything about them.
That distinction is the whole rule. Not "no prices". No prices before the diagnosis, which is the same argument I make about software, which is that recommending the tool before understanding the process is how you end up with an expensive version of the old problem.
The related thing I got wrong recently
Two days ago I published a piece here about holding a statistic back for six days and then discovering my reason for doubting it was rubbish, even though the conclusion was right. I have been thinking about whether this is the same error wearing a different suit. Right call, bad reasoning, waiting to be found out.
I do not think it is, and here is the test I applied. In that case there was a fact of the matter that I could go and check, and when I checked, my reasoning failed. Here there is no equivalent document to go and read. There is a judgement about whether a number given without context helps or misleads a specific reader, and the honest answer is that it depends on the reader and I will not always get it right.
So I am holding this one more loosely than the rule makes it look. The script enforcing it is rigid on purpose, because the failure mode of a soft rule is that it erodes on the busy days. But I would change it if somebody showed me the version where publishing the number serves the reader better than the conversation does. Nobody has yet.
Frequently Asked Questions
Is this not just a way to avoid an awkward conversation?
The opposite. Not publishing means having the conversation every single time, including with people who then say no. Publishing a number is the version that avoids the awkward conversation by filtering people out before it happens.
How do people know if they can afford you?
They ask, and they get an honest answer early rather than at proposal stage. The scoping conversation exists partly to establish that quickly, and if the answer is no, saying so in the first call is better for both of us than a polished document nobody wanted.
Would you publish prices for a productised service?
Yes, without hesitation. A fixed scope with a fixed output has a fixed price and hiding it would be pure friction. The rule applies to work whose size is genuinely unknown until somebody looks at it.
What would change your mind?
Evidence that readers are worse served by the conversation than by the number. Right now I have the opposite intuition and no data either way, which I would rather state plainly than dress up as certainty.