Buying Learning Tech: Why A Vendor's Refusals Predict Quality Better Than Its Feature List

Buying Learning Tech: Why A Vendor's Refusals Predict Quality

Recently, we have noticed that Learning Tech demos often sound too good to be true. I sat through a demo last year where the answer to every question was yes. Could it handle our legacy SCORM files? yes. Could it push completion data into a system the rep had not heard of before that morning? Yes. Could it be white-labeled by next quarter, absolutely? I left the call feeling good about it, which in hindsight was the tell. Nobody who knows their product that well agrees to that much.

We did not buy, for unrelated reasons. But the pattern stayed with me because I have since noticed how reliably it runs the other way. The vendors I trust most in any category are the ones who interrupt a requirements list to say that one item is not something they do, and here is why, and here is what they would suggest instead. That sentence costs them the deal sometimes. They say it anyway, which is the point.

A Yes Is Cheap And A No Has To Be Paid For

Saying yes in a sales call costs nothing at the moment it is said. The bill arrives later, in implementation, and it arrives at the buyer's desk rather than the vendor's. So the incentive structure of a demo rewards agreement regardless of whether the agreement is true.

A no works differently. To decline something in a live sales conversation, someone at that company has to have decided in advance where the boundaries are, written it down, and given the person on the call permission to lose revenue by respecting it. That is three organizational facts you can infer from one sentence. None of them appear on a feature matrix.

This is why I have stopped treating a refusal as a gap in the product. It is closer to evidence about how the company is run, which is the thing you are actually buying when you sign a three-year contract for learning technology.

The Same Pattern In A Very Different Market

I run a digital PR and link building agency, and I test this kind of thing on our own data because I distrust arguments that only work as anecdotes. Earlier this month we published an analysis of Canadian link building prices covering 231 publishers, which sorted the market by what each publisher was willing to carry.

Just under half of them would publish content from at least one legally restricted category. Slightly over half would publish none of it. Then the interesting part: for ordinary, unrestricted content, the publishers that refused everything quoted a median of $1,110, against $610 for the ones that accepted something. Same deliverable, 1.82 times the price, and the pattern held in every category we checked.

The refusal was not a niche policy. It tracked with having written advertising terms, a disclosure practice, and somebody in the building whose job was to approve advertisers before publication. Publishers who disclosed sponsored content charged $1,420 against $405 for those who did not. The companies with governance said no more often, charged more, and reached larger audiences.

I did not expect the gap to be that wide when I ran it. It changed how I read a supplier's willingness to accommodate me, in my own industry and elsewhere. An organization that has never built the capacity to decline anything has usually not built much else either.

Meanwhile The Metric Everyone Shops On Was Flat

The same analysis turned up something I think L&D buyers will recognize. Publishers are shopped on an authority score, and across the middle of the Canadian market, that score had stopped carrying information. Sites scoring 30 to 39 and sites scoring 50 to 59 quoted almost the same price, around $550, and the band in the middle actually came in slightly cheaper than the one below it. Price only moved at the top, where audience moved with it.

Learning technology has its own version of this. Course library size, number of integrations, badge counts on a website: these are the numbers vendors compete on, and they have been optimized for so long that they no longer separate good from mediocre. Everyone has ten thousand courses now. A metric that every serious competitor has maxed out tells you nothing about which one to pick, and a feature checklist made of those metrics will return a tie.

Refusals have not been optimized, because no vendor has ever been rewarded for them. That is exactly why they still carry the signal.

Five Refusals Worth Hearing

"We will not be your system of record for that." A vendor who insists their platform should own every piece of learning data in your organization either does not understand your existing stack or is pricing in lock-in. Drawing a boundary around what they store is a sign they have thought about where they sit.

"Our AI features will not do that without a human review step." If you ask whether their AI can generate finished assessments and the answer is an unqualified yes, ask who validates the output. I wrote about this workflow for eLearning Industry earlier this year. The honest answer involves a review gate, and vendors who have built one will tell you about it before you ask.

"That integration exists but it is fragile." The cost of an integration is in maintaining it, not in its existence. A vendor willing to describe one of their own connectors as brittle has given you something more useful than a logo wall.

"We cannot commit to that timeline." Almost every implementation slips. A vendor who will not promise a date they cannot hold is telling you they intend to be honest during the hard part of the project, which is when honesty actually matters.

"Your requirement is unusual and we would rather not build it." Custom work sold reluctantly is better than custom work sold eagerly by a company that has no plan to maintain it after launch. Bespoke features become unmaintained features faster than anyone expects.

How To Get A Vendor To Refuse Something

Standard demos are designed to prevent this, so you have to make room for it. Ask directly what kind of customer their product is a poor fit for. A good answer is specific and slightly unflattering to their own product. A bad answer is that they work well for everyone.

Ask what they turned down in the last year and what happened. Also, ask which of their features their own team recommends against using. Ask for a customer who left and why. None of these questions are aggressive, and all of them are hard to answer smoothly if the honest answer does not exist.

Then note what happens when you ask for something impossible. I sometimes put one requirement into a brief that I know cannot be met cleanly, not as a trap but as a test of what the company does when it hits a limit. The response is informative either way.

When A No Is Just A No

None of this means every refusal is virtuous. Some are dressed up product gaps, and some are a vendor with no engineering capacity trying to sound principled. The difference is usually in what follows. A governance no comes with a reason and an alternative. An evasive no comes with a change of subject.

I sit on both sides of this. We decline client work at ESBO Ltd most weeks, usually where someone wants a brand mention placed somewhere the content does not belong, and those conversations cost us revenue in a way that is easy to measure and hard to enjoy. What I get for it is the ability to tell the next client what we will not do and be believed. Judge vendors on whether they have made that trade, because you can hear it in a demo if you leave space for it.