Pricing validation: a source-based guide for 2027

You cannot validate a price without a refusal.

Every other kind of price research is people being helpful. Ask what someone would pay and they will produce a number, sincerely, from nowhere. It costs them nothing to be generous with a hypothetical budget, and nothing to be cautious with one either, which is why the answers scatter and why averaging them produces a figure with no relationship to anything.

Two data points have weight. Someone paid X. Someone was asked for X and said no. Collecting the second is the harder skill and the one that gets skipped.

What to take away

• You want a real number said out loud in a real conversation, and then silence.
• "Too expensive" and "not worth it" sound the same and mean opposite things.
• You cannot interpret a refusal without knowing what you need.
• Once something is selling, the cleanest test is to raise the price for new customers only and leave existing ones alone.

Why the direct question fails

Three separate mechanisms, all working at once.

No consequence. The person answering does not have to live with the answer. Nothing is deducted from anything.

Anchoring. Your question contains a number, or a range, or a framing that implies one. Whatever you put in front of them shapes what comes back. Ask about a subscription and you get subscription-shaped numbers, even from someone who would have happily paid ten times as much once.

Wrong person. In anything sold to an organization, the person who feels the problem and the person who releases the money are usually different, and the first one is guessing about the second one's tolerance. Those roles are separated properly in problem discovery.

None of this is fixable by asking better. It is fixable by asking for money.

Getting a refusal cheaply

You want a real number said out loud in a real conversation, and then silence.

State it plainly. Not "it would be somewhere around", not a range, not "for you we could probably". One number, in the way you would say it to a customer, and then stop talking. The pause after a price is where the information is. Fill it and you have destroyed the reading.

Then ask for the next step that costs something: a deposit, a purchase order, a start date. What you are looking for is not agreement. It is the specific shape of the resistance.

You will not enjoy this. Almost everyone finds a way to avoid it, usually by testing everything else first and leaving price for later. Price is not a detail to settle at the end. For most ideas it is the largest single unknown, and the rest of the plan is downstream of it.

Two very different noes

"Too expensive" and "not worth it" sound the same and mean opposite things.

Not worth it is a value problem. They understand what it does and do not want it that much. More proof, a better outcome, or a different customer. Lowering the price will not fix this. It will just get you a cheaper version of the same no.

Too expensive means they want it and cannot justify the amount. Now find out against what. Three diagnostics:

• Would they pay that number for a competing product? Then it is positioning, not price, and what the buyer is comparing you against is the subject of competitor validation.
• Do they pay comparable amounts for other things that solve problems of this size? Then it is proof, or timing.
• Is there a budget line this could come out of at all? If not, you have a category problem, and no price works until the buyer knows which pocket the money comes from.

That last one is the quiet killer. A thing that does not fit an existing budget category has to be paid for by someone deciding to create a new one, which is a much larger ask than the number suggests.

Know your floor before you ask

You cannot interpret a refusal without knowing what you need. Build the floor from your own numbers, not from anyone's benchmark:

• What it costs you to deliver one, in materials, licenses, and hours at a rate you would actually accept.
• The support and rework it generates after delivery, which is nearly always underestimated in the first year.
• Refunds, failed payments, and the ones who leave early.
• What it costs to acquire the customer through the channel you can actually repeat.

Keep the one-time items in a separate block from the recurring ones while you build it, for the reason the SBA gives when it sets out how to calculate startup costs: only the recurring block bears on whether the price works month after month.

If the floor sits above what people will pay, the answer is not a pricing tweak. Either the delivery model has to change or the customer does.

Structure is a separate question from level

Two decisions get collapsed into one and they behave differently. The level is how much. The structure is what the meter runs on.

Per seat, per use, per unit of the customer's result, a flat fee, a fee plus usage, an annual commitment. Each one changes who says yes, how the customer behaves once they are in, and what happens as they grow. A per-seat price makes customers ration access. A usage price makes them nervous about a bill they cannot predict. A flat fee is easy to approve and hard to grow.

Test structure by asking how the buyer wants to be billed, which is one of the few forward-looking questions worth asking, because it is about their process rather than their preferences. Asking it without leading them is an interview skill, set out in customer interviews. The answer is often shaped by their own accounting rather than by your product, and that is exactly the constraint you need to know.

Things that ruin the reading

Discounting during validation. Every discount you give makes the data point unreadable. You now know someone will pay the discounted number, which is not the number you wanted to test.

Manufacturing pressure to close the sale. A deadline that does not exist, a queue with nobody in it, a struck-through price nobody has paid. Each is a statement likely to mislead a reasonable person about something that matters to the decision, which is the test the FTC applies in its advertising guide for small business. It also destroys the reading, because a yes given under invented pressure says nothing about whether the number holds.

Testing the price on people who love you. Your network will pay to be supportive. That is a personal transaction, not a market one.

Reading one yes as the answer. A single customer at a high price may be an outlier who had an unusual reason. Two independent yeses at the same number are worth far more than one enthusiastic yes and a lot of interest.

A price so low it selects the wrong customers. Cheap attracts a different population: more price-sensitive, more support-hungry, more likely to leave. Then you conclude the segment is difficult, when in fact you recruited it.

Never testing upward. If nobody hesitates, the price is too low and you are getting no information at all. A pipeline where everyone says yes immediately is a pricing test that has stopped running.

Testing an increase

Once something is selling, the cleanest test is to raise the price for new customers only and leave existing ones alone.

It gives you a comparison over a period you control, it does not punish the people who took a chance on you early, and it avoids the awkward position of defending a change to people who are already paying. Keep the old price for existing customers only as long as you actually mean to. An indefinite promise you later break costs more than the revenue it protected.

Watch two things, not one: whether the conversion rate drops, and whether the kind of customer changes. A higher price often converts fewer people who then stay longer and ask for less, which is a better business and looks worse on the first chart you check.

What good pricing evidence looks like

You have something usable when you can say: this specific type of buyer, reached this way, agreed to this number, in this structure, more than once, and this other type of buyer said no for this stated reason.

That is a paragraph, not a number. It carries its own limits, which is what makes it worth having. Anything shorter is either a guess wearing precision, or a survey result about a budget nobody had to spend. Getting the refusals that fill the paragraph in is the job of offer testing.

Common questions

How many refusals do I need before the price means anything?

Enough that the refusals sort into groups. A handful that all say different things usually means you are talking to several segments, and that is itself the finding.

Can I validate a price without taking money?

Only if the commitment costs the person something real and you would honor it: a deposit, a signed order, a scheduled start. A verbal yes with nothing attached is an opinion about a future.

What do I say when somebody asks for a discount during a test?

No, plainly, for this period. Then ask the next-step question again. If the conversation ends there, record it as too expensive with the comparison they named. You learned something; a discount would have taught you nothing.

My price is public and I think it is wrong. Now what?

Change it, and honor the old number for anyone who already saw it. Test the increase on new customers only. The version where existing customers discover the change from a bank statement costs more than the revenue it raises.

Is a competitor's published price a useful reference?

As a signal about what buyers expect, yes. As a target, no. A list price is an opening position and tells you nothing about what their customers actually pay or what it costs them to deliver.