Industry
Offer testing: planning, execution and measurement
Offer testing from a five-slot sentence: change one slot at a time, climb the commitment ladder, and tell an execution failure from a demand failure.
Most tests that founders call an offer test are really a poster test. A page goes up, some people click it or do not, and the result gets read as a verdict on the idea. It is not. It is a verdict on a headline, shown to whoever happened to see it, with no way to tell which part failed.
An offer is a sentence with five slots. Testing it means finding out which slot is carrying the risk, then putting that slot in front of someone who can say no.
What to take away
- Each slot is a separate claim that can be wrong on its own.
- The temptation is to rewrite the headline, drop the price, and switch the audience all in the same push, because it feels efficient and there is never enough traffic.
- The value of a test comes from what the person gives up to pass it.
- You can test an offer for something you have not built.
The five slots
For [who], we [deliver what outcome], replacing [what they do now], for [what it costs them], and here is why you should believe it: [the proof].
Each slot is a separate claim that can be wrong on its own.
Who. The most common silent failure. A test that flops on the wrong audience returns an accurate number about people who were never your customer. Describing that group so a stranger could find twenty of them is the first gate in market validation.
Outcome. Not what the thing is, what it changes. People buy the after-state. If your outcome slot describes a mechanism instead of a result, you have not written an offer yet.
Replaces. Every purchase displaces something, even if that something is doing nothing. Naming it forces you to be honest about switching cost and about who has to be persuaded, which is the whole subject of competitor validation.
Cost. Money is only part of it. Setup time, learning, disruption, the risk of being the person who chose this. A low price attached to a high-effort change is not a cheap offer.
Proof. Why should someone believe the outcome will happen for them specifically? Absent proof is a legitimate reason for a good offer to fail, and it is fixable without changing anything about the product. It is also the slot where honesty is easiest to lose: an advertiser needs a reasonable basis for a claim before the claim runs, not a defense assembled after somebody challenges it.
Write the sentence out fully before you test anything. Then underline the slot you are least sure of. That is what you are testing this week. The others hold still.
Change one slot at a time
The temptation is to rewrite the headline, drop the price, and switch the audience all in the same push, because it feels efficient and there is never enough traffic. What you get back is a single number that could mean any of six things.
If you only have the budget or the attention for one clean test, spend it on the slot that would kill the idea if it were wrong. Usually that is who or cost. Almost never proof, which is the one people instinctively work on first because it is the most comfortable to improve.
The commitment ladder
The value of a test comes from what the person gives up to pass it. Rank your test by what it costs them, not by how many people you can put through it.
| What you ask for | What it costs them | What a yes proves |
|---|---|---|
| A click | A second of curiosity | The headline was interesting |
| An email address | Mild inbox risk | The promise was worth a small ongoing cost |
| A survey answer | Politeness | Very little on its own |
| A scheduled call | Real time, in advance, with a diary conflict | The problem is currently on their mind |
| A document or data from them | Effort, and a bit of trust | They want the outcome enough to work for it |
| A refundable deposit | Money, reversibly | They are prepared to be the one who chose this |
| A payment | Money, and internal justification | Someone will pay this price for this promise |
| A signed order or contract | Money and reputation inside their organization | The buyer, not just the enthusiast, agrees |
The top rows are cheap to collect and correspondingly weak. The bottom rows are the ones that change your mind. A test that only produces the top three rows can tell you your message is working, which is genuinely useful, and cannot tell you that anyone will buy.
Where you sit on this ladder should climb over time. Early, a scheduled call from a stranger is a good result. Later, anything that is not money is a stall.
Design a test that can fail
Before it runs, write down three things:
- The exact behavior you are counting.
- The threshold that means continue.
- The threshold that means stop or change the offer.
Set both numbers in advance and set them in ink. The failure mode here is not dishonesty, it is that human judgment quietly reinterprets an ambiguous result in the direction of continuing, and an ambiguous result is the most likely outcome of any small test.
If you cannot name a result that would make you abandon the current offer, you are not running a test. You are running a campaign, which is a fine thing to do, but it will not tell you anything.
A useful discipline: write the prediction down as a sentence, sealed before the test, and read it afterwards. "I expect at least three of the twenty people I speak to this week to book a paid pilot." You will be badly wrong often, and the surprise is the whole point.
Telling execution failure from demand failure
A bad test can kill a good offer. This is the expensive error, because it looks exactly like a decisive result.
Some ways to separate the two:
Run the same offer through a warm channel. If a direct conversation converts and the page does not, the offer is fine and the page or the traffic is not.
Check the traffic before you check the result. Look at who actually arrived. If the audience does not match the who slot, throw the result away rather than interpreting it.
Ask the people who declined. Not by survey. A short message to a handful of them. "You looked and did not book. What put you off?" The answers usually name a specific slot, and often it is the one you did not think you were testing. The rules for hearing that without leading them are the ones in customer interviews.
Watch the shape of the drop-off. People leaving before they see the price is a message problem. People leaving at the price is a pricing problem or a proof problem. Those need opposite fixes.
Limits you should not cross
You can test an offer for something you have not built. You cannot take money for something you have no intention of delivering, and you should not imply that something exists when it does not.
The honest version costs you almost nothing. Say what state it is in. Take a deposit for a place in the first group rather than a payment for immediate delivery. Give a date, and give the money back without argument if you miss it. That last part is not only good manners: where money is taken for goods before they ship, the order and shipment rule requires notice, consent, or a prompt refund when the stated window cannot be met. Tell people who ask that delivery is manual right now.
None of this weakens the signal. A person who pays a deposit for something that does not exist yet, knowing that, has given you a stronger reading than one who bought what they thought was a finished product.
After the test
The number is the smaller half of the result. The larger half is in the conversations around it: which word people repeated back to you, what they assumed it did, who they said they would have to check with, what they compared it to unprompted. That comparison in particular is worth recording, because it tells you which category the buyer filed you in, and the category sets the price expectation before you ever name one.
Then go back to the sentence, change one slot, and write down the new prediction. Testing the price properly is its own job, and it needs a number someone can refuse. Deciding how small the delivered thing can honestly be is another. Whether the problem underneath is real at all is upstream of both, in problem discovery.
Common questions
How many people does an offer test need?
Enough that one person changing their mind would not flip the conclusion. Run that check explicitly: move one event and see whether you would decide differently. If you would, you have a hypothesis rather than a result.
Can I test an offer for something that does not exist?
Yes, provided the buyer knows what state it is in, the date is one you can meet, and the money comes back if you miss it. What you may not do is imply that a finished thing exists.
What if almost nobody saw the offer?
Then the test measured the channel and the offer question is still open. Say so, and resist the reading that a small count is a rejection.
Which slot should I test first?
The one that would end the idea if it were wrong, which is usually who or cost. Proof is the one people work on first because it is the most comfortable to improve.
Is a survey ever an acceptable offer test?
No. A survey answer costs politeness. An offer test needs something the person gives up, and the whole value of the exercise is in what that cost was.