What a demand test can and cannot tell you
A demand test puts a described product in front of people who have the problem it solves and measures how many are willing to do something that costs them: join a waitlist, pay a refundable deposit, place a pre-orderAn order and payment taken before the stock is ready, with terms that state what the buyer commits to, when dispatch happens and what if the date slips.. Its strength is that it measures behaviour, not opinion. People who say in an interview that they would buy something often do not; people who hand over a deposit have shown more than an opinion. This is the difference between stated and revealed preference, and the test is built to collect the revealed kind.
Its limits are just as important. It measures response to a description, not to the product, so a product that wins the test can still disappoint. The people who commit early are enthusiasts, and their numbers do not scale in a straight line to the mainstream market. And it answers whether demand is enough for a decision, not how big the market is or which message works best. Once a working version exists, the tutorial on running a beta or early-access programme tests the product itself with customers from the target segment before launch.
| Question | Does a demand test answer it? | Where it is answered |
|---|---|---|
| Will enough of the right people commit at this price to justify launching? | Yes, it is the question the test exists for | This tutorial |
| How large is the whole market? | No | The market sizing guide |
| Which claim or message earns most interest? | Only incidentally | The positioning statement for a launch, which tests claims before launch |
| How should the launch be planned and phased? | No | The go-to-market guide |
| Is a waitlist on track during a launch? | No; that is a launch KPI | The launch objective hub |
| Will customers be satisfied once they have it? | No | Early customers, reviews and retention after launch |
The neighbouring launch and market decisions are covered in the strategy section and the pages linked above; this tutorial keeps to the first question.
Before you start
- Know the decision the test serves. Ordering stock, building the product, hiring for it or entering a segment. The size of that decision sets how strong the evidence needs to be.
- Have the economics to hand. The margin per unit, any minimum order or build cost, the price you intend to charge and what a visit from the chosen traffic source is likely to cost. The decision ruleA rule written before a test starts that says which result ships the change, which leads to another round and which abandons it. is built from these.
- Have a first segment in mind. The go-to-market guide covers choosing the first segment for a launch; the test is run on that segment.
- Have tracking ready. Analytics on the test page, a key event for the commitment and a naming convention for tagged links.
- Agree who decides. Name the person who will apply the decision rule and make the call, and agree that the rule will not be rewritten after the results arrive.
The steps
Write the hypothesis and the decision rule
Write the HypothesisA testable statement of the change to make, the effect expected on a named metric and the evidence-based reason for expecting it. in one sentence that names who, what and at what price: people in a stated segment, who need to get a stated job done, will commit to this product at this price. The jobs-to-be-done framework helps phrase the middle part in terms of what the customer is trying to achieve rather than what the product does.
Then write the decision rule: the result that means go, the result that means change something and test again, and the result that means stop. Take every threshold from the business's own numbers. Work backwards from the decision: how many units must sell to justify the order or the build, what each commitment is likely to be worth once some deposits are refunded or some sign-ups never buy, and what the traffic costs. The threshold is the commitment rate at which the launch would pay, not a figure borrowed from someone else's test.
Fix the period and the number of visits in advance as well, and write down how ranges will be read: for example, go only when the rate clears the go line and the lower end of its plausible range clears the stop line. Date the document. This is the hypothesis card of the experimentation loop, and it is what keeps the test honest when the results are ambiguous.
Gather desk evidence on search and competitors
Before spending anything, look at what is already visible. Search VolumeThe estimated number of times a query is searched in a period, usually per month. for the problem, the category and its substitutes shows whether the need is being expressed and in what words; the market sizing guide explains why it is stronger as a relative signal than as a count of buyers. Look at who already sells something similar, at what price, and how they describe it. Read reviews of the alternatives for what buyers complain about, because an unmet need in those complaints is often the best evidence that a gap exists.
Write the findings down as observations with their sources. They do not decide anything on their own, but they shape the page, sharpen the segment and sometimes reveal that the test is unnecessary, because demand is plainly there or plainly absent. Where the question is a new country rather than a new product, the business's own orders, enquiries and search queries from that country are desk evidence too, and the tutorial on reading the demand you already get from other countries reads them before any test is designed.
Talk to people who have the problem
Interviews reach the reasons behind behaviour that a page cannot. The customer interviews tutorial explains how to plan a small study around a decision and keep each conversation on what people have actually done: the last time they faced the problem, what they used, what it cost and what annoyed them. Do not pitch the product; people are polite about ideas, and a pitch turns an interview into a test of manners.
Listen for the words people use to describe the problem, the alternatives they compare against and the objections they raise. These become the page's headline, its comparison and the questions it answers. What interviews cannot tell you is whether people will commit, which is what the page tests.
Build an honest page that asks for a commitment
The page describes what is coming, says plainly that it is not yet available and asks for one commitment. The landing page brief template gives the fields to fill before anything is designed: the single action and the key event that records it, the promise the traffic source makes, the audience and its objections, the headline and the offer stated plainly with its conditions, proof, the form and the tracking.
Commitment What it asks of the visitor What it shows What the page must state Waitlist place An email address Interest, at the lowest cost to the visitor What they will receive and when, and how the address will be used Refundable deposit A small payment A willingness to part with money The amount, when and how it is refunded, and what happens if the product is not made Pre-order The full price, or part of it A purchase decision The expected dispatch date, what happens if it slips, and the cancellation and refund terms Honesty is not optional. The page never implies that the product exists or can be bought now when it cannot. The approvals and compliance guide records the principle that marketing must not mislead and that claims must be capable of substantiation, and the trust and proof guide is clear that only real proof belongs on a Landing PageThe page a visitor arrives on from a campaign, built to turn that visit into one action.: no invented reviews for a product nobody has used. Prices and deadlines must be real too; the pricing and packaging guide records that in the UK the Competition and Markets Authority enforces consumer protection law covering misleading pricing practices, including urgency claims that are not true, and that the Advertising Standards Authority's codes cover price statements. Taking deposits or pre-orders brings obligations that differ by country; this is educational guidance, not legal advice, so have the page and its terms reviewed before it takes money.
Choose a small, tagged traffic source
Choose one source that reaches the segment in the hypothesis: search ads on the problem and category terms found in step two, an email to an existing list of people who agreed to hear about new products, a post in a community where the segment gathers, or a partner whose customers fit. Avoid sources that bring the wrong people, such as friends, followers of the founder or a broad audience, because they produce commitments that will not repeat.
Tag every link with UTMURL tags identifying traffic source. parameters following the business's naming convention, so the test's visits and commitments can be separated from everything else. Set the budget from the number of visits fixed in step one and the expected cost per visit. Use one message: the best claim from any message testing already done, not several competing versions, because this test is about demand, not wording.
Run for the planned period without changing the page
Launch the page and the traffic together and leave both alone until the planned period or the planned number of visits is reached. Changing the headline, the price or the offer during the test mixes several tests into one and leaves a result nobody can read. Stopping early because the first days look good, or bad, is the same mistake in another form, because early results from small numbers swing widely.
Fix anything that is broken, such as a failing form or a payment error, and log the fix with its date. Check daily that the tracking is recording visits and commitments; do not check daily whether the result is winning.
Count commitments, not clicks
Clicks, visits, scroll depth and time on page show interest, and interest is cheap. The test is decided on commitments: confirmed waitlist places, deposits paid, pre-orders placed. Count them from the business's own records, such as the payment system or the email platform, not only from analytics.
Clean the count before reading it. Remove commitments from staff, friends and family, duplicates, test transactions and anything from untagged sources. Note refunds requested during the test, because a deposit reclaimed within days is weaker evidence than one left in place. Record who committed where you can, because commitments from outside the target segment say less about the hypothesis than those from inside it.
Compare the result with the decision rule
Divide commitments by tagged visits for the commitment rate, and read it as a range rather than a single figure. The guide to reading small numbers explains why: a rate built on a few dozen commitments has a wide band of plausible true values around it, and the conversion rate calculator handles the arithmetic. The planned Sample SizeThe number of users or sessions an experiment needs to detect a given effect reliably. was chosen so that this range would be narrow enough to decide; if it is not, the rule written in step one says what to do.
Place the rate and its range against the go, change and stop lines. A result clearly above the go line, or clearly below the stop line, decides itself. A range that straddles a line is a result too: it says the evidence is not strong enough for that decision, and the rule should already say whether that means a second test or a smaller commitment.
Decide go, change or stop
Apply the rule and write the decision down with the evidence beside it: the hypothesis, the rule, the count of visits and commitments, the rate and its range, and anything unusual during the test.
- Go. Plan the launch with the go-to-market guide, using the commitments as the first customers and the evidence as the starting point for launch targets. For a product sold in an online store, the tutorial on launching a new product in an online store sequences the page, feed, pre-order terms and launch-day order that follow.
- Change. Alter one thing the evidence points to, such as the price, the size, the segment or the offer, and test again under a new rule.
- Stop. Record what was learned and why, so the idea is not revived next year without new evidence.
A partial go is a legitimate outcome: one version, one size or one segment passes and the rest does not. Remember the limit noted in the positioning statement guide for launches, that waitlists over-represent enthusiasts; the same caution applies to early deposits, so launch targets should sit below what the test rate alone would suggest. If the product launches and then falls short, the diagnosis of an underperforming launch works through the causes.
Tell the people who signed up
Everyone who committed is owed an answer, whatever the decision. If the answer is go, tell them the dates, what happens to their deposit or pre-order and when they will hear next. If the answer is change, tell them what is changing and whether their commitment still stands, and offer a refund if it no longer fits. If the answer is stop, refund every deposit promptly, thank them and, if they are willing, ask a few of them why they signed up; their answers are the start of the next idea.
Use the contact details only for what people signed up for. A waitlist for one product is not permission for general marketing unless the sign-up said so.
A rug maker tests a new line (illustrative)
Nimbleweave Rugs tests washable runners before ordering stock
Nimbleweave Rugs, an invented maker of woven rugs sold online, is considering washable runner rugs in two sizes: a short runner at £85 and a long runner at £115. Its weaving partner's minimum order is 300 rugs per size, and contribution is about £34 on the short runner and £46 on the long one. Interviews with 12 recent customers who own hallway runners turn up the same complaint: muddy boots and pets, and rugs too heavy to wash at home.
The owner writes the decision rule before the test. Her search ads cost about £0.60 a visit on her own account. At 3 deposits per 100 tagged visits, a deposit costs about £20 in ads; if two in three deposits become orders, an order costs about £30, inside the contribution of both sizes. So, for each size: go at 3 or more deposits per 100 visits, provided the lower end of the range is above 1.5; stop below 1.5; anything between means change and test again. The test runs for four weeks or 1,000 tagged visits, whichever comes later.
The page reads "Coming soon: washable runner rugs. Not yet made." It offers a £10 deposit per rug, refundable at any time before dispatch and refunded automatically if the rugs are not made, and states that the decision will be announced at the end of the test.
| Measure | Short runner | Long runner |
|---|---|---|
| Tagged visits | 1,040 | 1,040 |
| Deposits after removing friends and duplicates | 34 | 9 |
| Deposits per 100 visits | 3.3 | 0.9 |
| Plausible range | About 2.3 to 4.5 | About 0.5 to 1.6 |
| Against the rule | Go | Stop |
She orders 300 short runners only. Everyone who left a deposit hears within a week: short-runner depositors get a dispatch month, and long-runner depositors get a full refund with the option to move their deposit to the short runner, which 4 of them take. Launch targets are set below the test rate, on the assumption that early depositors were keener than the buyers who will follow.
Common mistakes
- Writing the decision rule after the results. Any result can then be read as a success.
- Thresholds borrowed from other tests. The rate that justifies a launch comes from your margin, your order size and your traffic cost.
- A page that implies the product exists. It misleads visitors and invalidates the commitments it collects.
- Counting clicks as demand. Interest is cheap; decide on commitments.
- Sending the wrong traffic. Friends, followers and broad audiences produce commitments that will not repeat.
- Changing the page during the test. Several tests mixed into one cannot be read.
- Stopping early on a good start. Early results from small numbers swing widely.
- Treating test commitments as the launch forecast. Early committers are enthusiasts; plan below the test rate.
- Leaving sign-ups without an answer. Everyone who committed is owed one, and refunds are owed promptly.
