Back to the on-screen lesson ·
A useful test counts real customer behavior against a threshold written down in advance, often worked out from what the change costs, and a yes is the range of results from that threshold up to what the business can deliver.
Paper packet. Every task here also exists on screen, where it is checked automatically; answers written on paper are not assessed by Nydus. When you are back at a device, enter your answers there.
You will design a small test of real customer behavior that could change a stated decision, set the threshold in advance from the cost of the change, and read the result against it without moving the line.
You can find a segment, name its job and its alternatives, and write a promise backed by proof and bounded by honest limits. Whether enough customers want that offer is still a guess until something is tried. This lesson is about trying it in a way that settles the question instead of starting an argument.
| Term | What it means |
|---|---|
| Offer test | A small, real version of an offer, run to settle one decision. |
| Decision | The choice the test will settle, such as add breakfast or not. |
| Commitment signal | Customer behavior that costs them something: a payment, a pre-order, a booking. |
| Threshold | The result, written down before the test, that would mean yes. |
| Contribution | What one order adds toward costs: its price less the cost of making it. |
| Capacity ceiling | The most the business can deliver, the top of a yes. |
Most owners 'test' an idea by trying it and seeing how it feels. Four weeks later the result can be read either way, because nobody said in advance what would count. A test worth running has three parts, written down before it starts.
The decision it will settle: add a breakfast menu, or not. The behavior it will count: paid breakfast pre-orders, something that costs the customer something, not 'would you like breakfast?'. The threshold: the result that would mean yes.
The threshold is often not a matter of taste at all. Opening an hour earlier for a two-week trial costs Neighborhood Kitchen 150 dollars of staff time. If each breakfast contributes 5 dollars, the trial needs 150 ÷ 5 = 30 orders just to pay for itself. Below 30, the numbers say no, however cheerful the regulars were about the idea.
$$\text{orders that pay for the change} = \frac{\text{cost of the change}}{\text{contribution per order}}$$
Another way: table
A weak test and a strong one.
| Weak | Strong | |
|---|---|---|
| Behavior | 'Would you like breakfast?' | Paid pre-orders |
| Threshold | None | 30 orders in two weeks |
| Decision | 'See how it goes' | Add breakfast, or not |
| Length | Open-ended | Two weeks, then decide |
Another way: steps
When the test ends, compare the result with the threshold and act on it. The temptation, when a result falls just short, is to lower the line: 'twenty-six is nearly thirty, and people seemed keen'. Once the line moves after the result is known, the test has stopped testing anything. It has become a way of agreeing with yourself.
A result can also be too good. If Maya's starter set is pre-ordered 60 times and one kiln makes 40 sets a month, going ahead as planned would break 20 promises. So a yes is a range: from the threshold up to what the business can actually deliver. Above that range, the decision is a different one, such as a second kiln, a longer lead time or a waiting list, and it deserves its own numbers. Keep each test small and short, so that a no costs little, and keep the record, so the next test starts from evidence rather than memory.
Write the decision. One sentence that ends in a choice: add breakfast or not, keep the coffee corner or not.
Pick the behavior. It must cost the customer something: money, a deposit, a booked slot, a signed order. Opinions, likes and sign-ups that cost nothing are not counted.
Work out the threshold. Add up what the change costs over the test period and divide by the contribution of one order. Round up, because part of an order pays for nothing.
Note the ceiling. The most the business can deliver in the period, from its hours, ovens or kilns, as in the last lesson.
Run the smallest real version. A sign, a pre-order list, a single stop or a single weekend, for a fixed time set in advance.
Compare and decide. A result at or above the threshold and at or below the ceiling is a yes. Below the threshold is a no. Above the ceiling is a yes to demand and a new question about capacity.
Check the answer. Multiply the threshold back by the contribution per order: it must cover the cost of the change. Check that the threshold was written down with a date before the test started. And check that the behavior counted was paid for, not promised.
Dividing the cost of the change by the contribution per order is allowed because each order adds the same contribution toward that cost. When orders differ, use the average contribution from the business's own records, and say so beside the figure.
Counting only paid behavior is allowed, and needed, because the threshold is in paid orders. Comparing 30 paid orders needed with 45 people who said 'sounds great' compares two different things; only paid orders can be set against the cost.
Fixing the threshold before the test is what makes the comparison fair. A line drawn after the result is known will always be drawn where the owner hoped, so the comparison would tell nobody anything new.
Capping a yes at the ceiling is allowed because an order the business cannot deliver is not a sale; it is a promise that will be broken.
A test should be the cheapest real version of the offer, not a full launch. Before printing menus, buying a second oven or hiring a breakfast cook, Neighborhood Kitchen can take pre-orders on a sign by the register. Before building a website for the vegetable bag, Monica can use a clipboard. Before buying a second kiln, Maya can take deposits on a waiting list.
Small tests have two advantages. A no costs little, so the owner can afford to hear it. And because each test is cheap, the owner can run several in a season: a breakfast test in March, a delivery test in May, a price test in the fall. A business that runs four small tests a year learns more than one that bets everything on one big launch.
Small does not mean fake. Taking pre-orders for something the business cannot deliver is not a test; it is a broken promise waiting to happen. Every test offer must be one the business would honor if customers said yes, at the price and on the terms stated.
Set the length before the test starts, for the same reason as the threshold: a test with no end can be stopped on a good day and declared a success. The length should cover at least one full cycle of the business. A café test needs every weekday, because Mondays and Fridays differ. A farmers market test needs at least three Saturdays, so that one rainy morning does not decide it. A seasonal business must test in the season the offer is meant for: a snow-removal add-on tested in July says nothing.
Longer is not always better. A test that runs for six months ties up money and attention, and the world changes while it runs. Two to four weeks is enough for most small offers, as long as the threshold scales with the length: half the weeks, half the orders.
Write every test on one line of a simple log: the date, the decision, the behavior counted, the threshold, the result and what was decided. Add a short note on anything unusual, such as a storm, a holiday or a competitor's sale.
The log does three jobs. It stops the owner from quietly forgetting the threshold. It shows partners, lenders and staff why a decision was made. And over a year it becomes a record of what this business's customers actually buy, which is worth more than any survey.
A neighborhood bakery in Portland, Oregon, was asked again and again whether it delivered. The owner was tempted to sign up with a delivery app and buy an insulated van, but both would cost real money every month. Instead she wrote one line in her notebook: 'Decision: start Saturday delivery or not. Count: paid delivery orders, taken by phone, for four Saturdays. Yes if at least 24 in total.'
The threshold came from the costs. Hiring a driver with his own car for four Saturday mornings would cost 480 dollars. Each delivery order, at a typical 45 dollars of bread and pastries less ingredients and packaging, contributed about 20 dollars. 480 ÷ 20 = 24 orders just to cover the driver. She also noted her ceiling: the driver could manage about 12 drops a morning, so 48 orders over the four weeks.
She put a sign by the register and a line on the shop's social media page. The result was 19 paid orders. Several customers had said they would 'definitely order', but most of them did not. The owner did not lower the line to 19. She recorded the no, kept the driver's number, and noted that most of the 19 orders came from two apartment buildings nearby. The next test, run in the spring, offered pickup lockers in those buildings instead. It cost far less and passed its threshold in three weeks.
The approach in this lesson is close to what Eric Ries called a minimum viable product: the smallest version of an offer that lets a business learn from real customer behavior. The U.S. Small Business Administration's guidance on market research makes the same point in plainer words, urging owners to test demand with real customers before committing money to a launch. What both add to a guess is a number decided in advance.
A test succeeds when people say it sounds good. Only behavior that costs the customer something counts.
Decide the threshold once you see the result. Then any result can be read as a yes.
Bigger tests are better. A small real test that can say no cheaply is worth more than a big launch that cannot.
More demand is always good news. Beyond capacity it is a new decision.
A no means the test failed. A clear no is a success: it saved the cost of a launch that would not have paid.
Write the decision.
$\text{offer a two-hour express clean, or not}$
One decision, stated first.
Choose the behavior.
$\text{paid bookings in one month}$
Customers pay for the slot.
Work out the threshold.
$240 \div 16 = 15$
Setup cost over each clean's contribution.
Compare the result.
$19 \ge 15$
Read against the number written down.
Decide and record.
$\text{go ahead; note which segment booked}$
The record shapes the next test.
Write the decision.
$\text{add breakfast, or not}$
One choice to settle.
Find the trial's cost.
$10 \times 14 = 140$
Ten dollars of staff time a day for two weeks.
Work out the threshold.
$140 \div 5 = 28$
Five dollars of contribution per breakfast.
Turn it into a daily figure.
$28 \div 14 = 2$
Easy to count at the register.
Compare the result.
$23 < 28$
Twenty-three paid breakfasts in two weeks.
Decide without moving the line.
$\text{no breakfast menu for now}$
A cheap no, recorded for next year.
Write the decision.
$\text{add a café starter set, or not}$
One choice.
Choose the behavior.
$\text{pre-orders with a deposit}$
A deposit costs the café something.
Work out the threshold.
$600 \div 40 = 15$
Molds and photos over each set's contribution.
Note the ceiling.
$40 \text{ sets a month}$
What one kiln can make.
Write the range that means yes.
$15 \le x \le 40$
From the threshold to the ceiling.
Compare the result.
$52 > 40$
More demand than one kiln can meet.
Decide what to do.
$\text{take 40; waiting list for 12}$
A second kiln is a new decision with its own numbers.
Work out the threshold: opening costs 320 dollars a month; each job contributes 20.
$320 \div 20 = 16$
Cost over contribution.
Choose the behavior.
$\text{paid Saturday jobs over four Saturdays}$
Booked or walked in, but paid for.
Read a result of 21 jobs.
Neighborhood Kitchen is deciding whether to add a breakfast menu, which would mean opening an hour earlier. Which test could change that decision?
Complete the worked solution: opening early for a breakfast trial costs Neighborhood Kitchen $40$ dollars a day for $14$ days. Each breakfast contributes $5$ dollars. Find the trial's cost, the orders that would pay for it, and the orders needed each day.
Find what the whole trial costs.
$(\text{cost a day}) \times \text{fourteen} =$ f
Every day of the trial is paid for.
Find the orders that pay for it.
$(\text{trial cost}) \div (\text{contribution per order}) =$ n
Each order covers its contribution of the cost.
Find the orders needed each day.
$(\text{orders}) \div \text{fourteen} =$ d
A daily figure is easy to count against.
Say when the figure is written down.
$\text{before the first breakfast is served}$
A threshold set afterwards decides nothing.
Monica wants to test a pre-packed weekly vegetable bag. Put her steps in order.
Number the steps in order (write the number in the box):
Before testing a two-hour 'express clean', Bright Home Cleaning wrote down: 'go ahead if at least $15$ bookings are paid for in a month'. After a month, $31$ were paid for. What should it do?
Opening an hour earlier for a two-week breakfast trial would cost Neighborhood Kitchen $368$ dollars in extra staff time. Each breakfast order contributes $8$ dollars. How many breakfast orders over the two weeks would pay for the trial?
Answer:
A taco truck in Austin is testing a new lunch stop at an office park for five weekdays. The permit, parking and fuel for the trial cost $520$ dollars, and each lunch contributes $8$ dollars. How many lunches a day, on average, would pay for the trial?
Answer:
A bookshop owner wrote four test plans for ideas she is unsure about. Mark every plan that could actually decide something.
This task has no paper form; do it on a device.
Lesson test: one question per skill, one attempt each, no hints. Your answers are checked when you submit.
Maya tests a café starter set. She will go ahead if at least $19$ sets are pre-ordered in the test month, and one kiln can produce at most $31$ sets a month. For which numbers of pre-orders $x$ should she go ahead and deliver them all that month?
This task has no paper form; do it on a device.
You can turn 'let's try it and see' into a test with a decision, a behavior and a number. Tell someone why lowering the target after seeing the result spoils the test. Next: choosing the channel where customers already decide.
17. Your turn: Northside Repairs tests a Saturday opening, step 3
$21 \ge 16$
Above the line, so open on Saturdays.