How many postcards does a test need?
Enter the response rate you expect without a card and the lift you hope the card adds. The calculator returns how many cards to mail and how many customers to hold out so a real lift shows up in your results. It uses the standard sample size formula for comparing two proportions and runs in your browser.
Last updated: 22 October 2026
Runs in your browser. Nothing you type leaves this page.
How the calculator works
A postcard test compares two random groups from the same segment: customers you mail and customers you hold out. The response rate is the share of each group that orders within the same window, such as a 60-day attribution window. Chance alone moves both rates, so a small group can show a lift that is not there or hide one that is. The calculator finds group sizes where that risk stays inside the limits you set.
With equal groups, each group needs:
Here p₂ is the response rate without a card, p₁ is that rate raised by the lift, z₁ comes from the confidence level and z₂ from the power. When you hold out a share h of the audience, the holdout is k = h ÷ (1 − h) times the size of the mailed group, and the mailed group needs:
The calculator rounds each group up to a whole customer. It uses these normal values:
| Setting | z |
|---|---|
| 90% confidence (two-sided) | 1.64 |
| 95% confidence (two-sided) | 1.96 |
| 99% confidence (two-sided) | 2.58 |
| 80% power | 0.84 |
| 90% power | 1.28 |
The equal-groups formula and the example below follow Select Statistical Consultants' two-proportion calculator (checked 22 October 2026).
A worked example
Say 15% of the customers in a segment order without a card, and you hope a card lifts that to 25%. At 95% confidence and 80% power, with equal groups, you need 248 customers in each group: 248 cards to mail and 248 customers to hold out.
Lower response rates need bigger groups. This table shows the total number of customers in a 50/50 test at 95% confidence and 80% power:
| Rate without a card | +10% lift | +25% lift | +50% lift |
|---|---|---|---|
| 1% | 326,184 | 55,870 | 15,494 |
| 2% | 161,358 | 27,614 | 7,646 |
| 5% | 62,462 | 10,660 | 2,936 |
The split matters too. At 2% without a card and a 25% lift, a 50/50 test needs 27,614 customers in total. Holding out 10% instead needs 70,039: 63,035 cards and 7,004 held out.
Reading your result
- Use the response rate of the segment you will mail, taken from your own store.
- Decide the window and the group sizes before the first card goes out, then read the result once both groups reach their size and the window has closed.
- If the numbers are larger than your segment, test for a bigger lift with a stronger offer, or let the campaign run until both groups fill.
- A lift you can detect still has to pay for the cards.
Running the test in TouchDrop
To run the test with TouchDrop, split your segment at random before you send: mail one group and leave the other out. Count orders for both groups the same way, from your store's orders matched by customer over the same window. A card's discount code only exists in the mailed group, so counting code redemptions on one side and store orders on the other would bias the comparison.
Frequently asked questions
What response rate should I enter?
Your own: the share of customers in the segment you plan to mail who place an order within your attribution window without receiving a card. Read it from your store's orders for a similar segment over a past period of the same length.
Why does a smaller holdout need a bigger test?
The comparison is only as sharp as its smaller group. Holding out 10% instead of half leaves a small group whose response rate moves around a lot by chance, so you need more people overall to see the same lift.
What if my list is smaller than the result?
Test for a bigger lift, with a stronger offer or a segment the card should move more, or keep the campaign running until both groups reach the size. Stopping early and reading the result anyway gives you a number you cannot trust.
Does the calculator use TouchDrop's customer data?
No. It runs the standard formula on the numbers you type, inside your browser, and sends nothing to TouchDrop.