Lesson 2 of 5
Test, don't guess: simple experiments for small traffic
Simple tests that work even with little traffic
The idea
A pizza place wonders whether a lunch deal would bring more people in. It could argue about it for months. Or it could run the deal for four weeks, count the lunches, and compare with the four weeks before. That's a test.
Step 1
Guess
“A film on the first screen brings more enquiries”
Step 2
Change one thing
Only that, nothing else
Step 3
Wait
A set time, decided before
Step 4
Decide
Keep it, drop it, or test again
Big websites can test
- Button colours
- Single words
- Tiny differences
Small businesses should test
- A new offer or price
- A different first sentence
- A different first step
Three cheap ways to test
Before and after
Change it, and compare the same number of weeks before and after
Ask five people
Show two versions, ask which they'd click, and why
A small ad split
Two versions, the same budget, one week
Today, with AI
Making two versions of anything now takes minutes. What's rare is the patience to test one change at a time.
Analytics tools flag changes by themselves, so it's easier to see whether a test moved anything.
What AI does well
Writing the versions, keeping the test card, and comparing the weeks when the wait is over.
Where it goes wrong
It will call a winner too early. Hold it to the wait you set.
What stays yours
The guess worth testing, the patience, and the decision.
Check yourself
Do it in the studio
Try it here first
Your test card
Fill it in before you change anything. Keep it where you'll see it when the wait is over.
Yours
We believe a film on the first screen will show in more enquiries a week. We'll wait four weeks, and keep it if enquiries rise by a quarter.
- 01
Fill in the test card above for one change you're thinking about.
free - 02
Make the change, and note its date in In your own words → What is coming.
free - 03
When the wait is over, compare the weeks before and after on Results → Traffic.
free - 04
Optional: Read the numbers has the CMO read the result.
20 credits
Done when One change is being tested, with what counts and the wait decided before it started.
Go deeper
University courses on the same topic, if you want to go further.
University of Oxford, Saïd Business School
Oxford Digital Marketing, Module 7: Experimentation, measurement and analyticsRunning tests and measuring what they show
Stanford GSB
Marketing Strategy (MKTG 248)Making marketing decisions with experiments