Seehowwethink
A teardown is a close look at a product: how fast it is, where customers get stuck, and what we would change. Our first teardowns of real public products are coming. Until then, here are three worked examples on fictional products, so you can see exactly how we work.
Three steps, every time.
Observe
We use the product the way a customer would, on a mid-range phone and a laptop, and note every point of friction.
Measure
We time the key moments, such as how long until the page is usable, and repeat each test several times.
Recommend
We rank the fixes by customer impact and effort, and say what we expect to change.
Three products, three problems.
3 examplesExample · fictional product SpeedHarbor & Pine, a fashion shop
Product pages feel slow on phones.
Harbor & Pine, a fashion shop
Product pages feel slow on phones.
What we tested
A product page on a mid-range phone over a typical mobile connection.
What we found
- The first photo is far larger than the screen needs, and loads before the text.
- Seven outside scripts run before the Buy button responds.
- Size and colour options load last, so customers tap before they are ready.
What we would change
- Resize and compress photos, and load the first one first.
- Delay scripts that are not needed to buy.
- Load the options with the page, not after it.
Expected effect: Customers see the product sooner and can buy without waiting. We would compare completed purchases before and after.
Example · fictional product CheckoutQuillfare, a booking app
Customers abandon at the last step.
Quillfare, a booking app
Customers abandon at the last step.
What we tested
The full journey from choosing a time to paying, on a phone.
What we found
- The price changes at the end, because a service fee only appears at payment.
- Customers must create an account before they see the total.
- If payment fails, the booking details are cleared.
What we would change
- Show the full price on the first screen.
- Let people book first and create an account after.
- Keep the booking when payment fails, and explain what happened.
Expected effect: Fewer people give up at payment. We would measure the share who finish checkout.
Example · fictional product SearchFernway, a marketplace
Searches often come back empty.
Fernway, a marketplace
Searches often come back empty.
What we tested
Twenty everyday searches, including synonyms and common misspellings.
What we found
- Searching ‘sofa’ works, but ‘couch’ returns nothing.
- Misspellings return an empty page with no suggestion.
- Filters reset when the customer goes back to the results.
What we would change
- Treat common word pairs as the same thing.
- Suggest the closest match for misspellings.
- Remember filters when customers return.
Expected effect: Fewer empty searches, and more searches that end in a purchase.
How we keep teardowns fair.
Only what any customer can see
We never test private areas, use anyone’s data or break anything. If we can’t see it as a customer, we don’t write about it.
Company first, when possible
Where we can, we share findings with the company before we publish, and we only name things we can show.