The Checkout Trials
Sentinel
Your checkout isn't broken.
It's just broken for someone you've never met.
Built for the PayPal AI Hackathon · Braintree sandbox · Gemini
The problem
Shops are built on fast laptops by people with steady hands and good eyes.
Low vision
A shopper with low vision can't read pale text.
Slow signal
A shopper on one bar of signal gives up on a heavy page.
No mouse
A shopper without a mouse can't reach a fake button.
They leave quietly. The shop owner only ever sees a lower sales number, never why.
The idea
Don't read a report. Watch them try to buy from you.
Sentinel sends seven pixel-art shoppers through a real checkout. Each one stands for a real kind of customer, checks the page the way that customer would, and reacts when something gets in the way.
Every problem costs coins from their purse. Lose too much patience, lose the sale.
Meet the shoppers
Seven kinds of customer. 29 checks between them.
Pick who goes. Each one checks the page the way that customer would.
A trial
They walk the real page.
- Every number is measured by the browser on the live page.
- The shopper walks to each problem, reacts, and loses coins.
- The checkout is a real Braintree sandbox payment.
1.6:1 contrast on the checkout button, where 4.5:1 is needed
Different shoppers, different problems
Each one fails your checkout in a different way.
What it costs
Plain words first. Detail when you want it.
- Who bought, who left, and where.
- The worst problems, in plain words.
- An estimate of money at risk each month.
- An AI verdict from Gemini, written from the measurements.
$735 a month at risk from Grandma Iris alone, in the run pictured (an estimate, with its assumptions shown)
Fix it and run again
Every problem comes with a fix. Try it, and send them again.
Grandma Iris
9 problems, sale lost → 0 problems, sale made
Pell
about 15 seconds to load → about 4
Cog
can't read the product → buys by itself
How it works
PayPal and AI do real work.
Braintree sandbox
Real orders with line items, prices set on the server, declines, double-submit protection, and verified webhooks.
Gemini
Writes each verdict from the measurements. Cog's agent decides whether to buy. Suggested colours must pass the contrast maths.
Safe by design
Runs only on our demo store or sites whose owner added one line. Sandbox only. On owners' sites, shoppers only look.
- Pick shoppers
- They walk the page
- Measured checks
- Braintree sandbox sale
- Gemini verdict
- Try the fixes
- Run again
Who it's for
A small shop owner
Send all seven before the first customer arrives.
An agency
Paste one line into a client's site and hand over a clear list of fixes.
A developer
Try fixes in seconds and re-run before every release.
Where it's going
Today, checkouts. Next, whole sites and apps, for more kinds of people.
Whole journeys
Sign-up, search, booking and account pages, on websites and in apps.
More kinds of people
Colour blindness, dyslexia, memory and attention, older devices, bright sunlight.
Research data at scale
Simulated access needs give repeatable test data for accessibility research.
Safety as well as access
Security checks the owner switches on, which find common weaknesses and suggest fixes.
One safe connection
Link a site or app in a click. It still runs only where the owner puts it.
Simply priced
A clear standard of accessibility and safety, for any shop.
It doesn't replace testing with disabled people. It catches the measurable problems early, so those sessions go further.
Try it
Sentinel
sentinel-ten-ashy.vercel.app
github.com/FaatehHaneef/Sentinel · MIT licence