Statistical Configurations
Trust Every Winner with Statistical Configurations
A variation can lead all week and still be noise. Elevate calls a winner only when four requirements are met.
Four Requirements Decide Every Winner
Miss one and Elevate keeps your test running until you have it.
Confidence: 90%
How sure Elevate is the difference is real.
Visitors: 500 per Variation
Enough people in each group to mean something.
Conversions: 20 per Variation
Enough orders that a couple cannot flip the result.
Days Running: 3 Minimum
Long enough to cover how people shop across a week.
Catch These Before You Ship
- Shipping a change that only looked better
- Calling a test on a few hundred lucky visitors
- Trusting a result two or three orders could flip
- Ending a test before your weekend shoppers turn up
Cap Outlier Orders Automatically
Average order $100. One shopper spends $2,000. Elevate counts it as $300.
Shopper Still Counts
They bought, and that conversion stays in your result.
Set Your Own Limit
A multiple of your average order, a fixed amount, or a percentile of your order values.
Drop Outliers Instead
Capping is the default. Removing them from results is one click.
Always On
At three times your average order. Turn it off if big orders are normal, such as wholesale.
Keep Free Gifts Out of Revenue
Bundle and gift apps add $0 lines to orders. Your unit counts look bigger than they are and your average order value looks worse.
Exclude Free Gifts
Removes the lines from revenue, average order value and profit. The sale still counts.
Off by Default
Turn it on if you run a gift-with-purchase or bundle app on your store.
Raise Confidence for Decisions That Stick
Confidence runs from 80% to 99%.
Raise It
Elevate needs more evidence and your tests run longer. Worth it before price changes or anything near checkout.
Lower It
Elevate calls sooner and more of your results shift later. Fine for a headline you can swap back in one click.
How Elevate Decides
Bayesian statistics, with a probability to win for every variation.
Change Them When It Matters
| Price Tests | Raise confidence to 95%. Ship the wrong price and you lose margin on every order until you catch it. |
|---|---|
| Low-Traffic Pages | Check your traffic can reach 500 per variation. Elevate keeps running until it does. |
| Wholesale Stores | Turn outlier capping off. Large orders are your business, not a freak event. |
| Gift Promos | Turn on Exclude Free Gifts first. Fixing the numbers afterwards is harder. |
Set Once, Applies to Every Test
Store-Wide
Every experiment follows them, including ones your agency sets up.
Scales With You
Elevate raises your defaults as you grow.
Reversible
Reset to Default puts everything back.
Common Questions About Statistical Configurations
Get Started with Statistical Configurations
Ship a winner knowing it will still be a winner next month. Elevate holds every test to the same bar, so you are never acting on a lucky week or one unusual order.
