Bandit sorts your visitors into groups you can read, logs every request it counts, and sends more traffic to the version that converts.
Every feature on every plan. 14-day trial starts with your first event. No card needed to start; no charge until day 15.
Still shown a little, in case it catches up
Every request and reward, as it arrives
Converts 6% on “Gift-ready in 2 days” against 4% for everyone else.99/1,630 vs 161/4,020
Meta and Google build a model of exactly who converts, and they keep it. You get a report of totals, without the reasons.
A gift buyer, a returning customer, and someone from an email all get one headline.
Half your traffic sees the weaker version until the test ends. Then it often reports no clear winner.
Result: no significant difference
Who
Ask visitors one question and send the answer with each request. Every day, Bandit sorts them into groups, each one a plain rule you can read.
Buying for yourself or as a gift?
Bandit names the group
buying_for: a giftYour site also sends device, traffic source, and new or returning, so a rule can combine them. A group only appears when it converts differently on a version you're testing. You never scroll past segments that behave the same.
Each group shows its counts per version, the survey answers that describe it, and where its visitors are, by country and US state.
occasion: Mostly a birthday (62% of 73 answers)
Top states
What
Every funnel and campaign has a Data tab. It lists each version Bandit served and each reward that came back, with counts and when each was last seen.
Real rows. Each row is a request or a reward that happened, not a sample.
Credit that follows the sale. Funnels credit the final sale to every step that led to it.
Leavers still count. A visitor who leaves without buying counts as a $0, so the numbers include them.
| Time | Event | Step | Treatment | Detail |
|---|---|---|---|---|
| 14:11:52 | Treatment requested | Hero headline | 20% off your first order | v_2c91 |
| 14:09:05 | order_paid | Checkout button | Complete my order | 64 |
| 14:03:40 | Treatment requested | Checkout button | Complete my order | v_8f3aconverted |
| 14:02:11 | Treatment requested | Hero headline | Gift-ready in 2 days | v_8f3aconverted |
How
Test a headline, an offer, or a whole page, with headline, image, button copy, and button color together. Bandit sends each visitor a version and learns from what they do.
As soon as results come in, more visitors see the version that converts. The others still get a little, in case they catch up.
With forgetting on, older results count for less. December's winner doesn't stay locked in for March.
7-day half-life
14-day window
On Shopify, each order is scored by revenue minus unit cost, minus refunds. Orders with an unknown cost wait instead of guessing.
| Version | Revenue | Unit cost | Margin |
|---|---|---|---|
| Full priceWins | $48.00 | $31.20 | $16.80 |
| 25% off | $36.00 | $31.20 | $4.80 |
| 40% offLoss | $28.80 | $31.20 | -$2.40 |
Ask for new versions. Bandit writes them from your page and your live results. Every plan includes AI drafts each month.
Write a version for gift buyers that adds a deadline.
Gift-ready in 2 days. Order by Thursday.
A developer adds the TypeScript SDK or calls the REST API once. After that, you add versions and read results in the dashboard. The bandit CLI scripts the same steps.
Send this to your developerInstall the Bandit app, then add a Bandit Text block in the theme editor for the hero headline, subhead, main button, banner, or badge. No code. Orders come back with their margin, and billing runs through Shopify.
Bandit for ShopifyKeep the email service you already send with. Your send code asks Bandit for each recipient's copy, and every link carries a signed tag. When someone clicks, your site reports it and the next campaign starts from what this one learned. Bandit never holds your list or sends email.
Bandit for emailBandit does not replace your analytics, send your email, or touch your ad accounts.
On Shopify, no. Install the app and place Bandit blocks in your theme. On any other site, a developer adds the SDK once. After that, you run tests from the dashboard.
Bandit starts moving traffic as soon as conversions come in. Audiences take longer. A group needs at least 100 visitors in 30 days and has to convert differently from everyone else on a version you're testing, so expect no groups until the bandit has data.
It answers the same question. An A/B test splits traffic evenly until the test ends. Bandit shifts traffic toward the winner while it learns, so fewer visitors see the weaker version.
No. Keep them. Analytics and Klaviyo tell you who your customers are. Bandit shows what each group did on your page, in rules you can read, and changes what they see next. It doesn't send email or touch your ad accounts.
Yes. One login gives you a separate workspace for each client.
Growth is $99 a month and Scale is $299 a month, with unlimited team members and every feature on every plan. The 14-day trial starts with your first event, with no card needed to start and no charge until day 15. See pricing
Growth is $99 a month and Scale is $299 a month, with unlimited team members and every feature on every plan. The 14-day trial starts with your first event, with no card needed to start and no charge until day 15.