> ## Documentation Index > Fetch the complete documentation index at: https://docs.tented.ai/llms.txt > Use this file to discover all available pages before exploring further. # A/B Testing > Test up to five variations of a blast, then send the winner automatically ## What's an A/B test? An **A/B test** takes the guesswork out of a blast: instead of betting the whole audience on one subject line (or send time, or sender), Tented sends **up to five variations** to a small sample, measures which one performs best, and delivers the winning version to everyone else — automatically, or on your say-so. Every A/B test lives inside an [email blast](/email/email-blasts). Open a draft blast, head to the **Schedule** step, and find **AB Test** under **Additional Settings**. Click **Add Test**. A screenshot of the blast Schedule step showing the Additional Settings card with the Set as operational toggle and the AB Test section with an Add Test button.

A screenshot of the blast Schedule step showing the Additional Settings card with the Set as operational toggle and the AB Test section with an Add Test button.

Once a test is active, the blast's schedule follows the **test's** start time, and **Send now** is unavailable — a test needs time to collect results before the winner can go out. ## Choose what to test The **What should we test?** dropdown offers five test types: A screenshot of the What should we test dropdown open, listing Subject line, Whole email, From address, Day of week, and Time of day.

A screenshot of the What should we test dropdown open, listing Subject line, Whole email, From address, Day of week, and Time of day.

| Test type | Each variation is… | Good for | | - | - | - | | **Subject line** | A different subject on the same email | The classic — biggest lever on opens | | **Whole email** | A different **approved** email entirely | Layout, copy, or offer showdowns | | **From address** | A different from name, from address, and reply-to | "Founder vs. brand" sender tests | | **Day of week** | The same local time on a different weekday | Finding your audience's day | | **Time of day** | A different local time | Morning vs. evening sends | Changing the test type starts a fresh set of variations — your existing variation setup is discarded. ## Build the variations **Variation A** is always the **control**: your campaign's current setup. Add up to four more with **Add variation** (a test needs at least two). A screenshot of the Variations section with Variation A marked Control and prefilled with the campaign subject, an empty Variation B, an Add variation button, and a Suggest variations button.

A screenshot of the Variations section with Variation A marked Control and prefilled with the campaign subject, an empty Variation B, an Add variation button, and a Suggest variations button.

### Let AI write the alternatives For **Subject line** and **From address** tests, click **Suggest variations** and Tented generates alternatives alongside your control — each with a one-line rationale explaining what it's testing, so the test is a real experiment rather than five shots in the dark. Keep them, edit them, or **Regenerate variations** for a new set. A screenshot of AI-suggested variations, each with an AI suggested badge and a rationale line beneath the variation name.

A screenshot of AI-suggested variations, each with an AI suggested badge and a rationale line beneath the variation name.

### Whole-email variations For a **Whole email** test, Variation A sends the approved campaign email. For each other variation, pick another **approved** email — or click **Clone original email** to duplicate the control and edit the copy in the email editor. Clones must be approved before the test can be saved. A screenshot of a Whole email test where Variation A sends the approved campaign email with an Edit control email link, and Variation B has a Variation email picker.

A screenshot of a Whole email test where Variation A sends the approved campaign email with an Edit control email link, and Variation B has a Variation email picker.

### Timing variations * **Day of week** — set one **Send time for every variation**, then give each variation its own weekday. * **Time of day** — give each variation its own local time, interpreted as wall-clock time in the test's timezone. A **Winner delivery weekdays** setting controls which days the winning time may be delivered on — after winner selection, delivery waits for the winning local time on an eligible weekday. A screenshot of a Time of day test showing per-variation local time inputs and the Winner delivery weekdays radio options.

A screenshot of a Time of day test showing per-variation local time inputs and the Winner delivery weekdays radio options.

## Split the audience **Audience allocation** decides how much of the audience is the experiment and how much gets the winner: * **Test sample** (default **20%**) — divided **evenly** across your variations. * **Winner audience** (default **80%**) — held back until a winner is chosen, then receives the winning version. The two always total 100% (editing one rebalances the other), and both must be whole percentages between 1 and 99. A screenshot of the Audience allocation section with Test sample at 20 percent and Winner audience at 80 percent, above the Winner selection radios.

A screenshot of the Audience allocation section with Test sample at 20 percent and Winner audience at 80 percent, above the Winner selection radios.

The audience needs at least **one more qualified contact than you have variations** — every variation gets at least one sample recipient, and at least one contact is always held back for the winner send. ## Pick the winner (or let Tented) **Winning metric** is how variations are ranked: | Metric | Measured as | | - | - | | **Opens** | Unique opens ÷ delivered | | **Clicks** | Unique clicks ÷ delivered | | **Click-to-open** | Unique clicks ÷ unique opens | **Winner selection** has two modes: * **Choose automatically** — at the **Winner decision time**, Tented ranks the variations by your metric and sends the best performer to the winner audience. On a perfect tie, the variation with more data wins, and the control wins over a challenger that merely matched it. * **Declare manually** — you pick the winner yourself from the results view once the sample has collected. Tented emails you reminders after 6 and 24 hours, and a final reminder after 48 hours. The winner audience waits until you declare — nothing is sent without your pick. ## Set the timing Two datetime fields close out the setup: * **Test start time** — when the sample sends go out. This *is* the blast's send time. It can stay blank while you draft, but scheduling requires a start at least **2 minutes** in the future. * **Winner decision time** (automatic selection only) — when Tented evaluates results and sends the winner. It must land **after every variation has sent**. For subject-line, whole-email, and from-address tests the winner goes out right at decision time; day-of-week and time-of-day winners wait for the next eligible occurrence of the winning slot. A screenshot of the bottom of the AB Test configurator with Winner selection radios, Test start time and Winner decision time pickers, and the Save AB test button.

A screenshot of the bottom of the AB Test configurator with Winner selection radios, Test start time and Winner decision time pickers, and the Save AB test button.

## Save it, then schedule it Click **Save AB test** — the badge in the AB Test header flips from **Unsaved changes** to **Saved**. Then launch from the top of the Schedule step, where the usual Schedule send button now reads **Schedule test**: it approves the blast and schedules the test to begin at your start time. A screenshot of the Schedule step with a saved AB test, showing the Schedule the AB test card with a Schedule test button and the AB Test section marked Saved.

A screenshot of the Schedule step with a saved AB test, showing the Schedule the AB test card with a Schedule test button and the AB Test section marked Saved.

Back on the **Campaigns** page, blasts with a test carry a violet **AB TEST** badge next to their status: A screenshot of the Campaigns list where the draft blast Terra Collection Launch has a violet AB TEST badge beside its Draft status.

A screenshot of the Campaigns list where the draft blast Terra Collection Launch has a violet AB TEST badge beside its Draft status.

## Watch the results roll in Once scheduled, the campaign page gains an **AB Test** card that tracks the whole lifecycle — it refreshes itself, so you can leave it open: A screenshot of a completed AB test overview showing test timing tiles, the audience split, winning metric, selection mode, and the selected winner Variation C.

A screenshot of a completed AB test overview showing test timing tiles, the audience split, winning metric, selection mode, and the selected winner Variation C.

A screenshot of the AB test results table with five subject line variations, where Variation C is highlighted with a Winner badge at a 53 percent open rate.

For a **Declare manually** test, each row gains a **Declare winner** button once results are collecting. Declaring sends that variation to the winner audience — for content tests you also pick the **Send winner at** time. The choice can't be changed after winner selection. Second thoughts mid-test? **Cancel winner send** is available while the test is sending, collecting, or has a winner scheduled. The remaining audience simply never receives the blast, and the test completes with the sample results only — this can't be undone. The blast's regular [results dashboard](/email/email-blasts#watch-the-results-roll-in) keeps working throughout, aggregating the sample sends and the winner send together, with every metric clickable for a per-contact drill-down. **Free plan:** the daily 100-email limit and 1,000-contact allowance are re-checked at each send moment — the sample sends and the winner send each count against the day they fire. ## What's next? See how an email performed across every blast and flow it was sent in. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.