Run a Test across Campaigns

Outcome

You’ll end up with one Test that keeps everything about a question in one place: what you asked, the two approaches you compared, the Campaigns that ran the split, how each one came out, and what you concluded. Once you conclude, the question and results are locked in with your words. Then you can share it with the members you choose.

Sendfire reads the results. You still build and send the split tests in Klaviyo, exactly as you do today.

Before you start

• You need to be a marketer for the Brand. Only marketers can create or edit Tests.
• The Brand needs to be connected to Klaviyo before any results arrive. You can add a Campaign to a Test before it’s linked to its Klaviyo Campaign, but you can’t confirm which variant is which until the link exists and at least two variants have synced.
• A Campaign can only be in one Test at a time, and every Campaign in a Test has to use the same channel, email or SMS.
• Sendfire won’t tell you a result is statistically significant. It can’t confirm three things about a split today: when the numbers were last refreshed, whether they cover the window you asked for, and whether the winning variant was held back from the rest of the list. So every Campaign with paired results reads Split not verified, and the comparisons are yours to judge.

Procedure

1. Create the Test. Open Tests from Campaigns and choose New Test. Pick the Brand, give the Test a name, and write down the question you’re trying to answer. Then choose the channel, the metric that matters most, and a Test Category such as Subject Line or Send Time. Describe your two approaches in the Control Description and Variant Description fields, pick how many days after each send the results should cover (3, 5, 7, or 14), and choose Create Test.

2. Add Campaigns as you plan them. On the Test, choose Add planned Campaign and pick a Campaign. It doesn’t need a Klaviyo link yet. You can also start from the Campaign itself: its header shows Add to Test until it joins one, then Test setup. If a Campaign is already in another Test, you’ll see that Test’s name next to it and won’t be able to pick it.

3. Build the split test in Klaviyo. For each Campaign in the Test, set up the A/B split in Klaviyo with the same two approaches, and send it from Klaviyo. Keeping the approaches consistent from send to send is on you. Sendfire reads what ran; it doesn’t create, send, or stop anything.

4. Link each Campaign and confirm the approaches. Link the Campaign to its Klaviyo Campaign from the Campaign page. Once its variants have synced, open Manage participation on its row in the Test, or Test setup on the Campaign, and match each approach to the Klaviyo variant that ran it. Choose Confirm approaches. Until you do, the row reads Awaiting Klaviyo link or Confirm approaches, and that Campaign doesn’t count towards the results.

5. Read the results as they come in. Results show up once Sendfire has synced a Campaign’s variant results, and they’re recalculated every time you open the Test. Under Where it stands you get one plain sentence about which approach led and in how many Campaigns, the combined figures for your primary metric, and a note on what Sendfire couldn’t verify. All metrics breaks out control, treatment, and the difference for open rate, click rate, click-to-open rate, revenue per recipient, and unsubscribe rate, wherever Klaviyo supplies them. Campaigns lists every send with its own numbers and an Evidence label. The difference is always treatment minus control, in points. One Campaign leading isn’t an answer. Look for the same approach leading across your sends, and say in your conclusion how many it led on.

6. Exclude a Campaign that didn’t run as planned. Open its row, write the reason, and choose Exclude with reason. It stays in the Test with the reason showing, but drops out of the combined figures. Remove from Test takes it out completely; the Campaign itself isn’t touched.

7. Write the conclusion. Under Conclusion, choose the outcome that fits: one approach is supported, no meaningful difference, mixed results, inconclusive, or not enough evidence. The first option names your control or your treatment by its description. Then write what you learned, including the limits, and what to do next. Mixed or incomplete results are fine to conclude on. Save draft keeps your text private and survives a refresh. If you want a starting point, Suggest interpretation with AI drafts text you can edit; it uses your AI allowance and never concludes for you.

8. Conclude. Tick I have reviewed the current question, Campaign results and limitations. Once that’s ticked and you’ve written an interpretation, Conclude Test becomes available. Sendfire records the question, both approaches, and the results exactly as they stood, together with your conclusion and today’s date.

9. Share it with members. On the concluded Test, choose Share conclusion, pick the members who should see it, and confirm. Each of them gets a notification and can open that exact version. Concluding and sharing are separate steps on purpose: you can conclude something and keep it to yourself.

What happens next

• The Test is locked. You can’t edit the question, change which Campaigns are in it, or delete it. Later syncs don’t change the recorded results.
• Members read the shared version. Until you share, members who aren’t marketers see only the Test’s name, its status, how many Campaigns are in it, and how many have verified evidence, which is 0 today. After you share, they read the version you sent and can comment or ask a question on it. See What members see.
• Sharing again. Choosing Share conclusion again sends the same conclusion as a new round to whoever you pick that time. There’s no withdraw control on the Test page. Someone who received a version keeps it for as long as they have access to the Brand.
• Brand Memory is optional. Assess Brand Memory with AI proposes what this Test should teach the Brand’s memory. Nothing goes into Brand Memory until you accept the proposal, and a Test can be concluded and shared without ever being assessed.
• A follow-up question is just a new Test. There’s no follow-up Test type.

When it does not look right

• No Klaviyo link or Approaches never confirmed on a concluded Test. That Campaign was in the Test but never contributed results. It’s listed for the record and left out of the combined figures.
• Measurement window open until a date. That Campaign’s window hasn’t closed yet, so its numbers can still move.
• A Campaign has no paired counts. The limitation note under Where it stands names it and leaves it out of the combined figures. Check the Campaign’s Klaviyo link and its variant mapping.
• “The evidence changed. Your text is preserved; review the updated results before concluding.” New results arrived between you reading the page and clicking. Read the updated results and conclude again. Your draft is still there.
• You changed the question, approaches, metric, or window after confirming variants. Sendfire keeps your variant mappings and your draft, but asks you to look at the results again before concluding, because what you reviewed has changed.
• “Conclusion shared. A notification could not be sent.” The conclusion was shared; only the email failed. Choose Retry failed notifications. A retry never creates another version.
• “The conclusion could not be published. Check the selected members and current access.” Nothing was shared. Check that the members you picked still have access to the Brand, then try again.
• An AI suggestion or assessment fails. Your saved text and conclusion aren’t affected. Try again, or carry on without it.

Related

• Campaign Tests: what a Test owns, its states, how results are read, and what members can see.