Campaign Tests

What a Test is

A Test is one question about a Brand’s email or SMS, answered across several Campaigns rather than a single send. The marketer names two approaches, a control and a treatment, picks the metric that matters, and says how many days of results each send should cover. Each Campaign runs the two approaches as a Klaviyo split test, and the Test collects its paired results: the control figure and the treatment figure from the same send, side by side. When the marketer has seen enough, they write down what they concluded, and Sendfire locks the question and the results in with it.

The point is the pattern. A single split shows what happened once. A Test lets you see whether an approach kept working across several sends.

What members see

If you’re a member of a Brand but not one of its marketers, you follow Tests from the Brand’s published face.

• Before a conclusion is shared, you see the Test’s name, whether it’s Collecting or Concluded, how many Campaigns are in it, and how many have verified evidence. That last number is 0 today, because Sendfire can’t verify a split. The question, the approaches, the numbers, drafts, and any Brand Memory work stay with the marketer.
• When a marketer shares a conclusion with you, you get a notification that opens that exact version: the question, each Campaign’s evidence with a note on what couldn’t be verified, the combined results marked as descriptive, and the conclusion with its date. It’s the same recorded content the marketer concluded on; some labels and dates on your view are shown in their raw form.
• What you can do: comment on it, or ask a question about it. You can’t change anything, see drafts, or see whether Brand Memory was assessed.
• What won’t change: the version you received stays as it is, even when new results sync. You keep it for as long as you have access to the Brand. If the marketer shares again, that’s a new round, not a replacement.

What a Test owns and does not own

A Test owns

• the question, and the descriptions of the control and treatment;
• the primary metric and the measurement window;
• which Campaigns are in it, how each one’s Klaviyo variants map to the two approaches, and any exclusion with its reason;
• the marketer’s outcome, interpretation, and next steps, plus the results exactly as they stood when concluded;
• every shared version of that conclusion and who received it.

A Test does not own

• the split test itself. The marketer builds it in Klaviyo, and Klaviyo sends it and ends it. Sendfire reads the variant results and never writes anything back to Klaviyo.
• the Campaign. A Campaign can be in one Test at a time and keeps its own Brief, copy, review, and performance. Take it out of a Test, or delete the Test, and the Campaign and its Klaviyo data are untouched.
• the verdict. Sendfire does the comparisons; the marketer decides what they mean.
• Brand Memory. AI can propose a learning from a concluded Test, but nothing enters Brand Memory until the marketer accepts it.

States

A Test is either Collecting or Concluded. In the Tests list, a concluded Test is labelled Complete.

Collecting. The question and approaches can still be edited. Campaigns can be added, mapped, excluded, or removed, and the Test can be deleted unless an AI analysis has been saved on it. Results are recalculated every time the page opens. Each Campaign carries an Evidence label that says where it stands: Awaiting Klaviyo link, Confirm approaches, Measurement window open until a date, Split not verified, Not enough data, or Excluded with its reason.

Concluded. The marketer chose an outcome, wrote an interpretation and next steps, confirmed they’d reviewed the question, results, and limits, and chose Conclude Test. From then on the Test is read-only: no edits, no changes to participation, no deleting, no reopening. The page leads with the conclusion, then the question, then the results as they were recorded. Two things can still happen: the conclusion can be shared with members, and Brand Memory can be assessed.

Earlier verdict. A Test that was completed before Sendfire kept results alongside conclusions shows its old verdict and date under Earlier verdict, without a results snapshot. The marketer can review today’s results and record a new dated conclusion. Until they do, the Test can’t be shared with members. The earlier verdict stays in the Test’s history.

If the question, approaches, metric, or window change while a Test is collecting, the results the marketer reviewed are out of date. Sendfire keeps the variant mappings and any draft, and asks for a fresh look before concluding.

How results are read

Results come from the Klaviyo variant data Sendfire syncs for each Campaign. For every Campaign with paired results, you see the control value, the treatment value, and the difference as treatment minus control in points. That’s shown for the primary metric and for open rate, click rate, click-to-open rate, revenue per recipient, and unsubscribe rate, wherever Klaviyo supplies them. Combined figures pool the Campaigns that have paired counts, and one sentence says which approach led in how many Campaigns.

These are observations, not verdicts, and the page says so in three ways.

• Combined results are descriptive. Different Campaigns go to different audiences with different splits, so no overall significance is calculated.
• Sendfire won’t tell you a single Campaign’s result is statistically significant. Doing that honestly would need the original split confirmed: when the numbers were last refreshed, whether they cover the requested window, and whether the winner was rolled out to everyone else. Sendfire can’t confirm any of that today, so every Campaign with paired results is labelled Split not verified, and the page lists exactly what couldn’t be verified.
• The count of Campaigns each approach led on is not a confidence score.

The outcomes a marketer can record are: one approach is supported, no meaningful difference, mixed results, inconclusive, or not enough evidence.

Relationship to adjacent objects

• Campaigns. A Campaign can join a Test while it’s still only planned. Its header shows Add to Test before it joins and Test setup after. Confirming approaches needs the Campaign’s Klaviyo link and at least two synced variants.
• Brand Memory. A concluded Test can be assessed with AI. The assessment proposes a learning; the marketer accepts, edits, or dismisses it. A Test can be concluded and shared without ever being assessed.
• AI. AI can suggest an interpretation and can assess Brand Memory. It never concludes a Test, never shares one, and never changes anything in Klaviyo.

Related workflows

• Run a Test across Campaigns