We make it easy to run tests and get results you can rely on.

Zembula provides the testing capabilities you need to justify budget for personalization at scale.

  1. Assign at open
  2. Hold for the window
  3. Re-run on demand
A split test running in the daily email: a four week window of daily sends, the Smart Banner against a control that shows nothing, and the results.

Zembula customers

We power 30+ billion personalized messages annually

  • J.Crew
  • Madewell
  • NOBULL
  • Thrive Causemetics
  • SPANX
  • Joybird
  • Sephora
  • Pair Eyewear
  • Vail Resorts
  • Winn-Dixie

Pick the testing method that provides the evidence you need to justify budget

Every test has its place, but finance and executive teams want to see test results that prove incremental revenue.

Criteria
Revenue per send
Single-send A/B test
Block-level longitudinal split test

What it measures

How much revenue was attributed to a specific send.

Which of two creatives won with a single day’s audience.

The causal effect of a block strategy on channel revenue across a full window.

What it hides

If you send less email, your revenue per send will always go up.

Everything after the send: memory, fatigue, journey effects. The winner can flip next week.

Very little. Both groups live through the same sends, seasonality, and promos.

Incrementality

No

None. It moves credit between channels and sends.

No

Weak. One exposure, one moment, no durability.

Yes

Yes. A randomized difference is a causal one.

Repeatable

No

It isn’t a test at all.

No

Re-run it and the answer often changes.

Yes

Re-run the window and the number should come back.

CFO verdict

A reporting metric, not evidence.

Suggestive, not durable.

Evidence. Randomized, held, reproducible.

Revenue per send

  • What it measures

    How much revenue was attributed to a specific send.

  • What it hides

    If you send less email, your revenue per send will always go up.

  • Incrementality

    No

    None. It moves credit between channels and sends.

  • Repeatable

    No

    It isn’t a test at all.

  • CFO verdict

    A reporting metric, not evidence.

Single-send A/B test

  • What it measures

    Which of two creatives won with a single day’s audience.

  • What it hides

    Everything after the send: memory, fatigue, journey effects. The winner can flip next week.

  • Incrementality

    No

    Weak. One exposure, one moment, no durability.

  • Repeatable

    No

    Re-run it and the answer often changes.

  • CFO verdict

    Suggestive, not durable.

Block-level longitudinal split test

  • What it measures

    The causal effect of a block strategy on channel revenue across a full window.

  • What it hides

    Very little. Both groups live through the same sends, seasonality, and promos.

  • Incrementality

    Yes

    Yes. A randomized difference is a causal one.

  • Repeatable

    Yes

    Re-run the window and the number should come back.

  • CFO verdict

    Evidence. Randomized, held, reproducible.

Learn faster by running multiple independent tests at the same time

Zembula’s tests run at the content block level, so you can have many independent tests running in the same email.

One email with two blocks lit, the Smart Banner and the Product Grid, each wired to its own longitudinal split test card with its split and result.

When audiences are made at the email open, you get better results

The hardest part of a longitudinal split test is managing the audiences over time, and we make that easy.

An open email above a dotted ring: people who opened are navy dots inside, tagged Opened, in; the rest are hollow dots outside, tagged Not yet, out.

Accurate audience

Who to include

Setting the audience at first email open makes sure the audience is as accurate as possible.

A new subscriber above a level balance: pan A holds nine navy dots, pan B nine periwinkle dots, and a lime Balanced tag sits under the two pans.

New subscribers

Retain balanced audiences

Assigning new subscribers at first open makes sure your audiences stay perfectly balanced.

The path to an order: search, social and an ad as lavender steps, then email as a lime step tagged Last click, joined to the navy order node.

Last click channel attribution

Reliable results

Last click channel attribution captures the most conservative and accurate view of results.

Analytics steer strategy day to day. Longitudinal tests prove channel revenue lift

Zembula’s day-to-day analytics answer where and how: which placements, which offers, which designs. The longitudinal test answers whether the whole thing pays.

Analytics

Design and personalization learnings

Use analytics for day-to-day design decisions, decisioning logic, and what data to use together within design. RPM and CTC tell this story effectively.

  • RPM: Revenue per thousand impressions
  • CTC: Click to conversion

Longitudinal tests

The proof your finance team needs

The longitudinal split test measures transactions and revenue from each audience over weeks, so that you can determine incremental transactional and revenue lift.

  • Held-out control: Compare exposed vs. control groups
  • Full-window read: Measure the transactions and revenue gap over time
  • Repeatable verdict: Re-run with fresh assignment to prove it again

Prove what actually drives revenue

Run a longitudinal split test and measure the real revenue lift.