Klaviyo Attributed Revenue vs Incremental Revenue: What Should You Measure?

A Klaviyo report can accurately assign revenue to a message without proving that the message created the purchase.

That distinction matters when you decide whether to increase frequency, add SMS, extend a discount or pay for a new personalisation tool. Attribution describes credit under a set of rules. Incrementality asks what changed because of the intervention.

You need both, alongside customer economics. The mistake is asking one number to answer all three questions.

What Klaviyo attributed revenue tells you

Klaviyo uses qualifying interactions and attribution windows to associate conversions with messages. The account’s model, window and interaction settings affect which messages receive credit.

Read the current message attribution documentation and check the actual account settings before comparing results. Document any changes to windows or interaction exclusions. Otherwise a reporting change can look like a performance change.

Attribution is useful for daily operations. It helps identify campaigns, products and journeys associated with customer action. It does not observe the alternative world in which the customer received no treatment.

What incremental revenue asks

Incremental revenue is the additional revenue caused by an intervention relative to what would have happened without it.

A loyal customer might receive an email and then place an order they already intended to place. The email may receive credit even if it did not change the decision. Another message might influence a later purchase outside the attribution window.

Neither possibility makes attribution useless. It means causal questions require a different design.

Comparing recipients with people who happened not to receive a message is usually inadequate. Eligibility, purchase intent and customer value may differ before the send. Random assignment among eligible customers gives a stronger comparison.

A worked example

Consider a hypothetical test with 20,000 eligible customers. Randomly assign 10,000 to treatment and 10,000 to a control, keeping the rest of their experience comparable.

Over the agreed observation window, treatment produces £55,000 in net sales and control produces £50,000. Revenue per eligible customer is £5.50 versus £5.00. The estimated difference is £0.50 per customer, or £5,000 across the treatment group.

The platform might attribute £18,000 to the message. That is a different measurement. It is not automatically the additional revenue created by sending.

These invented figures illustrate the calculation, not a statistically validated result. A real analysis needs uncertainty estimates, adequate sample size and a design that addresses repeat observations and unequal spending.

Contribution changes the decision again

Suppose the same hypothetical test produces £28,000 in contribution before messaging cost in treatment and £27,500 in control. The difference is £500. If the treatment costs £600 to deliver, it has reduced contribution by £100 over that window.

Calculate contribution consistently from net sales after discounts and refunds, then subtract the relevant product, fulfilment, payment and other variable costs. Do not subtract discounts twice if net sales already reflects them.

A test can lift revenue while failing the commercial decision. It may also shift orders earlier rather than create durable additional demand. Choose a window long enough to examine that possibility.

Design the holdout around one question

Define the decision before the test. “Does this replenishment reminder increase profitable repeat purchasing?” is answerable. “Does email work?” is too broad for a small experiment.

Document the eligible population, assignment unit, treatment, control, duration, primary outcome and guardrails. Keep assignments stable when the same customer becomes eligible again, unless the design explicitly accounts for repeated randomisation.

Decide whether you are testing one flow, an added channel or the whole promotional programme. A control for one flow can still receive other campaigns. A programme-wide holdout answers a different question.

Do not remove essential transactional or service communication to test promotional effectiveness. Keep promises and necessary customer support intact.

Watch for contamination and misleading wins

If control customers receive an equivalent SMS or a campaign with the same offer, the experiment measures the difference between those experiences, not treatment versus silence.

If treatment receives a larger discount, the result combines messaging and incentive effects. That may be the decision you want to test, but name it honestly.

Agree the analysis plan before launch. Repeatedly checking results and stopping at the first favourable number can produce misleading confidence. Avoid changing audiences, offers or attribution settings midway without recording the effect on interpretation.

For low-volume brands, a reliable test may take longer than the team expects. An inconclusive result is not proof that the programme has no value.

What a sensible lifecycle scorecard includes

Use an operating view for attributed revenue, orders per recipient, delivery and unsubscribes. Use an experimental view for lift and uncertainty. Use a customer view for second purchase, product progression and contribution at equal cohort age.

The views should explain each other. Strong attributed revenue with weak experimental lift is a reason to investigate cannibalisation or targeting. Strong short-term lift with poor later margin is a reason to review the offer.

Our perfectwhitetee efficiency analysis separates reported revenue growth from revenue-per-delivery changes and their comparison periods. Those outcomes are useful evidence, but are not presented as controlled causal estimates.

Frequently asked questions

### Should we stop reporting attributed revenue?

No. Use it for the operating decisions it supports. Label it clearly and avoid treating it as a causal estimate.

Is changing the attribution window a way to improve performance?

It changes measurement. It does not itself create orders. Record the change and avoid joining incompatible periods into one trend.

What should we test first?

Choose a material, recurring decision with enough volume and uncertainty to justify the test, such as an additional reminder or incentive. Do not begin with a tiny creative variation and expect it to resolve the economics of the entire programme.

If your reporting is driving investment decisions without this distinction, book a consultation. Thrivelia can help connect lifecycle measurement to the customer behaviour and contribution the business needs.

Continue with the worked email revenue-share guide and retention marketing services.