Is Klaviyo's attributed revenue accurate?

How does Klaviyo attribution actually work?

Klaviyo uses a cooperative last-touch model: the most recent qualifying touch whose window is still open receives full credit for an order. Credit is never split by default; a linear multi-touch option exists only on Klaviyo’s paid analytics tiers.

The defaults, for accounts created after 9 October 2024:

Touchpoint	Default window	Counts by default?
Email click	5 days	Yes
Email open	5 days	Yes, including Apple auto-opens
SMS click	5 days	Yes
SMS open	1 day	Yes
SMS delivered	12 hours	Yes, no engagement required
Push open	24 hours	Yes
Form submission	2 hours	Separate system, can double-link with a message

Three details in that table do most of the work in any accuracy argument, and most coverage misses all three. The window starts when the message is delivered, not when it is engaged with. A delivered SMS can claim an order with no click and no open at all. And accounts created before October 2024 may still be running older defaults, such as 1-day SMS windows, so two brands comparing “Klaviyo revenue” may not be comparing the same measurement. When Everboost audits an account, the attribution settings page is checked before any revenue number is read, for exactly this reason.

So is the number accurate?

Wrong question, and getting the question right defuses most of the argument. Attributed revenue is a claims model: it answers “which message most recently touched this buyer”, not “which message caused this purchase” and certainly not “how much cash came in”. Judged as a ledger it will always fail, because it is not one. Every ESP on the market works the same way; Klaviyo is simply the one your finance team is looking at.

Judged as a claims model, its generosity depends entirely on settings. The documented ways the default number runs hot: refunds and cancellations are not subtracted (Klaviyo’s own Shopify documentation names this as a reason its figure exceeds Shopify’s), order value includes shipping, Apple’s automated opens count as engagement unless excluded, a delivered text can claim an order within 12 hours untouched, and subscription renewals arrive as ordinary Placed Order events, so a routine newsletter opened two days before a scheduled renewal charge can claim that renewal’s full value. That last one matters enormously for subscription brands: a flow can look heroic while measuring little more than billing cadence.

But attribution error is not one-directional, and honesty requires saying so. Klaviyo’s model counts nothing it cannot see: cross-device journeys Google Analytics misses, and misconfigured accounts that under-credit. When Everboost rebuilt Mersey Raw’s reporting, fixing attribution increased measured campaign revenue by 25.1%; the previous setup had been quietly dropping credit the programme had earned. Settings cut both ways.

Why does Klaviyo never match Shopify?

Because they are answering different questions, and Klaviyo documents the biggest mechanical gap itself: “Shopify subtracts canceled and refunded orders from their revenue calculation, while Klaviyo does not.” Add session-based versus profile-based identity, shipping inclusion, and the separate forms system that can link the same order to both a form and a message, and a permanent gap is not a bug, it is the design.

The practical rule Everboost applies across client reporting: Shopify is the ledger, Klaviyo is the influence estimate. Report Klaviyo-attributed revenue as a share of Shopify net revenue on locked settings, and never feed Klaviyo’s figure into CAC or ROAS maths built on ledger numbers.

What settings should you actually run?

Klaviyo ships more correction controls than most platforms, and most accounts never open them. The configuration Everboost recommends, from running this across client accounts:

1. Keep opens, but exclude Apple MPP opens. Some advisers go clicks-only, and Klaviyo’s own blog concedes that a clicks-first setup “avoids over-reporting”. We think that overcorrects: plenty of real buying starts with an email that gets read and no link clicked, and once Apple’s automated fetches are excluded, the opens that remain are genuine human signals. Discarding them under-credits the programme.
2. Shorten the open window to 1 to 3 days. This is the compromise that makes keeping opens honest. An open is a weaker signal than a click, so it deserves a shorter causal claim than the 5-day default. Clicks keep their 5 days.
3. Keep bot-click exclusion on, and consider unticking delivered SMS unless you genuinely want received-equals-touched.
4. Preview before saving with the built-in Model comparison tool.
5. Document the date and the old settings before changing anything. Klaviyo recalculates your entire history under new settings, taking up to 36 hours, and there is no frozen record of what was previously reported. Undocumented changes are how “why did last quarter’s revenue change?” conversations happen.

One caution: tightening will drop your attributed revenue share, and that drop is a measurement change, not a performance change. Make the change between reporting periods and say so.

How do you verify the number is real?

Three habits, in ascending order of effort.

Triangulate with conversion events. Attributed revenue should move with attributed orders and with checkout starts. Revenue rising while attributed orders fall is a settings artefact, not a performance story.

Reconcile monthly. Klaviyo-attributed revenue as a share of Shopify net revenue, on locked settings, tracked as a trend. The absolute share is debatable; the trend is not.

Test incrementality on big claims. Klaviyo’s blog argues attribution debates are performance theatre, and on this point it is right: no model changes how much money you made. Hold out a comparable audience from the flow or campaign in question and compare purchase rates. Caused revenue is the only number that settles an argument with a CFO.

The number that matters more

Here is the position that makes this whole debate smaller: attributed revenue, however configured, measures credit assignment between your own messages. It cannot tell you whether customers are actually coming back more often, which is the thing the programme exists to change. A brand can grow attributed revenue for two quarters while its repeat purchase rate stands still, usually by sending more.

Everboost’s job is turning one-time buyers into repeat customers, so client programmes are judged on repeat purchase behaviour, cohort by cohort, with attributed revenue as a steering metric rather than a scoreboard. Configure the window honestly, reconcile it against the ledger, and then spend your attention on the repeat purchase rate your whole programme exists to move. And if you use the Klaviyo MCP server for reporting, the same discipline applies: ask for attributed metrics explicitly, and check the settings before trusting anyone’s number, including the AI’s.