Attribution and Reporting Without Fooling Yourself
A source-aware way to read email and SMS attribution: record the model, compare stable cohorts, separate reporting layers, and validate material decisions with stronger evidence.
- Choose the right retention metric and denominator
- Turn performance signals into an operating decision
Platform attribution is a configured reporting convention, not a causal proof that one message created every credited order. It can still be useful, if the team records the live model, compares like periods, separates channels and journeys, and uses a stronger method when the decision is material.
Every dashboard should state its channel, attribution model, conversion window, denominator, date range, exclusions, and owner. If those fields are missing, the number is not ready to compare.
How a message can receive platform credit
- 01A customer receives or interacts with a message
The event is evaluated under the account’s configured channel and attribution settings.
- 02Other touches may occur
The customer journey can include direct visits, paid media, referrals, organic search, or other messages.
- 03A qualifying order occurs
If the configured attribution conditions are met, the platform records credit according to its model.
- 04The team interprets the report
Use the result as a consistent operational signal, then validate an important causal claim with a stronger design.
Illustrative reporting sequence. Check the current account configuration and the linked platform documentation before using a default setting.
Why is my attributed email revenue inflated?
Attributed revenue can be higher than incremental contribution because a reporting model may credit an order that would have happened without the message. Klaviyo documents a configurable last-touch model for new accounts, including default lookback windows for email interactions; verify the live account settings because the defaults can change and may have been edited.
Klaviyo's current documentation describes separate, configurable channel windows and notes that its model can attribute a conversion to the last qualifying message interaction within an open window. It also documents controls for excluding Apple Mail Privacy Protection opens and bot clicks from attribution. That makes the setting itself part of the metric definition.
A platform report can be decision-useful without being an incrementality estimate. Do not promise that attributed revenue is always inflated by a fixed amount, or that a credited order would definitely have happened without the message. Use an experiment or a credible counterfactual when that distinction changes budget, discount, or staffing decisions.
How do I check email attribution against real store lift?
Compare platform-attributed email and SMS revenue with total-store movement over the same period, then use a holdout or other causal design when you need an incremental estimate. If the store does not move with the attributed dashboard, investigate the denominator, attribution window, channel overlap, and promotion mix before claiming lift.
| Check | Keep stable | What it can tell you |
|---|---|---|
| Platform attribution | Channel, model, window, date range, exclusions | The platform-reported contribution under that configuration |
| Store movement | Store-revenue denominator, promotion mix, and date range | Whether the reported pattern is directionally consistent with the business |
| Cohort or holdout | Eligibility, treatment, control, and outcome window | A stronger estimate of incremental effect when the decision warrants it |
Do not change campaign volume, offer, audience, attribution window, and reporting period at once and then declare the winner. State the test question first.
Why do blended email metrics hide problems?
Blended metrics can let one healthy channel, flow, or audience conceal deterioration elsewhere. Report email, SMS, flows, campaigns, and major audience cohorts separately before using the combined number as an executive summary.
- Useful for a high-level trend
- Can conceal a weak flow, campaign, provider, or audience source
- Needs a drill-down before it drives a tactical change
- Separates flows from campaigns and email from SMS
- Compares defined cohorts under the same settings
- Shows outcomes alongside unsubscribes, complaints, and delivery signals
Split reporting does not make the number perfect. It makes it possible to see where the next question belongs.
What goes in a weekly email report vs a monthly review?
Use a lightweight recurring pulse to surface problems, and a deeper review to decide what to change. The exact meeting rhythm should match the program's volume and decision cadence; do not make a calendar ritual more important than the action it produces.
| Review | Primary purpose | Include |
|---|---|---|
| Operating pulse | Surface an issue that needs attention now | Delivery and complaint signals, what shipped, what is scheduled, and a concise outcome view |
| Decision review | Choose the next priority or test | Stable revenue definitions, flows versus campaigns, audience movement, list health, margin context, and the next test question |
A good reporting meeting ends with an owner, a decision, and the definition that will judge the result.
Which email metrics actually guide decisions?
Use a small metric stack with stable definitions. Revenue per eligible recipient helps compare commercial output under one recorded attribution model. The canonical email-and-SMS share and flow-share cards below give ZHS operating context, but neither card is a universal target or causal claim.
Forecast a send without confusing it with incrementality
Eligible recipients × observed qualifying-order rate × average order value = attributed-revenue scenario- Use a comparable historical cohort and the same attribution definition.
- State the offer, channel, audience, and send context.
- Treat the result as a scenario; validate incremental impact separately when it matters.
A planning calculation, not a claim that every attributed order was caused by the message.
- Revenue per eligible recipient. A normalized output measure when the eligible audience and attribution settings are stable.
- Attributed email-and-SMS share. A directional operating signal under the documented model.
- Flows versus campaigns. Separate these before deciding whether journey logic or planned messaging needs work.
- Reliable engagement and permission guardrails. Use clicks, purchases, complaints, unsubscribes, delivery, and consent context rather than privacy-affected opens alone.
- Incrementality evidence. Use a holdout or credible causal method for high-stakes decisions.
What are the most common attribution reporting mistakes?
The recurring failure is treating a configured dashboard number as though it were a self-explanatory causal result. The checks below keep the report tied to a definition and a decision.
- Reporting attributed revenue as causal fact. Label the model and use stronger evidence where the causal distinction matters.
- Forgetting the account settings. Record the live windows, exclusions, and changes with the report.
- Blending unlike programs. Separate flows, campaigns, channels, cohorts, and meaningful audience sources.
- Changing too many variables at once. A result is not diagnostic when audience, offer, timing, and model all moved together.
- Ignoring permission and delivery guardrails. Commercial output without list health is not a durable win.
- Treating revenue as profit. Bring margin, incentive cost, and fulfillment context into decisions that change the offer.
Get Expert Help
Our team can build a reporting system that makes attribution settings explicit, separates the layers that need different decisions, and defines the next test before the dashboard is read.
Need help implementing this?
We build and manage complete email & SMS programs for DTC brands. Get a custom plan for your brand.
Apply Now