Rules
Part of How to think about digital product funnels
Digital product funnel baselines: how to build your own before you trust a range
Digital product funnel baselines, built stage by stage: the counts, denominators and cohort dates to record, and why published ranges stay unsourced until you can name their source.
What to take away
- Digital product funnel benchmarks fail when a rate arrives without its denominator, cohort date and maturity window.
- No published range is used here. None could be traced to a named study, edition or dataset, so the figure leaves each value for your own cohorts.
- Build a baseline from your own cohorts, one stage at a time, then flag any stage that moves outside its own history.
- Delivery and first-value rates matter more than page progression, because they predict refunds.
- Publish counts beside rates so a reviewer can recompute the figure.
Why no published ranges are used
The figure below lists the stages to baseline for a self-serve digital product on paid and organic traffic. The publication could not trace any published range to a single study or a named platform dataset, so the figure gives no values to copy.
Funnel stage benchmark ranges
Typical range
- Visit to email capture
- 25-45%
- Email capture to checkout start
- 8-20%
- Checkout start to purchase
- 40-70%
- Visit to purchase
- 2-5%
- Refund rate within 30 days
- 3-10%
- First-value completion within 14 days
- 30-60%
Outlier low
- Visit to email capture
- under 10%
- Email capture to checkout start
- under 3%
- Checkout start to purchase
- under 20%
- Visit to purchase
- under 1%
- Refund rate within 30 days
- under 1%
- First-value completion within 14 days
- under 15%
Outlier high
- Visit to email capture
- over 60%
- Email capture to checkout start
- over 35%
- Checkout start to purchase
- over 85%
- Visit to purchase
- over 15%
- Refund rate within 30 days
- over 20%
- First-value completion within 14 days
- over 80%
Use the table below to decide what to record for each stage, and what to check first when a rate moves. Each stage should be tracked against its own history, using the same entry rule and window.
| Stage | What to record | Check first when it moves |
|---|---|---|
| Visit to email capture | Visitors, captures, traffic source | Source mix and any page change |
| Email capture to checkout start | Captures, checkout starts, offer shown | Offer and price change |
| Checkout start to purchase | Checkout starts, purchases, payment method | Payment step and currency |
| Visit to purchase | Visitors, purchases, population scope | Retargeting in the denominator |
| Refund rate within 30 days | Refunds, orders in the cohort, delivery method | Mismatch between sales page and delivered file |
| First-value completion within 14 days | Buyers, completions, definition of value | Onboarding steps and content order |
A visit-to-purchase rate far below your own history usually points to a mismatch between the traffic source and the offer. Price is rarely the first cause. A rate far above your history on cold paid traffic can mean the campaign is retargeting existing buyers, so the denominator is not new visitors.
Publish counts before rates
A 3.1 percent visit-to-purchase rate from 2,000 visitors is 62 orders. The same rate from 200 visitors is 6 orders. The 95 percent interval around 6 orders from 200 visitors runs from about 1.1 to 6.4 percent, which is wide enough to change any decision.
Same rate, different confidence
2,000 visitors
- Visit-to-purchase rate
- 3.1%
- Orders
- 62
- Confidence interval
- narrow
200 visitors
- Visit-to-purchase rate
- 3.1%
- Orders
- 6
- Confidence interval
- 1% to 7%
Report the count, the denominator and the interval together. Annotate every change to product, price, traffic source, event rules and consent.
A rate shift that lands the same week as a checkout change is a suspect, not a result. Compare matched internal cohorts first, and treat external figures as questions to investigate.
Separate acquisition scope from campaign scope
In Google Analytics 4, first-user dimensions describe how a user was first acquired, while session dimensions describe each visit. Comparing a first-user source against a session campaign mixes two denominators, and the resulting rate is meaningless. Pick one scope and label it in the export.
Delivery and value rates predict refunds
For automated delivery, correct entitlements divided by paid orders should be close to 100 percent. A shortfall is a defect to find and fix, and it tends to show up as refund requests within days.
First-value completion means the buyer reached the outcome the product promises. Measure it with a stated definition of value, not a page view, and report it inside the same 14-day window for every cohort.
Track refund rate inside a 30-day window. A rising refund rate usually points to a mismatch between the sales page promise and the delivered file. Keep refund counts tied to the cohort they came from, so a late refund is not read against the wrong month. The wider stage model is in how to think about digital product funnels.
Make the comparison reproducible
State the population, the period and the inclusion rules before comparing. Add the calculation, the currency and the maturity window. A single store's funnel result is a local observation. It does not show cause, and it does not transfer to another product without the same definitions.
Recalculate the figure from the cited source or your governed dataset, then record any mismatch. Keep rejected comparisons in the record beside the chosen one, and set the next review date.
Common questions
What is a good checkout completion rate?
No single figure is supported by a named source here. Compare checkout start to purchase against your own history for the same population and payment method. A sudden drop usually points to a payment method, currency or form problem before it points to price. Check the exact start and end states before comparing.
Should medians replace averages?
Use the distribution and the summary that fits the decision. A median hides a long tail of buyers who waited days to complete checkout, and an average hides the same tail differently. Report both when the tail matters.
Can competitor rates be used?
Treat competitor rates as weak context. A competitor's visit-to-purchase rate is comparable only if the definitions, traffic mix, product price, period and data quality match. Otherwise it is a prompt to check your own instrumentation.
Are any published ranges used?
No. The publication could not name a source for any range, so none is given as a target. Your own cohort is the only comparison that supports a decision. Recalculate from your governed dataset before acting.
