A conversion rate is a defined action divided by a defined eligible population. Compare rates only when events, counting units, filters and periods align. Percentage-point change describes the distance between rates; relative change describes that distance relative to the starting rate.
Your deliverable: Calculate and compare rates without mixing events, populations or reporting periods.
Use compatible counts where each opportunity can contribute at most one event. This tool does not model every native store metric.
Illustrative defaults, not client performance. Descriptive arithmetic only: no significance test or causal inference. Inputs stay in this page and are not saved.
01
Write the fraction in words.
Every rate needs an event in the numerator and an eligible population or opportunity in the denominator. “Conversion is 20%” is incomplete. Specify the report, period, filters and whether the event represents intent, a completed action or another stage.
The calculator in this guide compares two compatible event-to-opportunity rates. It does not recreate a store’s native metric or decide whether the change was caused by your listing.
02
Read the current source definition.
Apple’s reference defines its App Store conversion measure using downloads and pre-orders over unique-device impressions. It is not simply installs divided by page views. Preserve the platform’s exact scope in your report.
Google’s listing-performance documentation now describes click-based measures and distinguishes them from legacy acquisition reporting. Check which report and period you are using; an install click is not a completed install.
03
Distinguish percentage points from relative change.
A rise from 20% to 25% is five percentage points. Relative to the original rate, it is a 25% increase. Both descriptions can be mathematically correct, but they answer different questions and should not be used interchangeably.
If the initial rate is zero, relative change is undefined. Report the actual rates and percentage-point difference. Never replace an undefined comparison with an impressive-looking percentage.
04
Align the inputs before calculating.
Use the same event definition, audience scope and counting method in both periods. A change from unique users to events or from downloads to clicks breaks comparability. Keep known reporting changes visible rather than smoothing them into one trend.
Do not confuse arithmetic with evidence of effect.
The calculator describes the difference between the numbers entered. It does not test significance, adjust for mix or establish causality. Small samples and concurrent changes can produce differences that do not support a rollout decision.
Use a written experiment hypothesis and an appropriate evaluation plan. The next question after a rate changes is what evidence explains it, not which headline makes it sound largest.
WORKED EXAMPLE / FICTIONAL APP
The same app can have several valid rates
Consider a fictional reporting sample with 10,000 impressions, 2,000 listing visitors, 400 install clicks and 300 completed installations. These are deliberately simplified counts, not a claim about a particular store’s report. Different ratios answer different questions, and none should be labelled merely “the conversion rate” without explanation.
How to reason through it
In this example, click-to-visitor rate is 400 divided by 2,000, or 20%. Completed-install-to-visitor rate is 300 divided by 2,000, or 15%. Completed-install-to-impression rate is 3%. The arithmetic is simple; the analytical risk is silently changing the event or denominator while retaining the same label.
The next practical move
Write the formula beside the chart and identify the reporting population. If a visitor-based rate moves from 15% to 18%, that is a three-percentage-point increase and a 20% relative increase. It is not proof that a creative change caused the movement. Check audience mix, dates and concurrent activity before moving from description to attribution.
Swipe horizontally to compare the decision map →
The same app can have several valid rates: illustrative decision map
Decision area
What to inspect
What it means for the work
400 / 2,000
20%
Install clicks per listing visitor in this example
300 / 2,000
15%
Completed installations per listing visitor
300 / 10,000
3%
Completed installations per impression
The trap to avoid
Do not compare rates across different stores, sources or time periods until the definitions are aligned. An attractive benchmark can be irrelevant if it counts a different action or excludes a different population.
When the situation gets more complicated
Inspect composition before interpreting an overall rate. If high-intent branded visitors become a smaller share of traffic, the combined rate can fall even when each segment behaves consistently. Compare meaningful segments using consistent definitions and avoid attributing a mix shift to a screenshot. When data is sparse, state that segmentation may be too noisy to support a firm decision.
Reporting definition check: Google Play’s current listing-performance documentation distinguishes click intent from completed acquisitions. Preserve the actual event name in your worksheet and comparisons.
FROM METHOD TO DELIVERABLE
Why the overall rate can fall without a segment getting worse
This constructed example keeps both segment rates constant while changing their shares of traffic. All actions and visitors use the same hypothetical definitions. It illustrates arithmetic, not an observed client result or a native store funnel.
Swipe horizontally to read the complete example →
Synthetic audience-mix example: each segment is unchanged, but the combined rate declines.
Segment
Period A: actions / visitors
Period B: actions / visitors
Higher-intent visitors
160 / 400 = 40%
80 / 200 = 40%
Other visitors
60 / 600 = 10%
80 / 800 = 10%
Combined population
220 / 1,000 = 22%
160 / 1,000 = 16%
Change in combined rate
Baseline: 22%
−6 percentage points; approximately −27.3% relative
Calculate the combined rate from counts
Add the actions and eligible visitors, then divide. Do not take an unweighted average of segment percentages unless the denominators are equal. Period A has 220 actions among 1,000 visitors; Period B has 160 among 1,000. Both groups behave the same within the example, so changing the listing is not an explanation for the constructed decline.
State the comparison precisely
The combined rate falls from 22% to 16%, a six-percentage-point decline. Relative change is (16 − 22) / 22, approximately −27.3%. Include the starting rate so a reader can reproduce the calculation. If the baseline is zero, relative change is undefined; if the eligible denominator is zero, the rate itself is undefined.
Use segmentation without manufacturing certainty
In real data, inspect meaningful supported segments, their sizes and comparable definitions. Very small groups can produce unstable rates, and overlapping groups cannot simply be added. A mix explanation is a hypothesis until the underlying data supports it. Check that a reporting-definition change or duplicate events has not created the apparent composition shift.
A team reports a move from 10% to 12% as a “2% uplift.” Rewrite the statement so the arithmetic is unambiguous.
Reveal the reasoning
The rate increased by two percentage points, equivalent to a 20% relative increase. Add the event, denominator, audience and periods before presenting it as a real result. The statement still does not establish causation.
Your working brief
Use your own app for this exercise. Fill in the six fields below and attach the evidence behind each answer. If you cannot complete a field, write the missing question and the person who can resolve it. Do not replace unknowns with optimistic assumptions.
Source report and exact metric name
Numerator event
Denominator and eligibility
Dates, audience and filters
Absolute percentage-point change
Relative change and attribution limits
For each field, distinguish an observation from an interpretation. Add the date and source where available, then identify the decision that follows. A colleague should be able to trace your recommendation back to the evidence without needing you to explain the entire project again.
