# A finance app needs more than a green ranking arrow

Companion to /learn/measurement/aso-kpis/

## Outcome
Build a metric dictionary that connects discovery, listing choice and downstream quality.

Fictional example, not a client result.

## Scenario
BudgetBench is a fictional budgeting app. Its weekly report highlights improved keyword positions, but the product team cares about people completing a first budget. The ranking change may be useful, yet the report currently skips the steps between visibility and a useful customer outcome.

## Reasoning
Choose measures that correspond to different questions. Discovery signals describe whether relevant people may encounter the app. Listing interaction measures describe a defined action in a defined population. Activation or retention measures describe later behaviour. Connect the questions without pretending that every platform report can be joined at an individual-user level.

## Action
Write a metric dictionary with event, denominator, store, audience, source and time window. Use it to review the current report and remove misleading labels. Give each KPI a decision owner: what action might change when this measure changes? Keep diagnostic measures available, but do not present every observable number as an executive success metric.

## Trap to avoid
Do not let a metric become a target just because it is available. A visibility score may help monitor a research set while remaining a poor measure of revenue or user satisfaction. Explain what each metric can and cannot establish.

## Advanced considerations
Create a definition-change log alongside the performance log. If the reporting platform changes an event, preserve the break in the time series instead of drawing an uninterrupted trend. Where different measures cannot be reconciled, display them separately. Honest discontinuities are more informative than a smooth line whose meaning changes halfway across the chart.

## Exercise
A dashboard combines Apple downloads and Google Play install clicks under “installs.” What should happen before trend analysis?

## Suggested answer
Separate the events and correct the labels. Confirm the definitions for each source and historical period. A chart with an inconsistent numerator cannot become reliable through a more attractive visual design.

## Your worksheet

### Business question and decision owner
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Metric label and exact event
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Numerator and denominator
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Store, source, segment and exclusions
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Window, timezone and definition changes
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Interpretation limits and next action
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

## A metric dictionary for BudgetBench
This fictional dictionary demonstrates how to specify measures without merging incompatible sources. The product event is an illustrative instrumentation requirement, not an assertion that a store report supplies user-level attribution.

Illustrative metric dictionary. Source definitions and product instrumentation must be verified for your app.
Question and measure | Definition to preserve | Decision owner and limitation
Are relevant searches finding us? / tracked positions | Fixed query set, store, market, device context and observation date | Research owner; a position does not count installs
Are listing visitors acting? / selected listing action rate | Exact source event divided by its eligible denominator, with filters | Growth analyst; do not merge rates with different definitions
Do users reach initial value? / first-budget completion | Defined product event and eligible new-user cohort in a fixed window | Product analyst; verify tracking and attribution limits
Was the work implemented? / release status | Approved asset version compared with the visible published version | Release owner; completion is delivery evidence, not growth

### Make the dictionary part of the reporting contract
For each KPI, retain the source export, event definition, counting unit, filters and timezone. State whether it counts users, devices or actions. Add an owner who can explain the definition and an action the measure could inform. If nobody can name a decision it supports, move it out of the executive summary and into diagnostic material.

### Treat missing data as a finding
If first-budget completion is not reliably instrumented, label it unavailable and assign a tracking investigation. Do not substitute total sessions simply because the chart is easy to produce. Check event duplication, eligibility and timing before interpreting a newly introduced measure. The first useful deliverable may be a trustworthy baseline rather than a performance conclusion.

### Avoid inventing a joined funnel
A store-level report and an app analytics cohort may describe different populations. Present them as separate evidence when they cannot be reconciled. Do not divide one source’s users by another source’s devices and label the result activation. Explain what is known at each stage and what additional access or measurement would be required to connect them.

## Continue reading
- /learn/measurement/conversion-rate-definitions/
- /learn/measurement/monthly-reporting/
- /learn/measurement/organic-downloads/