How do you set audit objectives and a baseline?

What you will learn

By the end, you will understand:

Translate a business concern into answerable audit questions
Create a comparable quantitative and qualitative baseline
Document data quality, seasonality and attribution limits

Visual explainer

See the idea clearly.

A business goal connects to scoped journeys and measures; a baseline captures date range, definitions, segments, configuration and external context before future comparison.

Move from symptom to question

Symptom	Audit Question
Traffic is down	Which sources, pages, queries, markets and dates changed, and is the change real after data-quality checks?
Ads do not work	Are the chosen outcomes valid, campaigns eligible, search/audience fit sound, destinations usable and downstream leads valuable?
The site feels outdated	Which priority tasks fail clarity, trust, accessibility, device support or business accuracy?

Write the audit brief

Decision to support
Stakeholders and users
In-scope properties/accounts
Priority journeys
Markets/languages/devices
Date range and comparison
Business outcome definitions
Evidence access
Known changes/incidents
Exclusions
Risk and testing limits
Deliverable and decision date

A baseline is a reproducible snapshot

Record definitions and configuration with the number. “500 conversions” is not comparable if the event changed, duplicate tags were fixed or reporting identity changed.

Combine quantitative measures with qualitative task evidence. A baseline may include completed purchases, qualified calls, task success, accessibility barriers, index coverage and response time.

Choose measures as a chain

01
Business outcome
02
Customer behavior
03
Journey step
04
Observable event
05
Implementation/tag
06
Report definition
07
Quality validation
08
Decision threshold

Segment before averaging

Averages can hide severe differences by device, geography, page type, new/returning visitor, campaign or customer quality. Preselect segments relevant to the question rather than searching endlessly for a favorable cut.

Small segments carry more uncertainty and privacy risk. Suppress or aggregate where needed.

Record context and change history

Seasonality
Price/offer changes
Stock or service capacity
Website releases
Tracking changes
Consent changes
Campaign launches
Search/platform updates
Outages
Competitor/market shifts
Sales process changes
One-off events

Baseline limitations belong in the result

Missing historical data cannot be recreated exactly. Modelled conversions, sampled data, consent gaps, offline decisions and platform attribution should be stated.

A useful baseline can still be created with multiple imperfect sources if definitions are clear and uncertainty is not hidden.

Real-world example

Example: define a qualified enquiry baseline

Example

A services company currently counts every form success as a lead. The audit samples CRM outcomes, removes tests and duplicates, defines a qualified enquiry, records response time and matches source where permission and identifiers allow. It reports both total forms and qualified enquiries so the change in definition remains transparent.

Try this

Build a one-page baseline card

For one objective, record metric definition, source, date range, comparison, segments, configuration, data owner, known gaps, important changes and the decision threshold. Ask whether another reviewer could reproduce it.

Common questions

Questions beginners ask.

What is an audit objective?

The specific decision or question the review is intended to support.

What is the difference between a goal and a KPI?

A goal states the desired outcome; a KPI is a selected measure used to judge progress toward it.

How long should a baseline period be?

Long enough to represent the business cycle and volume while accounting for seasonality and changes; there is no universal duration.

Should last month always be compared with the month before?

No. Seasonal businesses may need year-over-year or matched-period comparisons plus recent trend context.

Can a baseline use modeled data?

Yes if its definition, eligibility and uncertainty are stated and it is not presented as directly observed individual behavior.

What if historical tracking is broken?

Document the gap, use other records where reliable, establish a clean forward baseline and avoid false retrospective precision.

Why segment a baseline?

To reveal meaningful differences that a total or average can conceal.

Can the baseline change during an audit?

Definitions may be corrected, but changes must be documented and earlier comparisons restated consistently where possible.

Assessment

Check what you understood.

5 questions · instant explanations

1. Which is an answerable audit objective?

2. What makes two metrics comparable?

3. Why record change history?

4. What should happen if old tracking is unreliable?

5. True or false: one overall average always represents every important user group.

Sources