Learn to diagnose performance changes systematically before making optimisation decisions.
Most bad optimisation starts too early.
Performance changes create pressure to act. CPA rises, ROAS falls, conversions slow down or spend suddenly accelerates — and the instinct is to change something immediately.
The problem is that the same headline result can be created by very different underlying causes.
CPC increased
Competition, auction mix or bidding behaviour may have changed.
Conversion rate fell
Traffic quality, landing pages, seasonality, offer strength or measurement may be involved.
Traffic mix changed
Different devices, locations, queries, audiences or campaign types may now represent more spend.
Tracking changed
Reported performance may deteriorate even when the underlying business outcome has not.
Outcome → Driver → Segment → Context → Evidence → Action
Outcome
State the change precisely: CPA +21%, ROAS -17%, conversions -14%.
Driver
Break the outcome into CPC, conversion rate, volume, value and mix.
Segment
Find where the change is concentrated: campaign, device, query, location, product or audience.
Context
Check seasonality, pricing, stock, website changes, tracking changes and market conditions.
Evidence
Identify which explanation is best supported and which remains uncertain.
Action
Choose the smallest action that directly addresses the supported diagnosis.
CPA rises 26%. What actually changed?
Account data
- Spend: +18%
- Clicks: +17%
- CPC: +1%
- Conversion rate: -20%
- Reported conversions: -5%
The most obvious numerical driver is conversion rate, not CPC.
But the diagnosis is still incomplete. You would now segment the conversion-rate decline by campaign, device, query, landing page and conversion action before deciding what to change.
Where to look before making a major change.
Diagnose before you optimise.
- Choose one recent account performance change.
- State the headline outcome without explaining it.
- Identify the main numerical driver.
- Segment the driver to find where it is concentrated.
- List three contextual explanations.
- Identify what evidence would confirm or reject each one.
- Choose the smallest justified action.
Diagnose the problem first.
1. CPA rises 25%. CPC is stable, but conversion rate falls 21%. What is the strongest first conclusion?
2. Conversion rate falls only on mobile after a website release. What is the most useful next step?
3. Why is segmentation useful during diagnosis?
4. What is the strongest principle in this lesson?
That discipline becomes essential as account complexity increases.