Decide what question the export must answer
Before opening Ads Manager, decide whether you are comparing clicks, lead submissions, qualified inquiries or purchases. Build a report around that objective. Combining incompatible outcomes under one “cost per result” column can make an analysis meaningless.
For local lead generation, a practical first set of fields may include ad name or ID, campaign, reporting dates, impressions, amount spent, link clicks, link CTR, recorded leads and cost per result. Select only the columns the account actually tracks.
Export ad-level data with enough context
In Meta Ads Manager, choose the Ad level, the intended date range and a consistent reporting preset, then export a CSV. Interface labels can change. Keep a note of which attribution settings and filters were used, especially when combining reports from different times.
If exporting data for several clients, keep accounts separate. Avoid including individual names, email addresses or phone numbers unless absolutely necessary for the workflow.
Match results to the exact asset version
A filename or ad name convention is a simple starting point. For example: roof-inspection-problem-v1 and roof-inspection-problem-v2. If the same name is reused for different creatives, any analysis that matches by name becomes unreliable.
Preserve a creative identifier where one exists, and manually review unmatched rows. A correct match matters more than a polished dashboard.
- Keep original versions instead of overwriting them.
- Record the date each version went live.
- Review duplicate names, missing fields and currency differences.
- Do not merge distinct objective types into one performance average.
Read the numbers in the correct order
Start with the primary outcome and its volume, then inspect costs. Use CTR and CPC to diagnose the earlier part of the journey; use qualified lead or purchase outcomes to decide whether the ad contributed to the business goal.
Segment by placement, audience or market only when there is enough information to support a sensible comparison. A tiny sample can make any group appear exceptional.
Create a learning record, not just a leaderboard
For every comparison, write down what was changed, what happened, which conditions may have influenced it and what should be tested next. That record makes it easier for a team to avoid repeating failed experiments or blindly copying a creative that performed well only in one context.
AdScoreLab supports CSV or manual performance entry alongside creative reviews. Treat those records as a practical aid to analysis, not a substitute for valid tracking and informed decision-making.
Common questions
Can a CSV identify the best creative automatically?
It can help rank observed metrics after the correct matching and filtering. It cannot establish that creative alone caused the outcome.
What if an ad cannot be matched to a creative?
Check naming and IDs, then resolve it manually. Do not silently assign results to the wrong asset.