DEV Community

Cover image for Ad Measurement Cheat Sheet: 90% of "Bad Performance" Is a Data Problem, Not a Bidding Problem
xiaolinzi11111@163.com
xiaolinzi11111@163.com

Posted on Edited on

Ad Measurement Cheat Sheet: 90% of "Bad Performance" Is a Data Problem, Not a Bidding Problem

Ad Measurement Cheat Sheet: 90% of "Bad Performance" Is a Data Problem

If you've run paid ads, you've probably seen this: ROAS drops 30% overnight, and your first instinct is to swap creatives or tweak bids. After spinning your wheels, you discover the real issue wasn't the creative at all — the data foundation was broken from day one.

This is a one-page cheat sheet covering metric validation, funnel diagnosis, optimization priority, common pitfalls, and platform-specific landmines. Scan it before opening any dashboard.


1. Metric Validation (5-Minute Self-Check)

# Check Ask Yourself Consequence if Wrong
1 Attribution Window Same window across all accounts? Cross-account ROAS is fake (±30%)
2 Date Basis Click date vs conversion date? Media and finance reports diverge
3 Currency/Exchange Unified exchange rate? Spend inflated/deflated 8%+
4 Actual vs Platform Spend Platform Cost = credit card charge? CPA built on sand
5 Additivity Reach/Frequency being SUM'd? Reach inflated 218%
6 Ratio Math CTR = total clicks ÷ total impr? Small-volume skews average

Quick call: Mixed attribution windows → separate accounts, never SUM. Use click-date internally, conversion-date for finance. Non-additive metrics → MAX/AVG, never SUM.


2. Funnel Diagnosis (Layer by Layer)

Layer Core Metrics Drop-off → Diagnosis
Impression Impr / Reach / Freq No impr = low budget/bid; Freq>3 = fatigue
Click Clicks / CTR / CPC Impr but no click = creative/targeting
Landing LP View / Bounce Click but no pageview = speed/tracking loss
Conversion Conv / CVR / CPA Pageview but no conv = offer/trust/friction
Value AOV / ROAS / LTV Conv but unprofitable = AOV/backfill/attribution

Rule: Cut where the drop is steepest.


3. Optimization Priority

Order: Cut waste → Improve efficiency → Scale

Phase Action What to Do
Cut Pause zero-conv Export CSV → sort by CPA → pause bottom 20%
Cut Add negatives Search terms → high spend, zero conv → exact negative
Cut Exclude bad placements Placement report → spend, no conv → exclude
Improve Swap creative Freq>3 + CTR dropping 3 days straight → swap
Improve Landing page High bounce → speed/mobile/friction audit
Improve Bid strategy 50+ conv/week → tCPA/tROAS; else → Manual
Scale Raise budget +20%~30% each time, watch 3-5 days
Scale Expand audience 1% LAL proven → expand to 3%-5%
Scale Add variants 3-5 variations of winning creative

4. Six Common Pitfalls

Pitfall One-Liner How to Catch
Cross-account sum Different windows can't be added List attribution windows, flag mixing
Date mismatch Click date ≠ conversion date Two books, never compare directly
Fake spend Platform Cost ≠ actual charge Credit card statement vs platform reconciliation
Currency hard-sum EUR treated as USD in SUM Convert at unified rate first
Inflated Reach Daily Reach summed across days Use deduplicated value, never SUM
Fake ratios CTR/ROAS naively averaged Total ÷ total, never average

5. Platform-Specific Landmines

Platform Watch Out For
Google Ads Display network bleeds to junk sites; Smart Bidding fails with thin data
Meta iOS 14+ attribution window shrunk; Advantage+ silently reshapes audience
TikTok Report delay 24-48h; limited attribution window options
Ocean Engine Deep-conversion backfill unstable; cold start slow
Tencent Ads Backfill delay; closed-loop inside WeChat but reports fragmented

6. Bottom Line

90% of "bad performance" is a data-layer disease, not a bidding problem.
Validate metrics → Diagnose funnel → Attribute to variables → Cut waste → Improve efficiency → Scale.


About the Author

Building Locply — a local CSV ad metric audit tool. No API, data never leaves your computer. Import Google Ads Ads CSVs, and it auto-flags attribution window conflicts, date-basis mismatches, currency hard-sums, non-additive metric SUMs, and more.

Top comments (0)