Evidence command board

Jul 1–28, 2026

Why did CPA rise 14.8% after Jul 15?

TL;DR

CPA increased while conversion volume fell, but qualified lead rate improved. Hold settings for 7 more days while validating device-level CVR.

Evidence timeline

Jul 1–28, 2026 (America/Toronto)

Conversions / day CPA (CAD)

All times shown in America/Toronto (ET)

Sample case loaded. Select any evidence source to upload your own CSV.

Paid media diagnostic guide

Campaign performance changed. Find out what changed with it.

Campaign forensics is the process of reconstructing a paid media performance shift from multiple evidence sources. Instead of assuming the most recent edit caused the result, it compares timing, platform delivery, traffic behaviour, measurement changes, lead quality, revenue, profit, and controlled-test evidence.

Short answer

When CPA rises or conversions fall, check whether spend, traffic mix, conversion rate, tracking, creative exposure, lead quality, or unit economics changed in the same window. Rank the explanations, identify contradictory evidence, and run the smallest test that can separate the leading causes.

Diagnostic coverage

What the Campaign Forensics Lab can diagnose

CPA increased

Separates higher auction cost, weaker conversion rate, traffic-quality changes, tracking shifts, and low sample volume.

Conversions dropped

Checks delivery, click volume, conversion lag, form completion, landing-page behaviour, and downstream lead quality.

Meta creative fatigue

Requires a pattern of repeat exposure, worsening response, rising CPM, and concentrated creative delivery—not frequency alone.

LinkedIn lead weakness

Shows observed economics and lead-form completion changes while leaving the root cause unresolved without audience and CRM evidence.

Tracking or attribution changed

Flags measurement contamination when conversion actions, tags, GA4 events, or attribution definitions change near the performance shift.

ROAS hides weak profit

Connects platform revenue to COGS, shipping, fees, discounts, refunds, and ad spend before calling growth profitable.

Evidence requirements

What data should you use to investigate a campaign change?

No single export is a complete source of truth. Platform files describe delivery and attributed outcomes; analytics, CRM, finance, and experiments test whether those outcomes were real and valuable.

Evidence sourceQuestions it can answerWhat it cannot prove alone
Google, Meta, or LinkedIn performanceDid spend, reach, clicks, conversions, CPA, CPM, frequency, or attributed value change?Whether the platform caused incremental business results.
Campaign change logWhich bids, budgets, audiences, creatives, offers, pages, or tracking settings changed near the inflection?Which overlapping change caused the movement.
GA4 landing-page dataDid sessions, key events, device mix, or landing-page conversion behaviour change?Lead quality, closed revenue, or profit.
CRM outcomesDid raw leads become qualified leads, sales opportunities, or revenue?Whether unattributed demand would have happened without ads.
Finance dataDid attributed revenue clear product, fulfilment, fee, refund, discount, and media costs?Incremental lift without a credible counterfactual.
Experiment or holdoutHow did treatment perform against a concurrent control?A trustworthy causal estimate when assignment, power, or measurement is weak.

Channel mapping matters: when multiple platforms are active, GA4, CRM, finance, and experiment rows must identify the relevant channel. Unmapped outcomes remain available but are not borrowed into a channel-specific conclusion.

How to use the tool

A four-step campaign forensics workflow

  1. 1

    Choose the investigation scope

    Start with Google Ads, Meta Ads, or LinkedIn Ads. Use the blended view only to find portfolio-level movement, then return to each channel before taking action.

  2. 2

    Add the evidence you actually have

    Upload CSV exports for performance, changes, GA4, CRM, finance, or experiments. Missing evidence lowers confidence instead of blocking the analysis.

  3. 3

    Review competing explanations

    Read what supports, contradicts, or is missing from the leading hypothesis. Check the runner-up and any confidence cap before accepting the TL;DR.

  4. 4

    Challenge, test, and document

    Challenge the logic, run the sensitivity test, and create a decision brief that records the evidence, falsifiers, unresolved question, and next test.

Campaign performance diagnostic tree showing how to locate a change in delivery, response, or conversion before checking account edits, GA4, CRM, finance, falsifiers, and the next test.
Campaign performance diagnostic tree: find the first broken stage, align the change timeline, reconcile platform and business evidence, then challenge the leading explanation.
Download the campaign diagnostic tree (SVG)

Transparent methodology

A diagnostic brain you can inspect

The Campaign Forensics Lab is deterministic. Channel adapters normalize Google, Meta, and LinkedIn evidence without applying one platform’s rules to another. The shared brain checks timing, measurement integrity, segment concentration, business outcomes, unit economics, experiments, and competing explanations. Every score and confidence cap is traceable; uploaded data never leaves the browser.

SupportsEvidence that raises a hypothesis score.
ContradictsEvidence that weakens the current explanation.
MissingInputs needed before confidence should rise.

Confidence is evidence coverage—not causal probability. When changes overlap, measurement changes are unresolved, the sample is small, or the top explanations are too close, the brain limits confidence and recommends the smallest useful test.

Channel context

How the diagnostic logic changes by ad platform

Google Ads performance changes

Google-specific checks include query exclusions, search intent, bid-strategy changes, Target CPA or Target ROAS constraints, impression-share pressure, conversion actions, and landing-page conversion rate. A query cleanup can reduce raw conversions while improving lead quality, so CRM evidence can materially change the decision.

Meta Ads creative and auction changes

Meta-specific checks require more than a high frequency number. The fatigue hypothesis strengthens when repeat exposure rises while CTR falls, CPM rises, and one creative receives most delivery. Audience edits, offer changes, and tracking changes remain competing explanations.

LinkedIn Ads lead-generation changes

LinkedIn often has smaller samples and longer B2B outcome delays. The tool can identify rising click cost, falling form completion, or zero-conversion periods, but it will not invent an audience-quality explanation without segment and CRM evidence.

Frequently asked questions

Campaign forensics FAQ

What is campaign forensics?

Campaign forensics is a structured investigation of why paid media performance changed. It aligns dated performance, account edits, site behaviour, downstream outcomes, business economics, and experiments to rank competing explanations without claiming more certainty than the evidence supports.

Why did CPA increase after a campaign change?

CPA can rise because media cost increased, conversion rate fell, traffic mix changed, delivery shifted, tracking definitions changed, or normal sample noise affected the comparison. A nearby edit is a timing clue, not proof that the edit caused the increase.

How do you diagnose Meta Ads creative fatigue?

Look for repeat exposure together with worsening response: rising frequency, falling CTR, rising CPM, and delivery concentrated in a small number of creatives. High frequency by itself does not prove fatigue.

What should I check when LinkedIn Ads leads drop?

Check spend, impressions, click cost, click-through rate, lead-form opens, form completions, audience or form changes, conversion lag, and CRM-qualified outcomes. Low volume should remain directional until more evidence is available.

Can before-and-after campaign analysis prove causality?

Usually not. Before-and-after analysis is useful for diagnosis and prioritization, but overlapping changes, seasonality, demand, competition, and measurement shifts weaken causal claims. A well-designed concurrent control is stronger evidence.

Does the Campaign Forensics Lab upload my advertising data?

No. CSV parsing and analysis run in the browser. The current tool does not send uploaded campaign, analytics, CRM, finance, or experiment files to a server.

Continue the investigation

Sensitivity analysis

Stress-test the leading hypothesis

Scores are evidence weights, not causal probabilities.

Decision record

Campaign forensics brief