A diagnostic for finding whether ownership, evidence, handoffs, measurement or correction—not strategy—is causing a marketing campaign to fail.
Short answer: answer: A sound strategy still fails when one of five execution controls breaks: ownership, evidence, handoffs, instrumentation or correction. Diagnose those controls in that order, and stop the campaign immediately if customer harm, unlawful processing or an unsupported material claim is present; otherwise give the team one bounded correction cycle before rewriting the strategy. Strategy should be reconsidered only when the intended audience, value proposition or economic premise remains unsupported after execution defects are removed.
The obvious explanation—“the idea did not resonate”—is attractive because it keeps the postmortem simple. It is often incomplete. An advertisement can express the planned idea accurately and still fail because nobody tested its meaning with the affected audience; a campaign can create extraordinary demand and still fail at the landing page; a dashboard can report its target while concealing the behaviour used to reach it.
CDM’s position is deliberately disputable: most teams abandon or defend strategy too early because strategy is more visible than the operating system underneath it. A creative rewrite cannot repair a missing consent boundary, a distorted review feed or a service that cannot complete the task promised in the launch message.
The Execution Fracture Atlas
The Execution Fracture Atlas separates five failure locations that are routinely collapsed into “bad strategy.” Use it as a sequence, not a score. A red flag in an early layer invalidates reassuring evidence from later layers.
Fracture 1: ownership
Ownership asks who can approve, stop and reverse the work. A named campaign manager is not enough. The decision owner must have authority over the consequence being created: privacy for data use, product reliability for a product launch, customer operations for fulfilment, and finance for an incentive structure.
The Wells Fargo board investigation filed with the SEC identified cultural, structural and leadership causes in the retail sales-practices failure. That record does not prove that every aggressive marketing target corrupts behaviour. It does show why target ownership without independent challenge is weak governance.
Fracture 2: evidence
Evidence asks whether the proposition is true in the conditions customers will encounter. It includes more than campaign research. Product claims need product evidence, privacy promises need verified data flows, and social proof needs a representative review process.
The FTC’s final BetterHelp order followed allegations that the service disclosed email addresses, IP addresses and health-questionnaire information for advertising despite privacy promises. The order required, among other measures, affirmative express consent for certain disclosures and a comprehensive privacy programme. The lesson is not that tracking technology is inherently improper. It is that copy and data architecture are two versions of the same promise.
Fracture 3: handoffs
Handoffs ask what changes when work crosses from strategy to creative, creative to legal, media to engineering, or platform data to reporting. A handoff fails when a necessary constraint becomes a comment, an assumption or somebody else’s problem.
Apple’s 2024 “Crush!” advertisement communicated product compression by showing creative objects crushed in a press. Apple’s marketing communications leader later said the company had missed the mark, and the planned television placement was withdrawn. The public record establishes the creative, reaction and apology; it does not reveal Apple’s internal approval chain. CDM’s inference is therefore limited: the released interpretation was not adequately challenged before distribution, whatever the internal process was.
Fracture 4: instrumentation
Instrumentation asks whether the team can observe the customer task and system condition that determine success. Reach and clicks are not enough when the promoted experience depends on account creation, inventory, support capacity or reliable product control.
Coinbase reported more than 20 million landing-page hits in one minute after its 2022 Super Bowl QR advertisement and said traffic had to be temporarily throttled. That is simultaneously evidence of exceptional demand creation and an execution gap. A campaign dashboard that celebrated scans without showing successful page loads and completed registrations would describe attention, not the whole outcome.
Fracture 5: correction
Correction asks whether the team can pause, contain, communicate, repair and learn before the failure compounds. Sonos launched a redesigned app on 7 May 2024. Its chief executive apologised on 25 July, listed missing or unreliable functions, and committed to updates roughly every two weeks. Sonos later disclosed delayed product launches and expected short-term recovery costs of up to $30 million.
The strategic ambition to modernise the platform was not, by itself, disproved. The launch system lacked a sufficiently protective correction path for an application customers needed to operate existing products. Correction capability belongs in the launch design, not in the apology draft.
The diagnostic and evidence classification table
Use this reader asset in the first postmortem meeting. Complete one row per material symptom and link every factual statement to a dated record.
| Atlas fracture | Question to answer | Minimum evidence | Stop or correction rule |
|---|---|---|---|
| Ownership | Who could stop this consequence before launch? | Named approver, authority and escalation route | Stop if no accountable owner can reverse the exposure |
| Evidence | What proves the promise under real conditions? | Primary record, test result, consent or substantiation file | Stop an unsupported material claim or unlawful use |
| Handoffs | Which constraint was lost between teams or tools? | Version history, approval record and final rendered asset | Correct before further distribution if the final differs materially |
| Instrumentation | Can we observe task completion and harm? | Baseline, success event, error rate and customer signal | Limit scale if failure cannot be detected quickly |
| Correction | How fast can we contain and repair it? | Pause control, incident owner, correction channel and review date | Pause when recovery time exceeds the declared tolerance |
Classify evidence as documented fact, CDM inference or CDM recommendation. A regulator’s order, a company release and a timestamped system record are documented facts. A conclusion about why an internal team behaved as it did is an inference unless the record supports it. A proposed threshold—such as requiring a rollback within 15 minutes for a high-volume landing page—is a recommendation, not a universal benchmark.
How to run the diagnosis in 45 minutes
Spend the first ten minutes writing the failed customer promise in one sentence. Do not start with channel metrics. Spend the next fifteen mapping the five fractures and marking facts, inferences and unknowns. Use ten minutes to select the smallest reversible correction, an owner and a deadline. Use the final ten to decide whether to continue, limit, pause or terminate distribution.
One correction cycle needs a predeclared test. For example: “Restore at least 99% successful form submissions for two consecutive peak-volume windows, with no duplicate CRM records, before reopening paid traffic.” The number must come from the team’s service need and risk tolerance; CDM is illustrating the form of a threshold, not prescribing 99% for every campaign.
Rewrite the strategy only after the corrected system produces credible evidence against its premise. If delivery works, measurement is valid and the intended audience still rejects the value proposition, strategy is now the leading diagnosis. Until then, changing the proposition may merely hide an operational defect.
Related guides
Frequently asked questions
How can I tell whether the strategy or execution caused the failure?
Test the strategy only after you can show that the intended experience was delivered and measured correctly. Confirm a named owner, valid supporting evidence, intact handoffs, observable customer completion and a functioning correction route. Then compare the corrected outcome with the pre-campaign baseline and decision threshold.
If the proposition still fails with the right audience under those conditions, the strategic premise is suspect. The caveat is that strategy and execution can fail together; repairing execution does not prove the positioning is sound, but it prevents a contaminated test from being treated as a strategic verdict.
What should a campaign postmortem examine first?
Examine the customer consequence first, then ownership. State what a customer saw, believed, disclosed, paid for or could not do, and identify who had authority to prevent or stop it. That order reduces the tendency to begin with impressions, sentiment or team intentions. After consequence and ownership, inspect evidence, handoffs, instrumentation and correction.
The exception is an active safety, legal or privacy incident: containment comes before diagnosis. Preserve records while pausing the risky activity, then conduct the structured review with legal, privacy or security specialists as appropriate.
Is poor campaign performance always an execution failure?
No. Weak demand, an undifferentiated offer or an uneconomic audience can be genuine strategy failures. The Execution Fracture Atlas is not a device for protecting a favourite strategy from evidence. It is a way to determine whether the evidence is trustworthy.
If the landing page worked, targeting matched the plan, exposure was sufficient, measurement was validated and the audience still did not respond, changing the proposition may be correct. A low-volume campaign is the caveat: insufficient data may support another bounded test, but it should not be dressed up as proof of either strategy or execution.
How many correction attempts should a team allow?
Allow one bounded correction cycle before escalation when the problem is reversible and no material harm is continuing. Define the change, success threshold, measurement window, owner and stop time in advance. A second cycle should require fresh evidence that the first test isolated the wrong cause, not merely optimism.
Do not allow a correction cycle for an unlawful data use, an unsafe claim or an expired right; stop the activity first. The exception is a complex infrastructure recovery where several technical stages are unavoidable, but management should still treat them as one controlled incident with explicit gates.
Can a campaign be successful and still expose an execution failure?
Yes. High reach, registrations or revenue can coexist with broken controls. Coinbase’s Super Bowl campaign produced reported traffic far beyond its previous benchmarks while also requiring temporary throttling. Wells Fargo’s cross-sell metric appeared to signal growth while the underlying incentives and controls contributed to misconduct. Judge success across customer outcome, operational integrity and economics—not the headline metric alone.
The caveat is proportionality: brief, planned throttling may be an acceptable protective control. It becomes a failure when the team did not define the trade-off, cannot measure lost completion or exposes customers to avoidable harm.
Who should own an execution failure that crosses several teams?
Assign one incident owner with authority to coordinate the response, while keeping specialist owners accountable for their controls. Marketing may own distribution; engineering may own availability; privacy may own lawful processing; customer operations may own remediation. A committee without a single decision owner slows containment and blurs responsibility.
Record who can pause the campaign, who approves resumption and what evidence each specialist must provide. The caveat is regulated or legally sensitive work, where counsel or a statutory role may control particular decisions even when another executive coordinates the overall incident.
Next decision: What Should a Marketing Team Do in the First Hour of a Public Campaign Failure?
Related reading: How Should You Analyze a Marketing Success Story Before Copying It? · What Should a Marketing AI Incident Runbook Include? · What Should a Marketing Team Do When a Critical Tool Fails on Launch Day?
Sources and research notes
- NIST Risk Management Framework — a useful primary model for assigned controls, authorization and continuous monitoring; checked 26 September 2026.
- Wells Fargo independent investigation, filed with the SEC — board-commissioned record of sales-practice causes and remediation; checked 26 September 2026.
- FTC final BetterHelp order announcement — regulator account of allegations and final obligations; checked 26 September 2026.
- Apple “Crush!” reporting and response — contemporaneous description of the creative and reaction; checked 26 September 2026.
- Coinbase campaign review — company-reported traffic, engagement and throttling; checked 26 September 2026.
- Sonos app update from Patrick Spence and Sonos fiscal 2024 filing — company chronology and disclosed consequences; checked 26 September 2026.
- Limitations: Public records show events and formal findings, not complete internal decision processes. The Atlas and thresholds are CDM analysis and recommendations, not findings made by the cited organisations or universal legal advice.
This article is editorial guidance. Apply the principles in proportion to your market, evidence, and responsibilities.
