Read CDM's preregistered protocol for measuring creator production delays by stage, cause, waiting time and rework without claiming results early.
Short answer: answer: No defensible CDM benchmark exists yet because CDM has not collected a multi-company dataset. This article preregisters the Production Delay Decomposition: measure active work, waiting and rework for every stage and cause, then report distributions only after consent, verification and methods review. Until the publication thresholds below are met, any delay number would be an unsupported claim.
“Campaigns take six weeks” is not a useful operations benchmark. It mixes contracting, creator work, brand queues, legal review and release scheduling, then encourages teams to blame whichever group they already distrust.
CDM’s disputable position is that approval delay must be measured by stage, cause and rework loop—not averaged into one campaign duration. This is a publication-ready protocol and data dictionary, not a completed study. The illustrative record is synthetic. Results must replace this protocol only after ethical collection, analysis and independent review.
Preregister the Production Delay Decomposition
The analytical spine treats each deliverable version as a sequence of state intervals. Every interval contains one state, one owner class and one primary cause. Calendar duration is then decomposed as:
elapsed time = active work + internal waiting + external waiting + scheduled hold + unresolved time.
Rework is counted separately as a return to a completed stage after material feedback. This prevents a two-hour edit after a four-day queue from being labelled “four days of creator production.” The protocol will be time-stamped before recruitment through a public registration service such as OSF; deviations will be dated and reported, never silently rewritten.
Research questions and units
The primary question is: what share of elapsed production time sits in each stage and cause? Secondary questions test which stages generate rework, how waiting differs by deliverable type, and whether delay patterns differ descriptively by team size, review route or rights complexity.
The primary unit is a deliverable-version stage interval, not a person or whole campaign. A campaign may contain several creators and deliverables; a deliverable may have several versions. Organization and campaign identifiers preserve nesting. The study will not infer individual productivity from response times because a timestamp cannot reveal workload, working hours or contractual priority.
The complete event-log instrument
Participating teams will record consecutive eligible deliverables in a supplied form or structured export. Required fields are:
| Group | Fields |
|---|---|
| study keys | organization pseudonym, campaign ID, deliverable ID, version ID, event ID |
| context | team-size band, market region, content format, paid/organic use, creator-count band, rights-complexity band |
| state | brief accepted, contracting, production, internal review, creator revision, specialist review, platform scheduling, released or cancelled |
| timing | entered-at UTC, exited-at UTC, participant working-time zone, active minutes, scheduled hold minutes |
| ownership | brand, agency, creator, legal/compliance, client, platform or other dependency |
| cause | work in progress, queue/capacity, missing input, scope change, evidence check, rights issue, technical failure, calendar hold or unknown |
| rework | prior stage returned to, feedback source, material/non-material, avoidability coded later |
| quality | source system, recorder role, contemporaneous/retrospective flag, missingness reason, verification status |
No names, email addresses, post copy, contract text, health information or free-text accusations are requested. A short controlled note may explain an “other” code after redaction. Entry and exit timestamps must come from an approved project system or a contemporaneous diary; reconstructed times are flagged and analyzed separately.
Sampling, consent and collection rules
Recruit organizations that run paid or contracted creator work and can provide an authorized study contact. Seek variation in geography, team size, agency use and campaign type; publish the recruitment channels and response flow. Within each enrolled organization, include every new eligible deliverable entering production during an eight-week window. Consecutive inclusion reduces cherry-picking of only smooth or painful campaigns.
The preregistered target is at least 30 organizations and 300 deliverables. Those are design targets, not achieved sample sizes. Publication as a market benchmark requires at least 10 independent organizations, 100 complete deliverables and no organization contributing more than 25% of analyzed deliverables. If those minima are not reached, CDM may publish a methods report labelled pilot, but not benchmark values.
Organization consent does not erase worker interests. Give individual recorders clear information about purpose, fields, risks, retention, withdrawal and publication. Do not use participation for performance management. Obtain ethics or institutional review where applicable, especially if private activity data concerns identifiable workers. The Belmont principles—respect, beneficence and justice—inform consent and recruitment, but local requirements still govern.
Verification and quality control
Run four checks before analysis:
- Sequence: exit cannot precede entry; stages cannot overlap without a documented parallel-work code.
- Control totals: stage intervals must reconcile to deliverable elapsed time within a declared tolerance.
- Provenance: label system-derived, contemporaneous diary and retrospective records.
- Outliers: query rather than automatically delete very long or zero-length intervals.
A second analyst will audit a random 10% of deliverables against redacted source evidence supplied by the organization. Any verification rate, discrepancy rate or exclusion count will be reported after collection; none exists now. The de-identification key stays with a restricted data custodian and is not included in the analysis file.
Prespecified analysis plan
Report median, interquartile range and selected percentiles for elapsed, active, waiting and scheduled-hold time by stage. Report shares of total elapsed time and rework frequency, always with denominators. Weight organizations equally in the primary organization-level summary so one prolific program does not define the market. A deliverable-level analysis will be clearly secondary.
Use cluster bootstrap intervals by organization if the final sample supports them. Stratify descriptively by format, rights complexity, team-size band and specialist-review route only when cells meet disclosure thresholds. Do not rank named organizations, convert associations into causes or publish a single “industry average.” Missing intervals remain a reported category; complete-case and sensitivity analyses will show how assumptions change results.
The protocol’s falsifiable expectation is that waiting will account for more variation than active creator production. That is a hypothesis, not a finding. The analysis must report contrary results with equal prominence.
The reader asset: data dictionary and benchmark release pack
The downloadable release pack should contain the frozen protocol, consent materials, instrument, codebook, validation rules, analysis code, synthetic demonstration data, data-flow diagram and a disclosure-controlled aggregated dataset. Raw logs and organization keys should not be public.
The report must label every table with observation period, participating-organization count, deliverable count, completeness and denominator. A revision history should distinguish protocol changes made before data lock from exploratory analyses added later. STROBE is a reporting aid, not proof of study quality; CDM will use its observational-reporting checklist to make selection, variables, bias, missing data and limitations visible.
Illustrative record—not a finding
Organization ORG-S07 records deliverable D014, version V2. It enters “internal review” at 14:00 UTC Monday and exits at 10:00 UTC Wednesday. Active review is 35 minutes; a 16-hour scheduled overnight/non-working period is coded; the remainder is internal queue time. Feedback sends the deliverable to “creator revision” because the usage-rights channel was missing from the brief.
This invented row demonstrates decomposition. It does not show that reviews normally take 44 hours, that rights cause delays, or that any real organization behaves this way. Synthetic examples will be marked and excluded from analysis.
Stop, review and publication criteria
Pause collection for an unexpected privacy risk, systematic coercion, exposed identity key or instrument defect that changes measurement. Log the decision and seek methods/privacy review. Stop recruitment at the registered window or target, whichever rule the final registration specifies; do not extend merely to obtain a preferred answer.
Publish benchmark results only after data lock, verification, disclosure-risk review, reproducible analysis and an independent research-methods review. Publish null, mixed and inconvenient results. If thresholds fail, publish the protocol, recruitment outcome and limitations without headline estimates. CDM must never backfill this article with attractive numbers from an informal poll.
Related guides
Frequently asked questions
What is the current creator campaign delay benchmark?
There is no CDM benchmark yet. No dataset, participating-company count or verified delay distribution has been supplied, so quoting a number would manufacture evidence. This protocol defines what CDM will collect and the thresholds required before reporting results. Until then, teams should baseline their own stages using the same definitions.
The caveat is that other vendors may publish surveys, but their definitions, sampling and commercial incentives may differ. Compare instruments and denominators before importing any external number into planning, and label it as external rather than a CDM result.
Why not measure total campaign duration?
Total duration hides the operational cause. A long campaign might include a planned embargo, contract negotiation, brand queue, creator revision or specialist evidence review; combining them makes the number difficult to act on. Decomposition assigns time to states and distinguishes active work from waiting.
It also shows whether an intervention moved delay rather than removed it. The caveat is that stage data adds recording burden and can still misclassify parallel work. Keep the taxonomy short, verify sequences and report unknown time instead of forcing certainty.
How will participating companies stay anonymous?
CDM should pseudonymize organizations, separate identity keys, remove direct identifiers, band rare attributes and suppress small cells before release. Public files should contain aggregated or disclosure-controlled records, not raw messages, contract copy or names. Access and retention must be documented, with deletion responsibility assigned before collection begins.
The caveat is that pseudonymized information may remain personal or re-identifiable when combined with public campaign details. A privacy review and motivated-intruder test should decide what can safely be shared; “anonymous” must not be promised casually.
Can retrospective project-management data be included?
Yes, but it must be labelled and analyzed separately from prospective records. System timestamps can be useful, while reconstructed active minutes and reasons may suffer recall or workflow artefacts. The protocol records source and contemporaneous status for every interval, then runs sensitivity analysis excluding weaker records. Results should show how estimates change when retrospective rows are removed.
The caveat is comparability: tools define statuses differently and teams may leave cards idle. Mapping rules and evidence audits are required before combining exports across organizations.
Will the study prove what causes production delays?
No. It is an observational operations study designed to describe where recorded time accumulates and which coded conditions co-occur with it. It cannot randomly assign approvals, staffing or rights complexity, so associations should not be described as causal. Unmeasured workload or campaign difficulty may explain an apparent relationship.
The caveat is that repeated patterns can still guide hypotheses and local experiments. Teams can use the taxonomy to test a specific intervention, but those intervention results need a separate design and should not be folded into the benchmark as proof.
When will CDM publish actual benchmark results?
Only after ethical collection, the registered minimum sample, data lock, verification, disclosure-risk assessment, reproducible analysis and independent methods review. CDM should publish a dated status if recruitment closes without sufficient diversity, but it must not turn a small convenience sample into an industry benchmark. The frozen protocol and analysis code should accompany any release.
The caveat is that meeting a numeric threshold does not guarantee representativeness. The final report must still disclose recruitment channels, response, geography, missingness and concentration so readers can judge transferability.
Next decision: A Realistic Creator Campaign Timeline and RACI for Lean Marketing Teams
Related reading: The Creator Campaign Operating Playbook: Strategy to Post-Campaign Learning · How to Design a Creator Content Approval Process That Does Not Kill the Idea · Original Research as Authority Infrastructure: From Question to Citable Asset
Sources and research notes
- Center for Open Science: Register Your Research — checked 26 September 2026; supports time-stamped registration and linked research records.
- AAPOR Transparency Initiative — checked 26 September 2026; disclosure standards for survey and public-opinion research, adapted here for transparent recruitment and methods reporting.
- HHS Office for Human Research Protections: Belmont Report — checked 26 September 2026; primary ethical principles for research participation and selection.
- EQUATOR Network: STROBE Statement — checked 26 September 2026; reporting checklist for observational research; it is not a study-design or quality-scoring tool.
- ICO Anonymisation Guidance — checked 26 September 2026; regulator guidance on identifiability, pseudonymisation, governance and disclosure.
Limitations: No CDM observations, organizations, deliverables or results are claimed. Proposed targets are protocol decisions, not attained samples, and may require ethics, legal or privacy revision before recruitment. Self-selection, inconsistent workflow systems, recall and sector mix will constrain any future estimates. Publish findings only after ethical collection and review, and retain this version as the preregistered methods record.
This article is editorial guidance. Apply the principles in proportion to your market, evidence, and responsibilities.
