What you will achieve
Create a scorecard that records the campaign problem, completed outputs, actual costs, quality findings and discovery observations. Finish with a recommendation to continue, change or pause a defined activity.
The scorecard is an editorial and operational working document. It is not a native AdCyte ROI dashboard or a method for assigning a monetary value to every citation.
Before you start
Choose a bounded campaign and evaluation period. Gather the agreed brief, task records, approved outputs, publication checks, actual charges and relevant analytics or AI observations. Identify the decision owner.
If no baseline exists, record that limitation. Do not create a retrospective estimate and present it as measured history.
1. State the decision
Write the decision in one sentence: “Should we repeat this workflow for the next two product campaigns?” Record the problem the pilot was intended to solve and the useful outcome expected.
For the worked example, the problem is an unclear product-page explanation of a removable cookware handle. The intended output is a reviewed explanation connected with the relevant demonstration and approved product guidance.
Check: the scorecard evaluates a decision rather than collecting every available number.
2. Define equivalent work
If measuring efficiency, describe the alternative process and the same finished deliverable. Include source review, drafting, approval, implementation and checking in both cases.
If the alternative was never measured, label any estimate clearly and keep it outside observed savings. The current pilot can establish a baseline for the next comparable assignment.
Check: a time comparison does not pit a generated draft against a fully reviewed publication.
3. Record the completed work
| Item | Evidence required | Status to record |
|---|---|---|
| Campaign evidence reviewed | Asset and relevant moments | Complete, partial or blocked |
| Product facts resolved | Current approved references | Accepted or unresolved |
| Draft reviewed | Exact accepted version | Approved or needs revision |
| Destination updated | Actual public result | Published and checked, or pending |
| Receiving team accepted handover | Named owner and decision | Accepted or outstanding |
Use actual records. A planned output is not a completed one, and an approved draft is not necessarily a published page.
Check: each completed item has inspectable evidence.
4. Log costs and effort
Record actual product charges and staff time by activity. Include briefing, source review, drafting, approval, implementation, rework and maintenance within the chosen period.
The following numbers are an illustrative calculation only: two hours of preparation, one hour of review and one hour of implementation total four hours. At an assumed allocated rate of £50 per hour, that is £200 of allocated labour. If actual service charges for the exercise were £30, the combined example cost would be £230.
Do not present those figures as AdCyte pricing, a typical campaign cost or a customer result. Replace every input with your own records and distinguish cash charges from allocated labour.
Check: the total can be recalculated from its stated inputs.
5. Record quality and rework
Note whether the output answered the question and passed the agreed standard. Separate substantive corrections, such as a wrong variant or unsupported claim, from ordinary style changes.
For the cookware example, a reviewer might remove an unsupported compatibility statement while accepting the explanation of the release action. Record the correction and whether the source brief could prevent it next time.
Check: a speed improvement is considered alongside the quality of the finished work.
6. Add discovery observations separately
Record relevant visits or AI-answer observations using the existing measurement method. For AI checks, retain the question set, provider, date, settings, eligible responses and failed checks.
An illustrative observation might be two citations in four eligible samples, with one additional failed check reported separately. This describes that sample only. It does not establish a market-wide citation rate or a causal effect of the page update.
Use no invented baseline. If the campaign has not been checked, record “not measured” and assign the observation task if it is relevant to the decision.
Check: observations remain distinct from work completed and from commercial outcomes.
7. Complete the decision record
| Field | Illustrative assessment |
|---|---|
| Original problem | Demonstrated handle feature poorly explained on product page |
| Useful outcome | One reviewed section and relevant source connection implemented |
| Cost basis | Actual task log and charges required; example arithmetic above is illustrative |
| Quality finding | Unsupported compatibility extension removed during review |
| Discovery finding | Insert actual observations or record not measured |
| Main uncertainty | Whether the workflow remains efficient across different campaigns |
| Recommendation | Repeat a bounded assignment with a clearer source brief |
| Next evidence | Comparable effort log, quality review and agreed discovery checks |
This example recommendation follows the operational learning, not an assumed revenue result. Your conclusion may reasonably differ if the actual work was too costly or the outputs did not resolve a useful question.
Check: the recommendation is traceable to the findings and names the next evidence needed.
8. Assign the next action and maintenance
Give each agreed action an owner and review point. Include the person responsible for maintaining the published explanation when product facts change.
If ROI cannot be calculated defensibly, leave it out. Report costs and useful outcomes directly. If your organisation has a valid incremental-profit estimate, apply its agreed ROI method with the period and assumptions stated.
Check: the scorecard produces a decision and owned work rather than another report awaiting interpretation.
Troubleshooting
| Problem | Response |
|---|---|
| No baseline | Establish the current workflow as a baseline; label estimates |
| Time logs omit review | Add review and rework before comparing efficiency |
| Approved copy was never published | Record the publication dependency separately |
| Failed checks inflate negative results | Separate failures from eligible answer observations |
| Revenue attribution is unavailable | Report costs, outputs and signals without invented ROI |
| Many outputs, little usefulness | Revisit customer questions and destination needs |
Completion checklist
- Decision, campaign and period defined.
- Comparison basis stated honestly.
- Completed outputs supported by records.
- Actual charges and human effort included.
- Quality and rework considered.
- Discovery and commercial observations separated.
- Assumptions distinguished from measured results.
- Recommendation, owners and maintenance agreed.
Evaluate a campaign with useful evidence
Explore AdCyte for marketing leaders to connect the work with a clear evaluation. To scope the next campaign, book a pilot discussion.
FAQs
Is the example cost an AdCyte price?
No. It is illustrative arithmetic. Use actual charges and your organisation's agreed treatment of staff time.
Should we combine everything into one score?
Only if the weighting has a clear decision purpose. Keeping quality, cost and observations visible often makes the trade-offs easier to understand.
Can a pilot be worth continuing without measured ROI?
Yes, if its useful outcomes and operating evidence justify another bounded test. State the remaining uncertainty rather than claiming a return that has not been established.

