Marketing Forecast Governance for SaaS companies: Audit Checklist

SaaS marketing forecasts are often discussed as if they were a single number. In practice, a forecast can combine campaign delivery, lead creation, sales acceptance, conversion assumptions, renewal context, and a management decision. When those layers are not separated, a neat spreadsheet can hide missing inputs, changing definitions, or a confidence level that no one has agreed to use.

This audit checklist treats forecast governance as a quality system. It asks whether the forecast has a clear purpose, stable definitions, traceable inputs, visible assumptions, named ownership, and a correction path. It does not produce a budget, a revenue projection, an investment view, or a universal SaaS benchmark.

1. Identify the forecast’s decision

Write the decision before reviewing the model. Examples include choosing which campaign to monitor, deciding whether an experiment has enough evidence to continue, planning a staffing conversation, or preparing a scenario for a board discussion. These are different uses and should not be collapsed into “the forecast.”

Record the audience, date prepared, horizon, update cadence, and decision deadline. A model used for a weekly learning review can tolerate different granularity from a scenario used in a formal planning process. If the audience and decision are missing, mark the audit as incomplete rather than judging the number’s appearance.

2. Audit definitions and boundaries

Check each material term: inquiry, marketing-qualified record, accepted opportunity, pipeline, booked revenue, expansion, churn, and campaign contribution. The label should have an inclusion rule, exclusion rule, owner, and source. “Qualified” without a rule is a discussion prompt, not a forecast input.

Separate observed values, assumptions, and scenarios. An observed value has a source and period. An assumption is a chosen rule that may be challenged. A scenario is a conditional illustration. A forecast governance audit should make these categories visible in the file or record.

3. Test the input chain

Trace each important input to its origin and last refresh. Capture the query, export, CRM view, spreadsheet tab, or approved source that produced it. Check whether the period, timezone, currency, account population, and status definitions match the stated horizon.

If analytics events are used, retain the distinction between an observed interaction and a business state. The GA4 Event documentation describes how events are represented; it does not define a qualified opportunity, contract, or revenue result. An audit should flag a model that silently treats an event count as a commercial outcome.

Use an input record:

| Input | Type | Period | Source owner | Freshness | Reconciliation check | Status | |—|—|—|—|—|—|—| | Campaign response | Observed | Named dates | Marketing ops | Date/time | Duplicate and exclusion check | Pass / fail | | Sales acceptance | Observed or assumption | Named dates | Revenue ops | Date/time | Stage dictionary check | Pass / fail | | Conversion rule | Assumption | Version | Model owner | Review date | Sensitivity note | Pass / fail | | Scenario output | Conditional | Horizon | Planning owner | Version | Formula replay | Pass / fail |

4. Review measurement transport

Some SaaS teams import conversion information between ad platforms, analytics systems, and a CRM. The Google Ads conversion import guidance is an implementation reference for one route. It is not evidence of incrementality, quality, or revenue. The audit should record the source event, transformation, destination field, timing, and reconciliation owner.

Look for silent transformations: a renamed status, a changed attribution window, a missing currency, or an import that counts the same record twice. If the transformation cannot be replayed with a synthetic record, mark it as a major control gap.

5. Audit assumptions and scenario logic

List every assumption that can materially move the output. Include the reason chosen, evidence available, owner, review date, and what would cause a revision. Avoid false precision. A scenario with a range and a clear limitation is easier to govern than a single number with an unexplained decimal.

Run a simple sensitivity review: change one assumption at a time, record the direction of movement, and identify whether the decision changes. The purpose is not to prove an outcome. It is to show which unknown deserves the next evidence request.

The GOV.UK Measuring Success guidance is a useful reminder to link measures to decisions. It is not a SaaS planning formula. Use it to ask whether each tracked value changes an action, not whether the dashboard has more measures.

6. Check governance and decision rights

Name four roles where they apply: input owner, model owner, decision owner, and reviewer. A finance or revenue leader may review a scenario without owning the marketing input. A marketing operations owner may maintain the data mapping without approving a strategic commitment.

Record how disagreements are resolved. If marketing and sales use different definitions, the audit should show the difference and the decision to reconcile, retain both, or hold the output. Do not average incompatible stage definitions into a number that appears neutral.

7. Test correction history and quality

A governed forecast can answer: what changed, who changed it, why, which inputs were affected, and whether an earlier decision should be revisited. NIST’s Information Quality Standards provide a quality lens for utility, integrity, objectivity, and correction history. They do not validate a forecast.

Audit the correction log for stale assumptions, unexplained overwrites, missing source dates, and changes made after a review without notification. A corrected input should remain traceable. Deleting the old value can make a later variance impossible to understand.

8. Apply pass/fail criteria

Mark a control PASS only when evidence is attached or linked. Use FAIL when the control is not satisfied, and N/A only when the owner documents why it does not apply.

• The forecast purpose, audience, horizon, and decision deadline are recorded.
• Terms and stages have inclusion, exclusion, source, and owner fields.
• Inputs can be traced to a dated source and replayed or reconciled.
• Assumptions are separated from observations and conditional scenarios.
• Material assumptions have a review date and a change trigger.
• Measurement transports and transformations are documented.
• Duplicate, late, missing, and contradictory records have a treatment.
• The model has a correction history and version owner.
• A reviewer can state what the output does not prove.
• The action priority and next evidence request are explicit.

9. Use severity and action priority consistently

Severity describes the control failure; action priority describes when the team should address it.

| Severity | Example | Default action priority | |—|—|—| | Critical | Output presented as a commitment while material inputs are untraceable | Immediate hold | | Major | Stage definitions conflict, duplicate records are possible, or assumptions have no owner | Repair before next decision | | Moderate | Source date, sensitivity note, or correction link is missing | Fix in current review cycle | | Minor | Label, formatting, or non-material documentation gap | Track and assign |

Do not use a low action priority to make a critical evidence gap disappear. Conversely, a minor formatting issue should not block a clearly bounded learning discussion when the limitation is written.

10. Review a hypothetical SaaS audit

Imagine a growth team presents a scenario using campaign responses, sales-accepted records, and an assumed conversion rate. The audit finds that campaign responses are dated by platform timezone, sales acceptance uses a newer stage name, and the conversion assumption was copied from a different segment. The formula itself is reproducible, but the input chain is not aligned.

The proper result is not “the forecast is wrong” or “the forecast is approved.” The result is a major finding: reconcile the period and stage dictionary, replace the borrowed assumption with a documented range, assign owners, and rerun the sensitivity view. Until then, the scenario can be labelled conditional and used only for the bounded decision it was designed to inform.

11. Copy-ready audit record

“text Forecast name, version, owner, and audience: Decision supported and deadline: Horizon, cadence, timezone, currency, and population: Definitions and stage dictionary: Observed inputs with sources and refresh dates: Assumptions, rationale, sensitivity, and review triggers: Transformations, imports, and reconciliation evidence: Correction history and version changes: Controls: PASS / FAIL / N/A with evidence: Findings by severity: Action priority, owner, and due date: What this forecast does not prove: Audit decision: pass / repair / hold / retire: “

12. Close the audit with a decision

The audit is complete when a reviewer can see the forecast’s purpose, definitions, evidence, assumptions, ownership, limitations, and next action without reconstructing them from conversations. A clean result does not promise accuracy forever. It means the current version is governable and has a known path for correction.

Sources and limits

This checklist uses GA4 Events, Google Ads conversion import guidance, GOV.UK Measuring Success, NIST Information Quality Standards, and FTC Advertising and Marketing as governance references. It does not provide a financial forecast, budget recommendation, investment advice, revenue guarantee, or universal SaaS benchmark.