GEO-131·Issue 11·Measurement & Growth·131 / 132
A GEO Scorecard: Six Numbers for Management
Six numbers a GEO scorecard needs, how each defines its numerator and denominator, and a one-page review that ends in a decision, not a status update.
Issue focusScale It: SOPs, Organization, and the 2027 Roadmap
In the first week of November, a DTC air purifier brand carries its 2027 budget into a review. Slide 7 lists 23 metrics: mention rate, citation rate, sentiment, AI sessions, and nineteen more. The founder flips to the last page: if we double this line next year, which of these 23 moves first? The room goes quiet. The growth lead says they all matter.
Management does not need 23 visibility metrics. It needs six numbers that each trigger a decision — visibility, citation, recommendation, traffic, conversion, payback — where the first three decide whether content is still worth writing and the last three decide whether this money is worth spending again.
Six numbers, not twenty-three
A management team has three moves: add money, change direction, or stop. A number that cannot force one of those is not a scorecard row; it is an appendix. Longer lists make it easier to promote the softest metric to progress. Mentions rise on one misquoted spec in a review or a single forum thread, and neither changes an order.
The familiar counterexample: 23 metrics as red, amber, green, thresholds set in the meeting. Six months later every light means something and nobody knows whether to rewrite a page or move a price.
The six have an order: visibility, citation, and recommendation on the answer side; traffic, conversion, and payback on the business side. Break the first three and the last three will not hold; miss the last three and the first three were decoration.
Six numbers, six definitions
| Number | Numerator / denominator | Decision it triggers |
|---|---|---|
| Visibility | Answers naming you / valid samples | Keep funding presence |
| Citation | Answers quoting your page, with source / valid samples | Write more, or fix the page |
| Recommendation | Answers shortlisting you, with a use case and exclusion / valid samples | Where review budget goes |
| Traffic | Identifiable AI sessions / all sessions | Channel budget split |
| Conversion | Orders and margin from AI referrals / sessions from that entry | Fix price, shipping, returns |
| Payback | Spend / monthly incremental margin, last three months | Add, cut, or stop |
Three habits belong in the dictionary: counts beside ratios, per-engine reporting, and “unattributable” when a session cannot be identified. Default channel grouping in most analytics tools has no slot for AI referrals; a forced allocation keeps traffic undercounted.
Payback needs a warning. Incremental contribution margin is an estimate, not an accounting figure: it ranks projects, not board-pack ROI, and should not be promised externally. Pushing it toward evidence takes the counterfactual design in incrementality experiments.
One page: three columns and a change log
One page, six rows, three columns: baseline, last quarter, this month. One number and one arrow per cell; charts go in the appendix. Put the denominator beside every ratio, or nobody can tell whether 23% is nine answers out of forty or ninety-two out of four hundred.
A fourth column is fixed: what changed this month — pages rewritten, reviews commissioned, templates moved. Without it, every move gets blamed on the algorithm.
A line at the bottom names the quarter’s main front. The other five are guardrails with a red line, not a target; six simultaneous targets guarantee the team chases the softest layer.
Five steps, two rhythms
| Step | Action | Owner | Done when |
|---|---|---|---|
| 1 Settle the six | Keep six; rest to appendix | Growth | Each row maps to one decision |
| 2 Write the dictionary | Numerator, denominator, sampling, exclusions | Data | Two people get one figure |
| 3 Wire the data | Sampling, channel grouping, order tags | Data, engineering | Six refresh without a sheet |
| 4 Set thresholds | Red lines from last quarter | Growth | Threshold has a stated origin |
| 5 Set the cadence | Monthly numbers, quarterly money | Owner | Review arrives with a request |
Monthly takes 30 minutes: the first three numbers and the change log, no finance. Quarterly takes 45 minutes: add traffic, conversion, and payback, then end with one sentence — add, change direction, or stop. A review that cannot produce it should not be scheduled.
The rhythm has to match the build. The six numbers arrive in the order of the gates in the 90-day roadmap: visibility, then answer assets, then instrumentation. Installing the full scorecard in week one returns six zeros, and a zero from an unbuilt pipeline reads like a zero from a lost market.
Example: reading a fictional quarter
Example — brand, market, and figures are invented; only the reading method carries over.
| Number | Q3 | Q4 | Note |
|---|---|---|---|
| Visibility | 34% | 46% | Brand terms first |
| Citation | 11% | 23% | Eight new passages |
| Recommendation | 4% | 5% | Set by outside lists |
| AI-referred share of sessions | 1.6% | 2.2% | Small base |
| Conversion from AI referrals | 2.1% | 2.4% | Returns fell |
| Payback | 16 months | 11 months | Half from a second market |
Citation doubled while recommendation barely moved: the content is used as evidence but does not reach the shortlist. What is missing is third-party comparison and exclusion conditions — the work in third-party “best of” lists, not ten more articles. Traffic rose while conversion held, so exposure is the gap. Payback fell to 11 months, half from reusing the same assets in a second market rather than scale; say so in the deck, or next quarter it becomes a promise.
The decision request: add budget to comparison pages and independent reviews, not output volume. Volume lifts the first three numbers and leaves the last three where they are.
Where this breaks
Too few samples. With single-digit valid samples a week, ratios are noise. Record direction, keep verbatim answers, report counts.
In-platform assistants and agent traffic stay unidentified. Write “unattributable” and leave it; an allocated figure gets challenged next budget review and takes the scorecard with it.
Six numbers with no guardrails. Refund rate, ticket volume, and return reasons belong under conversion, or the page turns green as the business worsens.
Six department KPIs. One number per owner means nobody owns the handoff. The six are segments of one thing, and one person answers for it.
Volume up, QA flat. The first three numbers rise, the last three fall. That calls for the content QA checklist, not more publishing.
This month, cut the deck to six numbers. For each, write the numerator, the denominator, and the decision it triggers. Next review, bring one page and one closing sentence naming that decision. Management will not remember 23 numbers; it will remember whether last time’s decision was honored. The rest of this line sits in measurement and growth.