Turning a Scorecard into One Priority Action

In this lesson

Reading a scorecard is half the job. The other half is turning it into an answer to "what do we do on Monday?" In this lesson we take one priority action from the Sample Coffee Co. scorecard in Lesson 1. Then, if you like, you do the same for your own brand.

The four parts of a good action

1. Evidence: which line of the scorecard? Question number, state, quote from the answer.
2. Diagnosis: why does that line look like that? Combine it with what you know.
3. Action: a concrete task you will do, in one sentence.
4. Check: how and when you'll look at the effect.

If one of the four is missing, the action is either a guess or a wish.

The three moves the scorecard suggests

The Lesson 1 scorecard suggests three moves: move brand evidence earlier, check the marketplace listing, and add JSON-LD. Each comes from evidence on the scorecard. But you don't have to do all three at once. Which comes first?

Worked answer (representative)

This is one defensible answer, not the only correct one.

Evidence. Sample Coffee Co. is absent from 2 of the 4 answered questions: question 3 (breakfast in Clifton) and question 4 (freshly roasted beans). Both were answered, so these are measured "Absent" results, not unmeasured ones.

Diagnosis. According to its facts, Sample Coffee Co. does both: it serves breakfast and sells beans. But the breakfast menu on the website is only a photo, not text. The answer to question 4 sent buyers to a marketplace (the quote names marketplace.example), where Sample Coffee Co. has no store. So two facts that answer real customer questions aren't written anywhere the AI can easily read them.

Action. Using Module 5's answer-first rules, add two Q&A sections to the website: "Do you serve breakfast?" and "Do you sell coffee beans?". Give the answer in the first sentence, then the hours, a few menu items, a price range and how to order. Update the fact sheet with the same facts.

Check. Don't check with a single query. With paid measurement, that means looking at comparable weeks with the same question and competitor sets (Module 6). If you use the free scorecard, the next one for the same business is available after 30 days. The questions are generated automatically and may not be the same, so compare the questions before comparing the two scorecards.

Why not the others?

• JSON-LD is Site Readiness work and adds no points to Discovery Visibility. It's useful, but structured data should mirror the page text. Writing the text first means the JSON-LD can carry the right facts. It's a good second step.
• The marketplace listing is channel work. But Sample Coffee Co. has no store there; opening one is a business decision, not a visibility fix.
• Moving brand evidence earlier rests on only 2 answers. And working on position in answers where you appear is a weaker place to start than answers where you're missing entirely.

Weak actions

• "Let's increase our SOV." There's no SOV on this scorecard, and that's a goal, not an action.
• "Let's add llms.txt." The scorecard doesn't count it as an issue, and no visibility effect has been shown.
• "Adding JSON-LD will raise our visibility." Site Readiness adds no points to visibility. Claiming an effect is a promise without measurement.
• "Let's beat Café A." Café A isn't a confirmed competitor yet, and this is a goal too.
• "We're missing from question 5, let's fix that." Question 5 wasn't measured. There's no result to fix.
• "Our site isn't being cited, let's raise owned-site citations." The free scorecard doesn't measure sources; "Not measured" isn't 0%. This scorecard supports no diagnosis based on a citation rate.

Optional: your own brand's scorecard

This step is optional and doesn't count toward your exam score. Your certificate doesn't depend on your own scorecard.

You can run a free scorecard for your own brand: free scan. The rules:

• It asks ChatGPT only, with 5 automatically generated questions that don't include your brand name.
• It can run once every 30 days for the same business. The Academy doesn't add extra runs.
• If you give your website, Site Readiness is measured as a separate diagnostic. If you don't, it shows as not measured.

If you can't get a scorecard, for example because one was already run for the same business in the last 30 days, carry on with the representative Sample Coffee Co. scorecard. The lesson and the exam are built around it.

If your scorecard comes back nearly empty (e.g. 0/4), that's a result, not an error. New or lesser-known brands may not appear in AI answers at all. Unmeasured questions, though, are never zero.

When you read your own scorecard, follow the Lesson 1 order: scope, headline and x/y, answer coverage, evidence table, other names (four boxes), source indicators (not measured on the free scorecard; don't read them as 0%), then Site Readiness on its own. Then write one action using the four parts above.

After the measurement

On paid plans the dashboard is grouped into three sections. MONITOR holds Tracked Questions and Competitor Radar; IMPROVE holds GEO Actions, Site Readiness and Brand Facts; PROVE holds Reports. Improvement work that comes from measurement evidence is listed in GEO Actions.

Summary

• A good action has four parts: evidence, diagnosis, action, check.
• A strong priority on the representative scorecard: write answer-first content on the two things Sample Coffee Co. really offers but hasn't written as text on its site (breakfast and coffee beans).
• Site Readiness work such as JSON-LD adds no points to visibility. Goals ("raise SOV") and results you can't measure aren't actions.
• Check the effect with comparable measurement, not a single query.
• Your own scorecard is optional and doesn't count toward your score; if you can't get one, continue with the representative example.