Start with the market, not keywords
Turn customer evidence into a prioritized map of the problems, comparisons, and questions that create pipeline. This is keyword research, done from the market inward.
Before you touch a keyword tool, find out what your buyers actually ask on the way to a deal, using your own calls, tickets, and wins. Then pick the ten questions worth winning first.
You will leave with one demand map that tells SEO, content, product marketing, and AEO what to build first.
Keyword research usually starts in a keyword tool, and keyword tools show what people type. They do not tell you which pains create urgency, which objections stall deals, or how buyers ask AI for a shortlist. Start with evidence from the market, then translate it into query families and prompts.
IF YOU SKIP IT / Skip this and every later module runs perfectly against questions that never touched revenue. This is the module that keeps the other nine honest.
- Keyword research
- Finding the queries and prompts buyers use for a decision. This module does it from the market inward: buyer language first, query families second, volume last as a tiebreaker.
- ICP
- Ideal customer profile: the type of company and person your product fits best. In this course you narrow it to one segment at a time.
- Closed-won / closed-lost
- CRM labels for deals you won and deals you lost. Both are evidence; lost deals often explain your objections.
- Demand family
- A group of related buyer questions that all point at the same underlying need, such as everything about automating client reports.
- Buying trigger
- The event that turns a nagging problem into an active search: a board meeting, a hire, a tool being cancelled, a busy season.
- Search volume
- How many times per month a keyword is typed into Google. Useful context, terrible master. This module deliberately demotes it.
Each phase feeds the next. Don't skip ahead, the artifact at the end is only trustworthy if every phase ran.
- 01Collect
Export 12 months of closed-won, closed-lost, and stalled opportunities with segment, role, ACV, and loss reason.
- 02Listen
Review at least 10 calls and tag exact problem, trigger, objection, comparison, and proof language.
- 03Expand
Add GSC queries, paid-search terms, competitor pages, review sites, communities, and AI prompt variants.
- 04Cluster
Group language by buyer job and decision, not by matching words.
- 05Score
Score revenue evidence, frequency, strategic fit, current visibility, and production effort separately.
- 06Map
Assign one primary asset and one owner to every priority demand family.
- 07Validate
Have sales and customer success challenge the map before content production begins.
- ✓You now have
Buyer demand map
Choose one revenue segment
A broad ICP produces a broad content plan. Pick one segment where the problem, buying process, and product fit are similar enough to write a precise answer.
Open your CRM and pull up the last twenty deals you won. Somewhere in that list a pattern is hiding: the same kind of company, the same buyer role, the same trigger, over and over. Most teams have never looked. They open a keyword tool on day one instead, it hands back 4,000 keywords, they pick the big numbers, and three months later there is traffic and no pipeline.
We are going to start where the money already is. Your closed-won deals are a list of people whose problem was urgent enough to pay for. Group them, score the groups, pick one. Not three. One segment for the next 90 days.
This feels restrictive and it is the opposite. Prompts, pages, outreach: everything you build later inherits its focus from this single choice, and every later module gets easier because you made it.
- 01
Export the last 20 closed-won deals and group them by company type, size, use case, and buyer role.
- 02
Score each group from 1-5 for win rate, ACV, sales-cycle speed, retention, and access to customer evidence.
- 03
Select one segment for this 90-day cycle. Write it as: [role] at [company type] who need [job] before [trigger].
- In HubSpot: Reports > Deals, filter Deal Stage = Closed Won, close date in the last 12 months, then Export as CSV. In Salesforce: build an Opportunities report with the same filters.
- Paste the export into a spreadsheet and add four columns you fill by hand: company type, size band, use case, buyer role. Fifteen minutes of typing, not a data project.
- Add the five scoring columns from the steps, score each group 1-5, and sum. The winner is usually obvious by row ten.
You can say your segment out loud in one sentence, with role, company type, job, and trigger, and your sales lead nods instead of adding "well, also...".
- NO BUDGET?
No CRM? Use your invoice list. Twenty customers, four columns, and a 20-minute call with whoever does the selling gets you the same answer.
- WATCH OUT
Do not pick the segment you wish you sold to. Score what actually closed. Aspirational segments belong in next year's plan, not this quarter's map.
- PRO TIP
Fewer than 20 closed deals? Use every deal you have ever won plus your three best-fit current customers, and weight retention heavily.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
HubSpot / Salesforce, plus tagged call evidence from Gong / call notes.
| Segment | Wins | Win rate | Median ACV | Sales cycle | Retention | Evidence | Weighted score |
|---|---|---|---|---|---|---|---|
| Mid-market agencies | 14 | 34% | $18k | 41 days | 93% | 8 calls | 23.4 / 30 |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Choose one revenue segment Objective: A broad ICP produces a broad content plan. Pick one segment where the problem, buying process, and product fit are similar enough to write a precise answer. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Opportunity export exported from HubSpot / Salesforce - Tagged call evidence exported from Gong / call notes - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Export the last 20 closed-won deals and group them by company type, size, use case, and buyer role. 2. Score each group from 1-5 for win rate, ACV, sales-cycle speed, retention, and access to customer evidence. 3. Select one segment for this 90-day cycle. Write it as: [role] at [company type] who need [job] before [trigger]. REQUIRED OUTPUT Return a table using these exact columns: Segment | Wins | Win rate | Median ACV | Sales cycle | Retention | Evidence | Weighted score. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: Head of Marketing at a 20-100 person B2B SaaS company who needs reliable attribution before the next board meeting.
One segment is named precisely. A named human owner must verify this before the lesson is complete.
Head of Marketing at a 20-100 person B2B SaaS company who needs reliable attribution before the next board meeting.
Mine buyer language
Your best query research is already hiding in sales calls, onboarding notes, support tickets, and reviews.
Somewhere in a sales call last month, a buyer leaned toward the mic and said the exact sentence they will one day type into ChatGPT. It is sitting in a transcript right now. Free. Nobody has read it twice.
That is why this lesson bans paraphrasing. 'Reporting is slow' and 'reports take all of Monday' feel like the same idea; they search completely differently, and only one of them is what a buyer actually says. Collect quotes verbatim, tally repeats, and when the same phrase shows up in five different mouths you have found the vocabulary of your category.
Expect this to take an afternoon, and expect at least five phrases to surprise you. The surprises are the point.
BUYER Honestly the biggest thing is that reports take all of Monday. Every client wants their own format, so my team rebuilds the same dashboard over and over.
BUYER We tried doing it in spreadsheets, but we can't see which campaigns actually drive pipeline, so the monthly call turns into us defending numbers instead of showing progress.
REP What would good look like by the next board meeting?
BUYER Something white-label we can send under our own brand, that pulls straight from Google Ads and GA4 without anyone touching it on a Sunday night.
- 01
Collect 10 sales transcripts, 10 support conversations, five reviews, and the notes from five lost deals.
- 02
Highlight exact phrases for pain, desired outcome, failed approach, evaluation criteria, objection, and trigger.
- 03
Keep the buyer wording. Add a tally each time the same idea appears; frequency reveals the category vocabulary.
- Call recordings live in Gong, Chorus, Fireflies, or your Zoom account (Zoom: Recordings > Cloud Recordings > Audio transcript). Download ten transcripts into one folder.
- Support language: in Intercom or Zendesk, export or skim the 30 most recent conversations tagged with your product area.
- Reviews: open your G2 and Capterra pages and your top two competitors' pages. Copy review sentences into the sheet verbatim, with a link to the source.
- Highlight in one pass per transcript: pain, outcome, failed approach, objection, trigger. One color each if your doc tool allows it.
You have at least 30 verbatim phrases with a tally next to each, and at least five of them surprised you.
- PRO TIP
Competitor 1-star and 2-star reviews are the goldmine. They are switching triggers written by your future customers, in the exact words they will type into ChatGPT.
- WATCH OUT
Never paraphrase while collecting. 'Reporting is slow' and 'reports take all of Monday' look similar and search completely differently.
- NO BUDGET?
No recorded calls? Sit in on three live sales calls this week and interview two current customers for 20 minutes each. Ask 'what were you doing before us?' and shut up.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
Gong / call notes, plus query export from GSC + Semrush / Ahrefs.
| Exact phrase | Role | Stage | Source | Frequency | Meaning | Human approved |
|---|---|---|---|---|---|---|
| Reports take all Monday | Ops lead | Problem | Call 184 | 7 | Manual consolidation | Yes |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Mine buyer language Objective: Your best query research is already hiding in sales calls, onboarding notes, support tickets, and reviews. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Tagged call evidence exported from Gong / call notes - Query export exported from GSC + Semrush / Ahrefs - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Collect 10 sales transcripts, 10 support conversations, five reviews, and the notes from five lost deals. 2. Highlight exact phrases for pain, desired outcome, failed approach, evaluation criteria, objection, and trigger. 3. Keep the buyer wording. Add a tally each time the same idea appears; frequency reveals the category vocabulary. REQUIRED OUTPUT Return a table using these exact columns: Exact phrase | Role | Stage | Source | Frequency | Meaning | Human approved. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: “We cannot explain which campaigns influence pipeline” is stronger source material than “marketing analytics software.”
At least 25 buyer questions are mapped. A named human owner must verify this before the lesson is complete.
“We cannot explain which campaigns influence pipeline” is stronger source material than “marketing analytics software.”
Build the demand ladder
Map the questions a buyer asks from first pain to final vendor validation. This prevents a content plan made entirely of awareness posts.
Follow one buyer for six weeks. Monday she is annoyed that reports eat a full day. By Wednesday she is searching 'how do agencies automate client reporting'. A month later it is 'best agency reporting software', then 'Swydo vs AgencyAnalytics', and finally, the night before she signs, 'is it white-label, is it SOC 2, who else uses it'.
Five stages, different questions at each, and here is the uncomfortable part: most content plans answer only the first one. Twenty awareness posts, zero pages for the person choosing a vendor this quarter. The demand ladder exists to make that gap visible before you spend a dollar filling the wrong end of it.
Why is reporting taking two days every month?MOST PLANS STOP HERE
How do agencies automate client reporting?
Best agency reporting software
Swydo vs AgencyAnalytics for a 20-client agency
Is it white-label? Is it SOC 2? Who else uses it?DEALS CLOSE HERE
- 01
Create five stages: problem, approach, category, comparison, and validation.
- 02
Write five real buyer questions at every stage, using evidence gathered in the previous lesson.
- 03
For each question, add the best page type, proof needed, business value, and whether buyers ask it in Google, AI, or both.
- Make a five-column sheet: Problem, Approach, Category, Comparison, Validation. Write questions down the rows, five per column minimum.
- Every question gets a source cell: the call, ticket, or review it came from. If you cannot fill the source cell, the question is a guess.
- To mark where each question gets asked, type it into Google (watch the autocomplete and the People Also Ask box) and paste it into ChatGPT. Note whether it produces links, an answer, or a shortlist.
Every stage has five questions, every question traces back to evidence from Lesson 2, and the comparison and validation rows are not empty.
- PRO TIP
Validation-stage questions hide in security questionnaires, procurement emails, and lost-deal notes. Ask sales for the last three questionnaires they filled in.
- WATCH OUT
Do not invent questions to make the grid look complete. An empty validation column is a finding: it means you have never heard how deals actually close.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
GSC + Semrush / Ahrefs, plus prompt families from ChatGPT / Perplexity.
| Problem | Approach | Category | Vendor | Validation | Target asset |
|---|---|---|---|---|---|
| Manual client reporting | Automate reports | Agency reporting software | Swydo alternatives | White-label proof | /agency-reporting/ |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Build the demand ladder Objective: Map the questions a buyer asks from first pain to final vendor validation. This prevents a content plan made entirely of awareness posts. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Query export exported from GSC + Semrush / Ahrefs - Prompt families exported from ChatGPT / Perplexity - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Create five stages: problem, approach, category, comparison, and validation. 2. Write five real buyer questions at every stage, using evidence gathered in the previous lesson. 3. For each question, add the best page type, proof needed, business value, and whether buyers ask it in Google, AI, or both. REQUIRED OUTPUT Return a table using these exact columns: Problem | Approach | Category | Vendor | Validation | Target asset. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: Problem: “Why is our attribution different in every tool?” Comparison: “Dreamdata vs HockeyStack for B2B SaaS.”
Every question has evidence and a stage. A named human owner must verify this before the lesson is complete.
Problem: “Why is our attribution different in every tool?” Comparison: “Dreamdata vs HockeyStack for B2B SaaS.”
Translate buyer language into query families
Buyer phrases and search queries are different strings for the same decision. The translation is where keyword research actually happens, and where most of it goes wrong.
Nobody types 'reports take all Monday' into Google. They type 'agency reporting software'. The buyer phrase and the search query are different strings that mean the same decision, and the translation between them is where keyword research actually lives. Skip it and you end up optimising for the words in your language library while buyers search for words you never mapped.
So translate deliberately. For every question on the ladder, pull the queries Search Console already shows you and the variants a keyword tool suggests, then group them by the decision they represent, not by the words they share. 'Reporting software for agencies', 'agency reporting tool', and 'client reporting platform' are one family. 'Reporting software' on its own is a different decision with a different page. Volume goes in as the last column, so it breaks ties instead of picking winners.
- 01
For each ladder question, pull the queries Search Console already shows you and the variants a keyword tool suggests, then group them by the decision they represent rather than by the words they share.
- 02
Give each family one canonical phrasing in buyer language, list its variants, and note the SERP type from the Module 03 check.
- 03
Record search volume as the last column, so it breaks ties between families instead of choosing them.
- In Search Console, Performance > Search results > Queries, filter to the last 12 months and export. These are the phrasings buyers already use to find you; group them under the ladder question each one answers.
- For questions with no Search Console data yet, use Google Keyword Planner or the keyword tool you already pay for to pull variants. Type the buyer phrase, not a guess at the keyword, and keep the ones that describe the same decision.
- Build the query-family sheet: buyer phrase, family name, canonical query, variants, SERP type (from the Module 03 check), volume, stage. One row per family, volume last.
Every ladder question maps to a named query family with one canonical phrasing, its variants, its SERP type, and a volume column that sits last.
- WATCH OUT
Clustering by shared words instead of shared decisions is the classic mistake. 'Reporting software' and 'reporting software for agencies' share three words and answer two different questions.
- PRO TIP
A query family with no volume yet is not dead. AI prompts often lead search demand by a quarter or two, so check the Module 05 prompt set before parking it.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
ChatGPT / Perplexity, plus opportunity export from HubSpot / Salesforce.
| Buyer phrase | Query family | Canonical query | Variants | SERP type | Volume | Stage |
|---|---|---|---|---|---|---|
| Reports take all Monday | Agency reporting software | reporting software for agencies | agency reporting tool, client reporting platform | Product + use-case pages | 1,900 | Category |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Translate buyer language into query families Objective: Buyer phrases and search queries are different strings for the same decision. The translation is where keyword research actually happens, and where most of it goes wrong. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Prompt families exported from ChatGPT / Perplexity - Opportunity export exported from HubSpot / Salesforce - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. For each ladder question, pull the queries Search Console already shows you and the variants a keyword tool suggests, then group them by the decision they represent rather than by the words they share. 2. Give each family one canonical phrasing in buyer language, list its variants, and note the SERP type from the Module 03 check. 3. Record search volume as the last column, so it breaks ties between families instead of choosing them. REQUIRED OUTPUT Return a table using these exact columns: Buyer phrase | Query family | Canonical query | Variants | SERP type | Volume | Stage. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: “Reporting software for agencies,” “agency reporting tool,” and “client reporting platform” are one family. “Reporting software” alone is a different decision and a different page.
The top 10 opportunities are scored and selected. A named human owner must verify this before the lesson is complete.
“Reporting software for agencies,” “agency reporting tool,” and “client reporting platform” are one family. “Reporting software” alone is a different decision and a different page.
Prioritize by revenue, not volume
A low-volume comparison can be worth more than a high-volume definition. Use a score that reflects commercial value.
Time to choose, and the counter-intuitive rule is that search volume is the least important number on the sheet. A comparison query with 40 searches a month, asked by your exact segment at the moment of decision, will out-earn a 2,000-volume definition every time. Small audiences close; big audiences browse.
So run the score honestly and let the math argue with your instincts. Fit, intent, evidence, sales value, minus difficulty. Take the top ten and park the rest. Parking is not killing; a parked idea just waits for a quarter that deserves it.
- 01
Score business fit, buying intent, evidence advantage, and sales usefulness from 1-5.
- 02
Subtract execution difficulty from 1-5. Use: fit + intent + evidence + sales value - difficulty.
- 03
Move the ten highest-scoring questions into the next modules. Park everything else in a later backlog.
- Add the four scoring columns plus difficulty to the ladder sheet and compute: fit + intent + evidence + sales value minus difficulty.
- Sort by score, take the top ten into a new tab named 'This quarter', and move everything else to a tab named 'Parked'.
- Book 30 minutes with your sales lead and walk the top ten. Their job is to veto anything that never comes up in real deals.
Ten questions are selected, each has a score you could defend to your CFO, and at least half are comparison or validation stage.
- PRO TIP
Break ties toward comparison and validation questions. They are closer to money and they double as AEO targets in Module 05.
- WATCH OUT
Be honest about difficulty. If a question needs data you do not have or an expert who has no time, that is a real cost. Score it.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
HubSpot / Salesforce, plus tagged call evidence from Gong / call notes.
| Demand family | Revenue evidence | Search / prompt evidence | Current gap | Effort | Priority | Owner |
|---|---|---|---|---|---|---|
| Automate client reports | 11 wins | 1,120 queries | No use-case page | Medium | P1 | Growth |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Prioritize by revenue, not volume Objective: A low-volume comparison can be worth more than a high-volume definition. Use a score that reflects commercial value. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Opportunity export exported from HubSpot / Salesforce - Tagged call evidence exported from Gong / call notes - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Score business fit, buying intent, evidence advantage, and sales usefulness from 1-5. 2. Subtract execution difficulty from 1-5. Use: fit + intent + evidence + sales value - difficulty. 3. Move the ten highest-scoring questions into the next modules. Park everything else in a later backlog. REQUIRED OUTPUT Return a table using these exact columns: Demand family | Revenue evidence | Search / prompt evidence | Current gap | Effort | Priority | Owner. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: A query with 40 monthly searches and a score of 16 outranks a 2,000-volume glossary term with a score of 7.
The top 10 opportunities are scored and selected. A named human owner must verify this before the lesson is complete.
A query with 40 monthly searches and a score of 16 outranks a 2,000-volume glossary term with a score of 7.
Write the identity sentence and attribute list
One sentence that says what you are, for whom, why you win, and what you are not, plus the attribute vocabulary every prompt, page, and profile will repeat.
Ask four people at your company what the product does and you will usually get four answers. Marketing has a tagline, sales has a pitch, the founder has a vision, and the G2 profile still has the 2023 version. A model reading all four does the only thing it can: it averages them into fog, and fog does not get recommended.
The identity sentence ends the argument. What you are, who it is for, why you win, and the clause nobody wants to write: what you are not for. That last part is what makes the rest believable, to a buyer and to a machine.
Then mine it into vocabulary. Categories, verticals, pains, use cases, integrations, roles. Write the list once, with evidence, and Modules 05, 07, and 08 will all draw from it instead of reinventing you every quarter.
- 01
Fill the template: "We are a [category] for [audience] who need [capability]. We win because [differentiator]. We are not [what you are not for]." Use buyer language from Lesson 2, not brand adjectives.
- 02
Build the attribute list in six groups: category terms, industries or verticals, pain points, use cases, integrations, and buyer personas. Pull every entry from evidence, with a tally.
- 03
Score each attribute from 1-5 for purchase influence, current AI representation gap, and influenceability; attributes scoring high on all three are your first targets.
- Open a fresh doc and fill the template in one sitting: 'We are a [category] for [audience] who need [capability]. We win because [differentiator]. We are not [what you are not for].'
- Send it to the founder, the sales lead, and one customer-facing person with a single question: 'What here would you refuse to say to a prospect?' Fix only that.
- Add a second tab for the attribute list, grouped into the six categories from the steps, each entry with its evidence tally from Lesson 2.
The identity sentence fits the template without a single adjective, and every attribute on the list points to a call, review, or ticket that produced it.
- PRO TIP
Read the sentence out loud. If you would be embarrassed to say it on a sales call, rewrite it for a human reader and remove claims you cannot defend.
- WATCH OUT
Ban adjectives. 'Leading', 'powerful', and 'seamless' carry zero information, and a model summarizing you will drop them anyway.
- NO BUDGET?
Stuck on wording? Study your three best reviews for customer language, then verify that every resulting product claim is current and representative.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
Gong / call notes, plus query export from GSC + Semrush / Ahrefs.
| Attribute | Group | Evidence | Purchase influence | AI gap | Influenceability | Priority |
|---|---|---|---|---|---|---|
| White-label reporting | Use case | 9 calls / 6 reviews | 5 | 4 | 4 | P1 |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 01: Start with the market, not keywords. Current lesson: Write the identity sentence and attribute list Objective: One sentence that says what you are, for whom, why you win, and what you are not, plus the attribute vocabulary every prompt, page, and profile will repeat. Required artifact: Buyer demand map BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Tagged call evidence exported from Gong / call notes - Query export exported from GSC + Semrush / Ahrefs - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Fill the template: "We are a [category] for [audience] who need [capability]. We win because [differentiator]. We are not [what you are not for]." Use buyer language from Lesson 2, not brand adjectives. 2. Build the attribute list in six groups: category terms, industries or verticals, pain points, use cases, integrations, and buyer personas. Pull every entry from evidence, with a tally. 3. Score each attribute from 1-5 for purchase influence, current AI representation gap, and influenceability; attributes scoring high on all three are your first targets. REQUIRED OUTPUT Return a table using these exact columns: Attribute | Group | Evidence | Purchase influence | AI gap | Influenceability | Priority. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: "We are an agency reporting platform for 5-50 person marketing agencies who need white-label client reports without manual rebuilds. We win because reports pull live from ad platforms on a schedule. We are not a BI tool for data teams."
The top 10 opportunities are scored and selected. A named human owner must verify this before the lesson is complete.
"We are an agency reporting platform for 5-50 person marketing agencies who need white-label client reports without manual rebuilds. We win because reports pull live from ad platforms on a schedule. We are not a BI tool for data teams."
What you are looking at below is a filled-in example of the artifact you are building: a demand map for a fictional reporting SaaS selling to mid-market agencies. The numbers are demo data. Your job is not to admire it. Rebuild every panel with your own CRM export and call notes, and use the tabs to see the sheet, the scenarios, and the exact workflow.
Buyer demand control room
Combine CRM outcomes, sales language, search demand, and AI prompts before deciding what to publish.
Demand by decision stage
Weight themes by commercial evidence, not search volume alone.
Turn demand data into a strategy.
Traffic is an input. The decision is where evidence, commercial intent, product fit, proof, and team capacity justify action. Changes save automatically on this device.
Start with modifiers, then inspect the live SERP, ranking page types, sales calls, and the next question. A human assigns the final intent.
CRM opportunities and closed revenue outrank modeled value. Use modeled pipeline only where attribution is missing, and label it as directional.
Rank every demand family, then limit the active list to actual quarterly capacity. Everything else is validated, supported, or parked.
Modifiers suggest intent. The ranking page types and buyer evidence confirm it.
| Intent | Query clues | SERP proof | Buyer decision | Likely asset |
|---|---|---|---|---|
| Diagnose | why, problem, meaning | Guides and definitions | Understand the problem | Expert guide |
| Solution | how to, automate, improve | Processes and use cases | Choose an approach | Use-case page |
| Category | software, platform, tools | Product and category pages | Choose a solution type | Product / category |
| Compare | best, top, vs, alternatives | Lists and comparisons | Build a shortlist | Comparison / best-of |
| Vendor | brand, reviews, demo | Vendor and review pages | Evaluate a named vendor | Vendor / proof |
| Pricing | price, cost, plans, ROI | Pricing and calculators | Validate commercial fit | Pricing / ROI |
| Validate | integration, security, migration | Docs, proof, implementation | Remove purchase risk | Integration / proof |
Use the strongest evidence available and keep weaker estimates visibly separate.
| Method | Evidence | How to use it | Confidence |
|---|---|---|---|
| Sourced revenue | Landing page or content is the recorded acquisition source | Use directly | Highest |
| Influenced revenue | Asset was touched before opportunity or close | Report separately with window | High |
| Opportunity evidence | Demand family appears in calls, forms, notes, or campaigns | Opps x win rate x ACV | Medium-high |
| Paid-search evidence | CPC, conversion, and opportunity quality exist | Use as commercial validation | Medium |
| Modeled potential | No direct CRM history exists | Volume x CTR x conversion x ACV | Directional |
Use your own GSC, analytics, and CRM rates. These assumptions affect modeled potential, not observed revenue.
| Now | Demand family | Intent | SV | CRM revenue | Opps | Win % | ACV | Fit | Evidence | Gap | Sales | Effort | Page state | Opp expected revenue | Modeled pipeline | Score | Next action |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| YES | $39,440 | $13,260 | 90.5 | Refresh + prove | |||||||||||||
| YES | $45,360 | $102,600 | 89.0 | Refresh + prove | |||||||||||||
| YES | $43,560 | $31,680 | 88.0 | Refresh + prove | |||||||||||||
| YES | $61,380 | $38,880 | 87.0 | Build revenue page | |||||||||||||
| YES | $43,320 | $17,670 | 85.0 | Refresh + prove | |||||||||||||
| LATER | $49,980 | $24,570 | 84.5 | Build revenue page | |||||||||||||
| LATER | $20,800 | $18,720 | 74.0 | Fix blocker first | |||||||||||||
| LATER | $17,600 | $60,000 | 69.5 | Consolidate URLs | |||||||||||||
| LATER | $4,320 | $86,400 | 44.0 | Park | |||||||||||||
| LATER | $0 | $104,400 | 31.5 | Park |
80-100: build or refresh now. 65-79: validate and queue. 50-64: supporting backlog. Below 50: park. Overlap and technical blockers override the score.
Best: sourced or influenced CRM revenue. Next: opportunity and win evidence. Last: modeled pipeline using explicit assumptions. Never present modeled value as revenue.
Check each item only after the artifact meets the standard. Progress saves on this device; use the academy-home backup to move it elsewhere.