Subscribe

Recipes / Case study shortlist: interview questions from closed-won customers

Case study shortlist: interview questions from closed-won customers

This recipe uses the official HubSpot, Gong and Notion MCP servers, and all of them work in Claude and ChatGPT. Steps that write, send or spend credits wait for your approval.

IntermediateMarketing

Who is this for?

Marketers who want a shortlist of closed-won customers for case study interviews, plus the outcomes Gong reports and a page of questions. This is for marketing content. The reference-match recipe finds a customer who looks like one prospect, for a rep to take to the account owner. This one starts from customers you already won and stops at interview questions. HubSpot can be swapped for Salesforce, Gong for Fireflies or Granola, and Notion for Airtable if the shortlist should live in a base.

What do you get?

Shortlist and interview questions, page not created yet. Sample output: every company and person in it is fictional.

Closed won in HubSpot, Jul 1 to Sep 30: 11 deals, 9 companies. Suggested for Gong, after you confirm: 4 companies. That is 4 ask_account calls. generate_brief is not included.

CompanyOwnerClosed wonAmountGong summary (not a quote)On the shortlist?
Larkspur Freightworks (fictional)Priya RamanAug 14$46KThe ops lead said they hear about a late load before the customer calls. Gong named the Aug 2 call. No percentage.Yes
Mossgate Labs (fictional)Sam WhitlockSep 3$28KGong could not tie a result to a call.No. Listed under no outcome found.
Quarry Lane Analytics (fictional)Dana OkaforJul 22$61KThe security lead said the review finished in one cycle. Gong named the Jul 9 call. No day count.Yes
Fernhill Software (fictional)Alex RiveraSep 18$19KLeft off the suggested list. Amount is under the $20K line you set.Not sent to Gong

No outcome found: Mossgate Labs. Private calls are excluded, so a missing answer is not proof of a bad result. Interview questions for Larkspur Freightworks (fictional), for you to ask. Not an email to the owner. 1. What did a late-load report look like before this? 2. Who on the team sees the alert on a normal morning? 3. Which part of that story are you willing to put in a case study, and which number should we leave out? Page draft (not created yet): 'Case study shortlist: Q3' Not a reference ask. No prospect. No note to the account owner. Nothing was created in Notion. No one was contacted.

How long does it take, and what does it cost?

Estimate About 30–45 minutes, most of it checking Gong's summaries and the questions.

Assumptions: One quarter, about four companies you approve for Gong, and one Notion page.

Exact credit costs depend on your plan with each vendor. Check the tool pages before a large run.

Which MCP servers does it use?

Every server is confirmed in: ClaudeChatGPT

How do I set it up in a chat app?

Every server in this recipe is confirmed for Claude and ChatGPT. Connect each one, then paste the prompts below into one conversation.

Developer setup

Auth setup and mcp.json for Claude Code and Cursor
  • HubSpot: Sign in with OAuth in your client (run /mcp in Claude Code). Do it once interactively, and the client stores and refreshes the token.
  • Gong: Sign in with OAuth in your client (run /mcp in Claude Code). Do it once interactively, and the client stores and refreshes the token.
  • Notion: Sign in with OAuth in your client (run /mcp in Claude Code). Do it once interactively, and the client stores and refreshes the token.

Put this in .mcp.json (Claude Code, project root) or ~/.cursor/mcp.json (Cursor). Swap the example keys for your own, and don't commit real ones. Remove the // comment lines if your client requires strict JSON.

{
  "mcpServers": {
    "hubspot": {
      "url": "https://mcp.hubspot.com"
    },
    "gong": {
      "url": "https://mcp.gong.io/mcp"
    },
    "notion": {
      "url": "https://mcp.notion.com/mcp"
    }
  }
}

Step-by-step prompts

  1. Find HubSpot deals closed won between {start_date} and {end_date}. Show a table with company, deal name, amount, close date and owner. Group the rows by company and give me the deal total and the company total. Don't change any records.
  2. From that table, suggest up to {n} companies for a case study. Prefer different owners, and skip deals under {min_amount}. One company is one Gong ask_account call, which uses Gong credits. Show the list and the call count, and wait for me to confirm it. Don't call Gong yet.
  3. After I approve the list, use Gong ask_account once per company. Ask what outcomes and results the customer mentioned, which calls those came from, and what Gong could not find. Label each answer as a summary, not a quote. If ask_account cannot find the account, stop on that company and say so. Don't call ask_deal or generate_brief unless I approve that call on its own.
  4. Draft a shortlist here: company, owner, the outcome in Gong's words, and three interview questions that ask the customer to confirm that outcome. Leave out anyone Gong could not tie to a result. These questions are for a case study interview. Don't draft a note to the account owner, and don't contact anyone.
  5. Show me the full page, titled {page_title}. After I approve it, create that page in Notion under {parent_page}. Don't change any other page.

Can I run it on a schedule? (optional)

Where does this go wrong?

Gotchas

Download

Get the prompts, guardrails and expected output as one Markdown file. Paste it into a Claude Project or ChatGPT project's instructions, or save it as CLAUDE.md for Claude Code.

Download case-study-shortlist.md

Vendor sites: HubSpot. If we join an affiliate program, these links will be labeled as affiliate links.