Steps to analyze customer data as a solo founder
This resource is based on The AI research stack behind a solo founder's Nordstrom deal, featuring Marianna Sachse of Jackalo, published on the AI Lab by ActiveCampaign.
Get the checklist
Marianna Sachse, founder of Jackalo (sustainable kids’ clothing), uses AI to analyze customer calls, survey responses, and Shopify data—without a dedicated analyst. The result: an 80% increase in website conversion and 250% increase in email conversion. Work through this checklist once and you’ll have a repeatable system for turning raw customer data into clear business decisions. Expect to spend about 2–3 hours the first time through; subsequent runs take 30–45 minutes.
Before you start
- Create accounts for at least two AI tools — Sachse cross-references ChatGPT, Claude, and Gemini to catch what any single tool misses
- Set up Fathom (or another call recorder) and connect it to your meeting tool (Zoom, Google Meet, etc.)
- Identify your data sources — which customer calls, survey tool (Google Forms or Typeform), and Shopify reports are most relevant right now
The workflow
Phase 1: Analyze customer calls
After this phase, you’ll have: A prioritized list of key learnings and action items from your customer conversations
- Record your next customer call using Fathom — it auto-transcribes and can summarize the call as it ends
- Export the transcript from Fathom once the call wraps
- Open two AI tools side by side — paste the transcript into both ChatGPT and Claude (or Gemini) separately
- Run the analysis prompt below in both tools:
- Compare the outputs — note where the two tools agree and where they diverge; divergences often surface nuances worth digging into
- Ask follow-up questions in whichever tool gave the more useful initial analysis:
- Layer on your intuition — review the AI’s output against what you remember from the call; note anything the AI flagged that you’d glossed over, and anything it missed that felt important to you
- Document your action items in Notion or wherever you track decisions
Analysis prompt
Here is a transcript from a customer call. I sell [PRODUCT/SERVICE] to [CUSTOMER DESCRIPTION].
Please analyze this conversation and give me:
- The top 3–5 key learnings about my customer’s needs, frustrations, or motivations
- Any patterns you notice in how they describe their problem
- Specific action items I should take based on what they said
- Any quotes that stand out as particularly revealing
Transcript:
[PASTE TRANSCRIPT]
Follow-up prompt
What did this customer say (or imply) about [SPECIFIC TOPIC: e.g., price sensitivity, fit concerns, competitor comparisons]? Is there anything in the transcript that speaks to [BUSINESS QUESTION]?
“I want to be able to give a half hour here and there for customer calls, but I want the analysis to be really fast.”
Phase 2: Review customer surveys
After this phase, you’ll have: A synthesized analysis of survey responses plus suggested improvements to the survey itself
- Export your survey data as a CSV from Google Forms (Responses > Download CSV) or Typeform (Results > Export)
- Record a short voice memo (2–3 minutes) summarizing the context: what you were trying to learn, who you sent this survey to, and what business decision you’re trying to make
- Transcribe the voice memo using your phone’s transcription or a tool like Otter.ai, then paste it into ChatGPT along with the CSV
- Run the survey analysis prompt below:
- Check for gaps in the analysis — Sachse notes AI “sometimes focuses on what it thinks is most important and will miss a whole part of the survey.” Skim the raw CSV yourself and ask: Did the AI address [section X]?
- Prompt it to fill the gaps.
- Ask for survey improvement suggestions:
- Document themes and survey improvements for your next send
Survey analysis prompt
I’m uploading results from a customer survey. Here’s the context:
[PASTE VOICE MEMO TRANSCRIPT OR BRIEF DESCRIPTION]
Please analyze the responses and tell me:
- The main themes and patterns in what respondents said
- Any surprising or unexpected findings
- The most important things I should act on
- What segments or groups of respondents seem to have different experiences or needs
[ATTACH CSV FILE]
Analysis gap prompt
You focused mainly on [TOPIC]. I also need you to analyze the responses to [SPECIFIC QUESTION OR SECTION]. What do those responses reveal?
Survey improvement prompt
Based on the analysis and the responses you’re seeing, where could I be asking different questions? What would give me more useful data next time?
Phase 3: Dig into Shopify data
After this phase, you’ll have: Answers to your most pressing business questions, grounded in your actual store data
- Identify the business question you want to answer before downloading anything — e.g., “Which products have the highest return rate?” or “What’s my conversion rate by traffic source?”
- Download the relevant Shopify report (Analytics > Reports — pick the report type that matches your question: Sales, Customers, Inventory, etc.)
- Open Claude — Sachse finds it especially strong for data analysis
- Upload the report and run the analysis prompt:
- Ask targeted follow-up questions to go deeper:
- Cross-reference with your own expertise — you know things the data doesn’t show (a slow month because you were at a trade show, a spike because of a press mention). Add that context to the conversation:
- Document decisions — note what the data showed, what you decided, and what you’ll check in 30–60 days to see if it worked
Analysis prompt
I’m uploading a Shopify [REPORT TYPE] report for my [DESCRIBE YOUR BUSINESS: e.g., sustainable kids’ clothing brand]. Here’s the business question I’m trying to answer:
[YOUR SPECIFIC QUESTION]
Please analyze the data and give me:
- A direct answer to my question
- Any patterns or anomalies in the data worth noting
- What this data suggests I should do or investigate further
Targeted follow-up
What does this data suggest about [SPECIFIC PRODUCT / CUSTOMER SEGMENT / TIME PERIOD]? Is there anything in here that contradicts what I’d expect to see?
Contextual prompt
[CONTEXTUAL NOTE: e.g., “Sales dropped in March because we were out of stock on our top SKU.”] Does that change your interpretation of the data?
“I’ve been pleased by the number of people who look at my company and think we must be 10 people.”
Quick reference
- Total time: 2–3 hours to set up; 30–45 minutes per data analysis session
- Tools needed: Fathom (call recording), ChatGPT, Claude, Gemini, Shopify (for reports), Google Forms or Typeform (for surveys), Notion (for documentation)
- Key output: A repeatable system for turning customer calls, survey responses, and store data into clear action items — without a dedicated analyst
Ready for the full story?
Read The AI research stack behind a solo founder’s Nordstrom deal, featuring Marianna Sachse of Jackalo, published on the AI Lab by ActiveCampaign.
Related
More data from the AI Lab.