Strategy & Planning

How to synthesize customer research interviews as CPG Founders

Strategy & PlanningFor CPG Founders4 apps11 steps~22 min to set up

You run 6-12 customer discovery calls a month — with buyers at Sprouts, DTC customers who churned, retail dietitians, and co-packer reps — and the synthesis never happens. You take notes in Google Docs, maybe record the call in Zoom, and by the time you surface from a week of production planning and deduction disputes, you've forgotten which buyer said your 8oz SKU was too expensive vs. which one said the shelf placement was the problem. The insight is in the recordings. The action never makes it to your product roadmap or your pitch deck. You're making packaging and formulation decisions on vibes instead of a clean read of what 30 customers actually told you.

Build this on Starch → See the apps used

Strategy & PlanningFor CPG Founders4 apps11 steps~22 min to set up

Outcome

What you'll set up

A searchable archive of every customer interview — transcribed, summarized, and tagged by theme (price sensitivity, packaging feedback, retail placement, flavor preference) — so you can pull quotes by topic in 30 seconds instead of scrubbing recordings

An AI-generated synthesis doc after every batch of interviews that surfaces the top 3 patterns, the sharpest verbatims, and the open questions your next round should answer

A one-click presentation you can drop into a board update or pitch deck showing what your customers actually said, backed by data from your growth and conversion metrics

The Starch recipe

Apps, data, and prompts

The combination of Starch apps, the data sources they pull from, and the prompts you use to drive them.

Apps used

Meeting Notes Knowledge Management Growth Analyst Presentation Agent

Data sources & config

Meeting Notes transcribes and summarizes calls as they happen. Interview summaries and tagged themes are stored in Knowledge Management (Starch connects directly to Notion where your existing interview notes live, via scheduled sync). Growth Analyst connects to PostHog from Starch's integration catalog — the agent queries it live — to pull traffic, conversion, and retention data that cross-validates what customers told you. Presentation Agent compiles the synthesis into a polished deck on demand.

Prompts to copy

Transcribe this customer interview recording, extract every mention of price, packaging, flavor, and purchase frequency, and write a 200-word summary with the three most actionable insights

Across the last 12 customer interviews stored in my Knowledge Base, identify the top 5 recurring themes, pull the strongest supporting quote for each, and flag any contradictions between what DTC customers say vs. what retail buyers say

Using my PostHog conversion data and the customer interview themes from this month, write me a one-page synthesis I can share with my co-founder: what customers are telling us vs. what the data confirms or contradicts

Build a 10-slide deck summarizing our Q2 customer research for our next investor update — include key themes, representative quotes, product implications, and the two decisions we're making based on this research

Run these in Starch → or paste them into your favorite agent

Walkthrough

Step-by-step

1 Connect Notion to Starch via scheduled sync so any existing interview notes, research docs, or Dovetail exports you've pasted in are automatically pulled into your Knowledge Base.

2 Install the Meeting Notes app and run your next customer call through it — it transcribes in real time and generates a summary with key decisions, action items, and verbatim highlights the moment the call ends.

3 After each interview, prompt Starch: 'Tag this transcript with the following themes: price sensitivity, pack size feedback, flavor preference, purchase occasion, retail channel friction, and churn reason' — Starch adds structured tags to every entry in your Knowledge Base.

4 After every 5-10 interviews, run the cross-interview synthesis prompt — ask Starch to surface the top recurring themes, pull the strongest supporting quote for each, and flag where DTC customer feedback contradicts retail buyer feedback.

5 Connect PostHog from Starch's integration catalog so the Growth Analyst can pull your actual conversion and retention data — this lets you cross-validate what customers say against what they do.

6 Run the cross-validation prompt: 'Customers in interviews say price is the main barrier to repeat purchase — does my PostHog data show higher churn at the point after the first order, or is drop-off earlier in the funnel?' Starch returns a short analysis with the relevant numbers.

7 Use Knowledge Management's AI search to pull every quote mentioning a specific topic — for example, 'find every interview where a buyer mentioned shelf placement or facings' — before a key retailer meeting so you walk in prepared.

8 When a contradiction surfaces between what customers say and what the data shows, log it as an open question in Knowledge Management so your next interview round is targeted at resolving it rather than just collecting more of the same signal.

9 Before a board meeting or fundraise, prompt Presentation Agent: 'Build a 10-slide customer research summary for our Series A deck — include our research methodology, top 5 themes with supporting quotes, the product decisions we're making, and the open questions we're still testing.' Export to PDF or PowerPoint in one click.

10 Set a monthly reminder automation: 'On the first Monday of each month, pull all interview summaries from the last 30 days, generate a one-page synthesis of new themes vs. themes we've seen before, and Slack it to me.' This keeps the research loop running without a dedicated researcher.

11 When a retail pitch or co-packer negotiation goes sideways, search your interview archive for relevant context — 'what have buyers told us about our minimum order quantities?' — so you're negotiating from evidence, not memory.

See this running on Starch

Connect your tools, describe what you want, and the agent builds it. Closed beta is free.

Try it on Starch →

Worked example

April 2026 DTC Churn Research Sprint — 8 interviews, 2 retail buyer calls

Sample numbers from a real run

Interviews conducted (DTC customers who churned after order 2)	8
Retail buyer calls (Sprouts regional buyer + one natural grocery chain)	2
Themes tagged across all 10 transcripts	6
Verbatim quotes extracted and stored in Knowledge Base	34
Open questions flagged for next research round	4
Hours spent on synthesis vs. prior manual process	2

After 8 DTC churn interviews and 2 retail buyer calls in April, the founder ran the cross-interview synthesis prompt. Starch surfaced that 6 of 8 churned DTC customers mentioned the same thing: they bought once as a 'try it' but didn't understand the subscription option existed. None mentioned price. The two retail buyers, separately, both flagged that the 4oz SKU was getting outsold by competitor 6oz packs at the same price point. The founder then ran the PostHog cross-validation prompt — PostHog data (queried live from Starch's integration catalog) confirmed that 71% of single-order churn happened before customers ever reached the subscription upsell email, which goes out on day 14. The synthesis: the churn problem is a day-1 education problem, not a price problem, and the retail problem is a pack-size problem, not a placement problem. Two decisions, grounded in 10 interviews and real funnel data, made in an afternoon instead of three weeks of spreadsheet wrangling. The founder dropped the output into Presentation Agent and had a 12-slide deck for the board ready before dinner.

Measurement

How you'll know it's working

Interview-to-insight cycle time (days from last interview to synthesis doc delivered to team)

Theme recurrence rate (how many themes show up in 3+ consecutive interview batches — signals a real pattern vs. noise)

Quote retrieval time before key meetings (seconds to find relevant verbatims on a specific topic)

Interview coverage ratio (what percentage of churned customers, new retail accounts, and repeat DTC buyers were talked to this quarter)

Decision traceability (how many product, packaging, or go-to-market decisions made this quarter have a linked interview insight vs. were made on gut)

Comparison

What this replaces

The other ways teams handle this today, and how the Starch version compares.

Dovetail or Aurelius

Purpose-built for UX research teams with $200-400/month price tags and workflows designed for researchers, not a CPG founder who needs synthesis connected to their growth data and board deck in the same tool.

Otter.ai + manual Google Docs tagging

Otter transcribes fine, but the synthesis is still manual — you're copying quotes into a doc, tagging by hand, and the insight never connects to your PostHog data or your next pitch deck.

Notion AI

Good for organizing notes already in Notion, but won't cross-validate what customers say against live PostHog funnel data, won't generate a polished deck, and won't run a monthly synthesis automation without you triggering it manually.

Hiring a part-time researcher or growth marketer

A $3,000/month contractor can do this well, but you're still the bottleneck for context — they don't know your co-packer constraints, your deduction disputes, or why the 4oz SKU matters more than it looks. Starch has the context because it's connected to all your data.

On Starch RECOMMENDED

One platform — meeting notes, knowledge management, growth analyst all running on connected data. Setup in plain English; numbers stay current via scheduled syncs and live agent queries.

Try it on Starch →

FAQ

Frequently asked questions

Can Starch transcribe calls I've already recorded, or only calls going forward?

You can upload existing recordings to Meeting Notes and Starch will transcribe and summarize them. If your recordings are sitting in Zoom or Google Drive, connect Google Drive from Starch's integration catalog and point the agent at the folder — it can process the backlog.

My interview notes are a mess — some in Notion, some in Google Docs, some just in my head. Does Starch need clean data to start?

No. Starch connects directly to Notion via scheduled sync and can reach Google Docs from Starch's integration catalog. Start by connecting both; the Knowledge Management app will pull in what exists and the AI search works even on unstructured notes. The synthesis will be imperfect at first and get sharper as you add structured transcripts going forward.

Will this work for retailer buyer calls, not just consumer interviews?

Yes. The tagging and synthesis prompts work for any call type. In fact, one of the most useful things Starch can do is run a comparison: 'What do DTC customers say about our 4oz SKU vs. what retail buyers say?' — that cross-segment view is exactly what's hard to do manually when your notes are scattered.

Is my interview data secure? These calls contain candid customer opinions I don't want leaking.

Starch is not SOC 2 Type II certified today — that's the honest answer. If your legal or partnership agreements require SOC 2 compliance for storing customer interview data, that's worth flagging. For most early-stage CPG brands running 5-20 interviews a quarter, it's not a blocker, but you should make the call with that information.

Can I share the synthesized research with my co-founder or an advisor without giving them access to Starch?

Yes. Presentation Agent exports to PDF, PowerPoint, or a shareable link. The Knowledge Management synthesis docs can be exported too. You don't need to give anyone else a Starch login to get value from the research outputs.

What if I only do 2-3 customer calls a month? Is this overkill?

Two to three calls a month is enough to build a real insight archive over time — the value compounds. Even at that cadence, the retrieval piece alone is worth it: being able to search 'what did any buyer say about our MSRP' in 10 seconds before a pitch is different from hoping you remember.