Continuous Discovery with Koji MCP — Always-On Research Pipeline

What Is Continuous Discovery?

Continuous discovery is the practice of conducting ongoing customer research rather than one-off studies. Instead of doing "a research sprint" every quarter, you make research a weekly habit — small, frequent touchpoints that keep your product decisions grounded in real user evidence.

Koji MCP makes continuous discovery practical by removing the friction. No switching between tools, no manual analysis, no building slide decks. You talk to Claude, and it handles the rest.

---

The Continuous Discovery Framework

Teresa Torres popularized the continuous discovery framework with three core activities:

1. Weekly customer touchpoints — Talk to customers every week, even briefly
2. Map the opportunity space — Organize what you learn into pain points, desires, and needs
3. Test assumptions — Validate your biggest assumptions before building

Koji MCP automates the overhead of all three.

---

Setting Up Your Continuous Pipeline

Step 1: Create a Long-Running Study

Instead of creating a new study for every question, create one "always-on" study for your product area:

"Create a Mom Test study called 'Continuous Product Feedback — Q1 2025' about understanding ongoing user pain points and workflow challenges. Keep it broad enough for general product discovery."

This becomes your persistent research channel. You can always create focused studies for specific questions, but this evergreen study collects ambient feedback.

Step 2: Customize for Your Brand

"Customize the landing page: headline 'Share Your Experience', description 'Help us build a better product by sharing your honest feedback in a quick AI-powered conversation', accent color #2563EB. Set the duration badge to '10 minutes' and add a custom badge with a heart icon that says 'Your feedback shapes our roadmap'."

Step 3: Set Up a Lead Form

If you want to segment feedback by user type:

"Configure a lead form with fields: name (text, required), email (email, required), role (select with options: Product Manager, Engineer, Designer, Executive, Other), and company size (select: 1-10, 11-50, 51-200, 201-1000, 1000+)."

Step 4: Distribute the Link

Embed the interview link in:

• Your product's feedback menu
• Post-onboarding email sequences
• Customer success check-in emails
• NPS follow-up flows
• Slack community channels

"Set the interview slug to 'product-feedback' so the URL is clean: koji.so/i/product-feedback"

---

The Weekly Cadence

Monday Morning: Review New Interviews

Start your week with a research check-in:

"Show me all completed interviews from my continuous feedback study in the last 7 days. What are the main themes?"

Claude pulls recent interviews, summarizes key themes, and highlights anything unusual. This takes 2 minutes.

Wednesday: Deep Dive Into Signals

If Monday's review surfaces something interesting:

"Show me transcripts from interviews that mentioned 'onboarding' in the themes. What are people struggling with?"

Claude finds relevant interviews and synthesizes the pain points across them.

Friday: Update the Research Record

If you have accumulated enough new interviews (5+):

"Generate an updated report from my continuous feedback study. How have the themes changed since the last report?"

Claude generates a report you can share with your team in the weekly product sync.

---

Advanced Patterns

Parallel Studies for Specific Bets

Run your continuous study alongside focused studies for specific product bets:

Continuous: "Product Feedback Q1 2025" (always collecting)
Focused: "Pricing page redesign validation" (time-boxed)
Focused: "Mobile app discovery" (time-boxed)

Use Claude to cross-reference:

"Compare the themes from my continuous feedback study with my pricing study. Is pricing showing up as a pain point in general feedback too?"

Monthly Synthesis

At the end of each month, do a comprehensive review:

"Get the study data from my continuous feedback study. Across all interviews this month, what are the top 5 pain points and which user segments mention them most?"

"Generate a report and publish it. I want to share a link with the leadership team showing our monthly research insights."

Importing Cohorts for Targeted Feedback

When you launch a new feature, target recent users:

"Import these 30 users who signed up this week to my continuous feedback study. I want to understand their first-week experience."

Each gets a personalized link. Their feedback automatically feeds into your continuous study.

---

Measuring Success

Track these metrics to know your continuous discovery pipeline is working:

---

Common Pitfalls

❌ Creating a new study every week

Continuous discovery works best with a persistent study. New studies = lost longitudinal data.

❌ Not segmenting respondents

Without a lead form or metadata, you cannot tell which insights come from power users vs. new users. Always collect at least role and tenure.

❌ Ignoring negative sentiment

Filter for negative sentiment interviews weekly. These contain the most actionable insights.

❌ Generating reports too frequently

Reports are snapshots. Generate them when you have meaningful new data (5+ new interviews), not daily.

---

Real-World Example: SaaS Onboarding Pipeline

A B2B SaaS team set up continuous discovery for their onboarding flow:

1. Study: "Onboarding Experience Feedback" with Customer Discovery methodology
2. Distribution: Link embedded in the post-onboarding email (day 3)
3. Lead form: Name, email, company size, which features they tried
4. Weekly review: Monday morning check-in via Claude
5. Monthly report: Published and shared in all-hands meeting

After 8 weeks (40+ interviews), they discovered that users were not struggling with the product — they were struggling with understanding the value proposition. This led to a repositioning of the onboarding copy, not a feature change.

That insight would not have surfaced in a one-off survey. It emerged from the pattern across dozens of open-ended conversations.

---

Next Steps

• PM Workflow Guide — Role-specific workflows
• Best Practices — Tips for effective MCP usage
• AI Interviews vs. Surveys — Why conversations beat forms