Product News

Product-Led Growth With Claude: Turning User Events into Customer Insights

Julie Beynon

27 January 2025

We're thrilled to announce that our AI Columns now support Claude, Anthropic's powerful AI assistant. This upgrade takes our event analysis capabilities to the next level, making our insights even more accurate and actionable. But before we dive into that, let's walk through how we're transforming product data into gold…

The Problem with Event Stream Data (aka Why Everyone Was Drowning)

Imagine this: your product is busy collecting every click, swipe, and interaction. Sounds great, right? Except now you're buried under a mountain of events in your data warehouse, and your sales and customer success teams are starting to look like they'd rather do anything else than dive into yet another dashboard.

Been there, felt that pain. At Census, we were swimming in data about models created, syncs configured, and destinations added. But turning that into something meaningful? That was the fun part.

Turning Data Overload into Insights People Actually Use

Step 1: Organize the Essentials

First, we had to bring order to the flood of events. For us, that means capturing moments like:

Creating a new model
Setting up a sync
Adding a destination
And many, many more.

For you, it might be tracking cart additions, purchase patterns, or repeated visits to a particular product page (we're looking at you, undecided shopper 👀).

Step 2: Turning Customer Events into AI-Ready Data

Ever try explaining 90 days of customer activity to Claude or GPT? Yeah, we need to make that digestible. Here's how we do it at Census:

WITH source AS (

SELECT

account_id,

event_name,

happened_at,

unique_id

FROM

census_on_census.production.product_usage_summary

WHERE

happened_at >= CURRENT_DATE - INTERVAL '90 days'

)

SELECT

account_id,

ARRAY_AGG(OBJECT_CONSTRUCT(

'event_name', event_name,

'unique_id', unique_id,

'happened_at', happened_at

)) AS events

FROM

source

GROUP BY

account_id

Copy

This SQL magic bundles up all customer events into a neat JSON array, with one row per account capturing recent activity. This gets the data neatly organized and ready for AI analysis.

Step 3: Bring in the AI Magic ✨ (Now with Claude!)

Our AI Columns prompt transforms raw JSON events into a structured report that actually drives action. Here's what we look for:

Negative Events: failed, deleted, paused
Exploratory Events: clicked
Positive Events: created, configured

The prompt analyzes these events across six key areas:

Overview of Last 30 Days
Growth & Potential Issues
Feature Usage Breakdown
Service Connection Type Concerns
Suggested Next Steps
Sales Talking Points

Summarize the product usage trends based on available event types, focusing on key activities and changes over the last 30 days compared to the previous 30 days. Highlight significant positive, exploratory, and negative actions with a focus on actionable insights for customer engagement and issue resolution.

Event Types:

• Negative Events contain: failed, deleted, paused

• Exploratory Events contain:: clicked

• Positive Events contain: created, configured

Data Sources:

•

Guidelines:

1. Overview of Last 30 Days:

• Briefly outline overall product usage, with a focus on significant positive, negative, and exploratory actions, compared to the previous 30 days.

2. Growth & Potential Issues:

• Identify growth signals (e.g., increases in positive events like new destination creation) and potential concerns (e.g., a rise in negative events like sync deletion) without repeating details from other sections.

3. Feature Usage Breakdown:

• Summarize engagement with key actions such as sync configuration, model creation, and exploratory clicks, highlighting notable trends.

4. Service Connection Type Concerns:

• Focus on any issues by service type related to negative actions like failures or deletions.

5. Suggested Next Steps (most detailed):

• Outline actionable next steps for Customer Success to address negative events, promote positive actions, and encourage further exploratory engagement.

6. Sales Talking Points (most detailed):

• Provide talking points for addressing any negative usage trends, reinforcing the benefits of positive actions, and encouraging deeper product exploration. Use relevant case studies or success stories where applicable.

Example Output Structure:

Overview of Last 30 Days:

<ul>

<li>Key Activity: Active model creation and new sync configurations, with some increases in negative events like sync deletion.</li>

</ul>

Growth Signals & Potential Issues:

<ul>

<li>🟢 Growth: 20% increase in new source creation, indicating stronger product adoption.</li>

<li>🟡 Concern: 10% rise in sync deletions, indicating potential disengagement.</li>

</ul>

Feature Usage Breakdown:

<ul>

<li>📄 Sync Configured: Increased configuration attempts.</li>

<li>📄 Model Creation: Consistent activity in new model creation.</li>

<li>📄 Negative Events: Sync deletions and pauses have increased.</li>

</ul>

Service Connection Type Concerns:

<ul>

<li>⚠️ Investigate causes for increased sync deletions and workspace deletions.</li>

</ul>

Suggested Next Steps:

<ul>

<li>🛠️ Address the increase in sync deletions by offering troubleshooting resources.</li>

<li>📧 Schedule a follow-up to discuss setup improvements and encourage continued feature adoption.</li>

</ul>

Talking Points:

<ul>

<li>📊 Emphasize the value of new sync configurations and model creation.</li>

<li>🔍 Address any negative trends with proactive support, using success stories as applicable.</li>

</ul>

<div>
Summarize the product usage trends based on available event types, focusing on key activities and changes over the last 30 days compared to the previous 30 days. Highlight significant positive, exploratory, and negative actions with a focus on actionable insights for customer engagement and issue resolution.
Event Types:
 • Negative Events contain: failed, deleted, paused
 • Exploratory Events contain:: clicked
 • Positive Events contain: created, configured
Data Sources:
 • 
Guidelines:
 1. Overview of Last 30 Days:
 • Briefly outline overall product usage, with a focus on significant positive, negative, and exploratory actions, compared to the previous 30 days.
 2. Growth & Potential Issues:
 • Identify growth signals (e.g., increases in positive events like new destination creation) and potential concerns (e.g., a rise in negative events like sync deletion) without repeating details from other sections.
 3. Feature Usage Breakdown:
 • Summarize engagement with key actions such as sync configuration, model creation, and exploratory clicks, highlighting notable trends.
 4. Service Connection Type Concerns:
 • Focus on any issues by service type related to negative actions like failures or deletions.
 5. Suggested Next Steps (most detailed):
 • Outline actionable next steps for Customer Success to address negative events, promote positive actions, and encourage further exploratory engagement.
 6. Sales Talking Points (most detailed):
 • Provide talking points for addressing any negative usage trends, reinforcing the benefits of positive actions, and encouraging deeper product exploration. Use relevant case studies or success stories where applicable.
Example Output Structure:
Overview of Last 30 Days:
<ul>
 <li>Key Activity: Active model creation and new sync configurations, with some increases in negative events like sync deletion.</li>
</ul>
Growth Signals & Potential Issues:
<ul>
 <li>🟢 Growth: 20% increase in new source creation, indicating stronger product adoption.</li>
 <li>🟡 Concern: 10% rise in sync deletions, indicating potential disengagement.</li>
</ul>
Feature Usage Breakdown:
<ul>
 <li>📄 Sync Configured: Increased configuration attempts.</li>
 <li>📄 Model Creation: Consistent activity in new model creation.</li>
 <li>📄 Negative Events: Sync deletions and pauses have increased.</li>
</ul>
Service Connection Type Concerns:
<ul>
 <li>⚠️ Investigate causes for increased sync deletions and workspace deletions.</li>
</ul>
Suggested Next Steps:
<ul>
 <li>🛠️ Address the increase in sync deletions by offering troubleshooting resources.</li>
 <li>📧 Schedule a follow-up to discuss setup improvements and encourage continued feature adoption.</li>
</ul>
Talking Points:
<ul>
 <li>📊 Emphasize the value of new sync configurations and model creation.</li>
 <li>🔍 Address any negative trends with proactive support, using success stories as applicable.</li>
</ul>
</div>

Copy

From Prompt to Action: The Output

Here's what our teams actually see (now with enhanced insights thanks to Claude's capabilities):

Product Usage Summary

Overview of Last 30 Days:

Key Activity: Active model creation (+15%), new sync configs (+10%)
Warning Signs: Workspace deletions up 20% 🚨

Growth Signals & Potential Issues:

🟢 Growth: 15% increase in model creation

🟡 Concern: 20% rise in workspace deletions

Feature Usage & Next Steps:

📄 Sync Activity: Configs up 10%

🛠️ Action Needed: Reach out to accounts with workspace deletions

📧 Support Focus: Create targeted troubleshooting guides

Copy

This structured format hits all the key points our teams need: what's happening, why it matters, and what to do about it. With Claude's support, we're now able to generate even more nuanced insights and contextual recommendations. Best part? It lands directly in your CRM, ready for action. No data science degree required!

The Impact (aka Why We're Obsessed With This)

Let's get real - we built this because flying blind with customer data just isn't cute anymore. This system, now enhanced with Claude's capabilities, has transformed how our teams work:

Sales has actual context for every call
CS spots fires before they start
Account managers look like mind readers
Everyone's working from the same playbook

Like any powerful tool, we're continuously refining and enhancing our AI Columns based on real-world feedback. Every update brings new capabilities and insights that help teams work smarter, not harder.

Pro tip: Start with a focused pilot – perhaps tracking feature adoption across your power users. Within weeks, you'll have insights that would have taken months to surface manually.

Want to see it in action? Book a demo or sign up for a free trial and we'll show you how companies like yours are using AI Columns to drive growth and retention. Let's make your product data work harder for you!