> AI agents: this is one page from PostHog's docs. Full index of Markdown docs for LLMs: https://posthog.com/llms.txt

# Which leads deserve the sales team's time? – Context Warehouse pocket guide

[](/pocket-guides.md)Aa

[](/pocket-guides/context-warehouse/acquisition-channels-retention.md)[](/pocket-guides/context-warehouse/feature-revenue-impact.md)

# Which leads deserve the sales team's time? – Context Warehouse pocket guide

A CRM full of leads doesn't tell sales who to call first; product usage does. This Skill scores CRM leads by recent product usage and surfaces the warmest ones sales hasn't yet reached: a product-qualified-lead list.

## Example data and how to read it

The Skill looks for patterns like Fig. 1: Priya was seen 6 hours ago and is still listed as a trial, not yet a marketing-qualified lead: exactly the kind of engaged, uncontacted account this list exists to surface before someone else notices.

| Lead | Score | Last seen | Stage |
| --- | --- | --- | --- |
| [jane@acme.com](mailto:jane@acme.com) | 34 | 2 days ago | MQL |
| [sam@northwind.io](mailto:sam@northwind.io) | 29 | 1 day ago | MQL |
| [priya@globex.com](mailto:priya@globex.com) | 26 | 6 hours ago | Trial |

Fig. 1 – Example data: leads scored by product usage, sorted by who's warmest right now.

## Doing this by hand

1.  **Connect [HubSpot](/docs/cdp/sources/hubspot.md) or [Salesforce](/docs/cdp/sources/salesforce.md)** via [Data pipeline > Sources](/docs/data-warehouse/sources.md).
2.  **Define your key [events](/docs/data/events.md)** under Data management > Events: define what signals real intent for you.
3.  **For a ranked list**, check each lead's activity in the [Persons](/docs/data/persons.md) explorer, or use the [SQL editor](/docs/data-warehouse/sql.md), or ask PostHog AI to score every lead in one query.

## Get an agent to do the work

Copy the Skill below and add it to your agent. The agent defines, with you, the handful of events that signal real intent. Then it scores each lead by recency and frequency of those events, weighted however you like. It joins the score to CRM lead records on email and filters to leads not yet marked contacted. The scoring stays transparent on purpose so that everyone using it can see where it comes from.

This ranks individual leads to action right now. For analyzing which onboarding *steps* drive trial-to-paid conversion in aggregate instead, that's [Onboarding conversion](/pocket-guides/context-warehouse/onboarding-conversion.md).

posthog-lead-scoring/SKILL.md

```markdown
# Which leads deserve the sales team's time?
**Question:** Which leads are most product-engaged — especially ones sales hasn't contacted yet?
**For:** Sales & Growth · **Difficulty:** Intermediate · **Shape:** a scored query
**Data sources:** PostHog events (key product actions) + HubSpot / Salesforce (lead records)
## What this produces
A saved PostHog insight scoring CRM leads by recent product usage and highlighting the most engaged ones that
sales hasn't yet reached — a product-qualified-lead list.
## Workflow
First read `references/posthog-workflow.md` for the shared setup: confirm the PostHog MCP is connected, ensure the
CRM source (HubSpot or Salesforce) exists (secure connect-link flow if not), and learn the real schema. Then:
### 1. Identify the pieces in this project
- **Key product actions.** With the user, define the handful of events that signal buying intent / real usage
  (`event-definitions-list`). Assign each a weight, or keep it simple (count of key events, recency + frequency).
- **Lead records.** From the CRM: lead/contact rows with email, owner, and a "contacted"/lifecycle field so you can
  tell who sales has already worked. Confirm the field that means "not yet contacted".
- **Join key.** CRM lead ↔ PostHog person by `lower(email)` (see join gotchas in the shared reference).
### 2. Build and validate the query
Adapt names and validate with `query-run`. Produce one scored row per lead, filtered/sorted to surface warm,
uncontacted leads.
```sql
-- Score leads by recent product usage; surface the warmest uncontacted ones.
-- Adapt: key events + weights, the CRM table/fields, and the "contacted" flag.
WITH usage_score AS (
    SELECT
        lower(person.properties.email) AS email,
        sum(multiIf(event = 'created_project', 5,
                    event = 'invited_teammate', 4,
                    event = 'ran_query', 2, 1)) AS score,
        max(timestamp) AS last_seen
    FROM events
    WHERE timestamp >= now() - INTERVAL 30 DAY
      AND event IN ('created_project', 'invited_teammate', 'ran_query', 'viewed_pricing')
    GROUP BY lower(person.properties.email)
)
SELECT
    l.email,
    l.lead_owner,
    us.score,
    us.last_seen,
    l.lifecycle_stage
FROM hubspot_contacts AS l
INNER JOIN usage_score AS us ON lower(l.email) = us.email
WHERE l.lifecycle_stage NOT IN ('customer')          -- still a lead
  AND (l.last_contacted IS NULL OR l.last_contacted < now() - INTERVAL 30 DAY)  -- "not contacted yet"
ORDER BY us.score DESC
LIMIT 100
```
Keep the scoring transparent and let the user tune weights — a lead score sales doesn't understand won't get used.
Note that this ranks by product signal only; a real PQL model may blend in firmographics.
### 3. Save the insight
Save as a SQL/HogQL table insight named "Product-qualified leads (uncontacted)", documenting the scoring weights
and window. Return the URL and name the top few leads to call today.
## Self-driving development (offer this)
With leads scored by real product usage, the user can catch the warmest ones as they heat up. Offer to help set up
an alert or a cohort that flags a lead when its score crosses a threshold — filling the pipeline while they sleep.
(Pushing the score back into the CRM is a separate reverse-ETL step; offer to explain it.)
```

Show full example

Fig. 2 – The Skill itself, copy and paste this into your agent to answer the question.

See also: [Onboarding conversion](/pocket-guides/context-warehouse/onboarding-conversion.md) · [Features & revenue](/pocket-guides/context-warehouse/features-drive-revenue.md)

[‹ Which channels bring customers who stick?](/pocket-guides/context-warehouse/acquisition-channels-retention.md)[All guides](/pocket-guides.md)p. 8 of 15[Did the feature we shipped move revenue? ›](/pocket-guides/context-warehouse/feature-revenue-impact.md)