Lead scoring, sent back to HubSpot

For a team whose leads live in HubSpot and who have, or want, a model that says which ones to call first. You end up with a score on every contact in HubSpot, refreshed as leads arrive.

1. Connect HubSpot. A service key with read scopes brings in contacts, companies and deals. Every property comes with them, every hour or every 15 minutes.

   In the app: Connect HubSpot (Connect a source).

2. Get leads in fast. Optional. A HubSpot workflow posts each new lead to a webhook, so it lands in seconds. A webhook source gives you an address. Without it a new lead arrives at the next sync.

   In the app: Add a webhook (Connect a source).

3. Train the model. Train it in your own tools on the contacts and deals, and save it as a pickle. Download the tables from Warehouse, or query them from your laptop with a key.

   In the app: Open the warehouse (The warehouse).

4. Upload the pickle. Upload the pickle into the dataset the model will live in. A new upload under the same name is the next version, and every version is kept.

   In the app: Upload the pickle (The warehouse).

5. Score in a Python model. It reads the contacts, opens the pickle, and writes one score per contact. Name the pickle in the model's files and read hubspot.contacts. Set it to run after its sources update, so a new lead is scored when it lands.

   file("lead_model")

   In the app: Start a Python model (Models).

6. Send the scores back. The scores go to HubSpot contacts, matched on email, as a property on each one. The property is made for you if it does not exist yet.

   In the app: Add HubSpot as a destination (Sending data out).

7. Watch it. The destination lists each send. An alert tells you if scoring stops. Set the alert on how many contacts were scored today.

   In the app: Make an alert (How Overscore tells you).

What you'll see when it's working

A new lead in HubSpot has a score within minutes, ready for workflows and views.

All playbooks