Ingestion
HubSpot contacts come in through the managed HubSpot connection, with the history of every property. A lead is graded as it stood when it became ready, so past leads are graded exactly as a new one would be.
- Connections
hubspot
Customer · an AI video company
Every demo request graded A to D, within hours.
An AI video company runs its lead scoring as code in its own SegmentStream workspace. A pipeline grades each demo request against standards written in plain language, and the sales team works the results in a dashboard with owners, statuses, notes and overrides.
- ~4 hours
- median time from demo request to grade
- A to D
- grades, with a probability for each
- 4
- grade standards, written as plain prose
- None
- models to retrain when a standard changes
From HubSpot contact to graded lead.
Grading
Two calls to a decision model through the evaluation connection. The first turns the lead's own words and company facts, with no personal fields, into a profile of fixed categories. The second matches that profile against four grade standards written as plain prose, and returns a grade with a probability for each. The standards were written from what a year of past demo requests and their outcomes showed.
- Connections
evaluation
Review
Grades stay in the warehouse. The ontology holds only what people decide: status, owner, notes, and a grade override with its reason, one review per HubSpot contact. Only invited members sign in to the dashboard, with an email code.
- Entities
lead_reviewnote- Actions
lead_review.setStatuslead_review.assignOwnerlead_review.addNotelead_review.overrideGrade- Roles
salesscoring
Changing the rules is an edit, not a project.
The standards are configuration, not a model to retrain. Editing the text changes how the next lead is graded.
- 1
Change the standards
The team edits the grade standards: plain paragraphs, one per grade.
- 2
Re-grade
One command re-grades a date range from the stored profiles, without asking the first question again.
- 3
Compare
Every row records which version of the standards graded it, so old and new grades can be compared.
People's decisions are records. Grades stay data.
Trimmed. The review joins the warehouse by contact id; it never stores the grade itself.
export const LeadReview = entity("lead_review", { label: "Lead review", description: "What the team decided about one scored lead: where follow-up stands, " + "who owns it, and a grade override when the model's grade is wrong.", properties: { contactId: p.text({ unique: true, description: "The HubSpot contact id." }), status: p.select(STATUSES, { default: "open", managed: true }), gradeOverride: p.select(GRADES, { optional: true, managed: true }), }, relationships: { owner: rel.one(User, { optional: true }) },}).actions(() => ({ overrideGrade: action({ description: "Replace the model's grade, with the reason." }),}));Enterprise pilot
Start a pilot.
Our engineers build the first version of your software in your own workspace, on your data, as code you own. Tell us what you run today and what you want to replace.
- One solution, on your data, in your own workspace
- First version live in two weeks
- Parity proven before anything is switched off
- Fixed scope and price, credited to your contract if you continue
Not ready for a pilot? Send the form anyway with your question. An engineer will reply, not a sales rep.
How the pilot worksPrefer to build it yourself? Join the waitlist.
Join the waitlist for self-serve:
Self-serve opens to the waitlist first.