How to build a lead scoring sync

What matters here

• Score fit and activity separately. A lead at 70 could be a perfect account that has done nothing, or a student who downloaded everything, and those need different follow-up.
• Only one process writes the score to the CRM. Every other system sends it signals, never a number.
• Activity should fade. A score that only ever goes up ranks a lead who went quiet a year ago above one who asked for a demo today.
• Crossing the threshold should start routing. The lead should reach a rep within seconds of qualifying.
• Stamp why a lead scored what it did. A rep who can see the reasons acts on the score, and one who can't ignores it.

Who this is for

You run marketing operations or revenue operations. You have a scoring model, probably in the marketing platform, and sales has stopped believing it. You want a score that means the same thing in every system, moves when the buyer moves, and sends qualified leads straight to a rep.

How it works in practice

What happens between a buyer doing something and a rep seeing the score change.

1. 1

   A signal arrives

   A form fill, a pricing page visit, a webinar attended, an intent surge on the account. Each one is recorded as an event with a date before any points are added up.

2. 2

   Fit is scored from the record

   Company size, industry, region and job title, from enrichment. Fit changes when the record changes, not when the buyer clicks.

3. 3

   Activity is scored from recent events

   Each event carries points that fade over time. A demo request this week counts in full; the same request six months ago counts for little.

4. 4

   One combined score is written to the CRM

   With the fit score, the activity score and the top three reasons beside it, so a rep sees why as well as how much.

5. 5

   Crossing the threshold sends the lead to routing

   The status changes to marketing qualified and the routing workflow runs at once. No overnight batch, no list for someone to export.

6. 6

   Dropping back below it is recorded too

   A lead that went quiet returns to nurture with a reason, rather than sitting in a rep's queue for a quarter.

Paste these into Claude Code or Codex with the Tray Headless plugin installed. Each stage runs on its own. The systems named in them are the worked example rather than a requirement, and every prompt says so.

Once per project, run /tray-workflows:set-workspace to pick the workspace these build in. Point it at a sandbox first.

1. 1

   Set up, and see what is actually connected

   Before anything is built, confirm the ground truth.

   Headless skills build-workflow

   Copy

   Use build-workflow. The systems in play are Marketo, Salesforce,
   Demandbase, Snowflake and Slack, or whatever we run in those seats.
   Before you plan anything, tell me which are already authenticated in
   the workspace. I expect Marketo, Salesforce and Slack.
   
   Then list every field in Salesforce and Marketo that currently holds a
   score, and which process writes to each one. I want to see the overlap
   before we change it.

   The list of current writers is usually the finding. Three processes writing one field is why the number moves for no reason.

2. 2

   Record signals as dated events

   Activity has to be stored with its date before it can fade.

   Headless skills build-workflow

   Copy

   Use build-workflow. Trigger on Marketo activity for: form fill, web
   page visit on /pricing or /demo, email click, webinar attended. Also
   take a daily Demandbase intent export for accounts in Salesforce.
   
   For each one, write a row to a Snowflake table with lead id, activity
   type, points and the date it happened. Do not add points to a running
   total anywhere. Points per type come from a table I can edit:
   
     demo request 40, pricing page 15, webinar attended 20,
     email click 3, intent surge on the account 10
   
   Webinar registration with no attendance scores zero.

3. 3

   Score fit and activity, then combine them

   The step that makes the score explainable.

   Headless skills build-workflow

   Copy

   Fit score, out of 100, from the Salesforce lead: employee count band,
   industry on our target list, country in a sold region, and title
   seniority. Recalculate when any of those fields change.
   
   Activity score, out of 100: sum of event points where each event
   halves in value every 30 days. Recalculate nightly and on every new
   event.
   
   Combined score: fit band A to D against activity band 1 to 4, from a
   grid I can edit. Write Fit_Score__c, Activity_Score__c, Lead_Grade__c
   and Score_Reasons__c (the top three events or fit factors) to the
   Salesforce lead. This workflow is the only writer of those fields.

4. 4

   Turn qualification into routing

   A score only pays off when something happens the moment it changes.

   Headless skills build-workflow tray-patterns

   Copy

   When a lead moves into grade A1, A2 or B1, set Status to Marketing
   Qualified, stamp MQL_Date__c, and call the lead routing workflow
   straight away with the lead id.
   
   When a qualified lead falls below B2 with no open task, set Status
   back to Nurture, stamp the reason, and tell the owner in Slack.
   
   If the same lead qualifies twice in one day, route it once.

5. 5

   Handle the four things that break it

   Each one has made a sales team stop reading the score.

   Headless skills tray-gotchas

   Copy

   Use tray-gotchas, then handle these explicitly:
   
   The marketing platform still writes its own score. Stop its sync to
   the Salesforce score field, and keep its number as a signal only.
   
   Enrichment arrives after the lead. Re-run the fit score when the
   enrichment fields fill in, not only at creation.
   
   Existing customers score as new leads. If the lead matches an account
   with a closed won opportunity, send it to the account owner instead of
   the new business queue.
   
   Bulk imports flood the threshold. Leads created by a list import score
   fit only until they show activity of their own.

6. 6

   Check the model against what closed

   The step that keeps the score honest a year from now.

   Copy

   Monthly, from Snowflake: for leads that qualified 90 or more days ago,
   show the rate that became opportunities and the rate that became
   closed won, by grade. Flag any grade where a lower grade converts
   better than a higher one.
   
   Then open the workflow in Tray Build so marketing operations can
   change the points table and the grid without a coding assistant.

   If grade B converts better than grade A for two months running, the model is wrong, and this report is how you find out before sales does.

What it connects to

Scoring reads signals from everywhere a buyer shows interest and writes one answer to the place reps work.

Named systems are the ones most teams run, not the only ones that work. Each is an authentication in your Tray workspace, referenced by name, so the workflow never holds a credential. Where we have a connector page, the name links to it.

Questions people ask

Further reading

Background on the same subject, for the case rather than the build.

• Lead scoring: why the score stops matching reality

Related guides