Score leads into explainable bands

A visible Formula combines fit, intent, and engagement into one lead score, writes it back to Attio, and routes each band to the right action.

A visible Formula combines fit, intent, and engagement into one lead score, writes it back to Attio, and routes each band to the right action. Here is the flow in plain terms. A product or analytics event arrives on a Webhook received trigger. Parse JSON unpacks the payload, the Formula reduces the three signals to a single number, and Switch reads the band. Update record writes the result back to Attio, and the A band flows through Round robin and Enroll in sequence. The score is plain arithmetic, not an opaque model output. That means a sales development rep can open the run and read exactly why a lead is an A.

Difficulty	Intermediate
Time to build	About 15 minutes
Trigger	Webhook received, or Attribute value changed when the signals already live in Attio
Blocks used	Parse JSON, Formula, Switch, Update record, Round robin, Enroll in sequence
Credit cost per run	1 credit on the C band, up to 3 credits on the A band
Plan required	Plus or higher

Prerequisites

• Attio Plus plan or higher, which runs the new Workflows engine.
• A Person object (or Company) with three attributes already created. You need lead_score as a number, and lead_band as a single-select with options A, B, and C. You also need a matching key such as email or domain.
• A source for the event payload. The webhook path needs a product-analytics or reverse-ETL tool that can POST JSON to Attio’s generated webhook URL. Reverse-ETL means extract, transform, load. The attribute path needs the signal fields to already be writing to the record.
• An active sequence to enroll A-band leads into, and a sales team configured for round-robin assignment.
• Workspace member access with permission to publish workflows. The webhook source is the only external credential you need. No API keys are required inside Attio.

How it works

The trigger fires on one event. Parse JSON turns the raw payload into typed variables. That means each value carries its proper type, such as a number rather than text. The Formula then reduces fit, intent, and engagement to one number on a 0 to 100 scale. Switch reads the resulting band and sends the run down exactly one branch.

Every branch writes the score and band back to the record, so the value is durable and filterable. The A band also picks an owner and starts outreach. The C band stops after the write. Each step passes its output to the next through the engine’s variables, which the Variables & data passing concept explains in full.

1. Webhook received fires on the inbound event, or Attribute value changed fires on a signal field.
2. Parse JSON extracts fit, intent, and engagement.
3. Formula computes the combined lead_score.
4. Switch branches on the score band.
5. Update record writes lead_score and lead_band on every branch.
6. On the A band only, Round robin assigns an owner and Enroll in sequence starts outreach.

Build steps

1. Add the trigger

   Create a new workflow and add the Webhook received trigger. Copy the generated webhook URL into your analytics tool so it can POST events to the workflow. When your signals already live in Attio, swap in the Attribute value changed trigger instead. Scope it to the Person object and the signal attribute.

   Field	Value
   Trigger	Webhook received
   Object context	Person
   Sample payload key email	[e-mail]
   Sample payload key company_domain	acme.com

2. Parse the payload

   Add the Parse JSON block. Point its Source field at the trigger body, then define a schema with the three signal inputs and the matching key. Mark anything your source may omit as optional, so a missing field does not fail the run.

   Field	Value
   Source	{{ trigger.body }}
   Schema field	email (string, required)
   Schema field	fit (number, 0–40)
   Schema field	intent (number, 0–40)
   Schema field	engagement (number, 0–20)

3. Compute the score with a visible Formula

   Add the Formula block. Keep the expression readable so reps trust the result. Transparency is the entire point of this recipe, unlike a black-box machine-learning score where the math is hidden. Each component is capped, so the total lands on a clean 0 to 100 scale.

   Field	Value
   Output name	lead_score
   Expression	min(fit, 40) + min(intent, 40) + min(engagement, 20)
   Output type	Number
   Example result	82 (fit 38, intent 32, engagement 12)

4. Branch on the band with Switch

   Add the Switch block and create one branch per band, reading the score from the Formula step. Order matters here. Switch takes the first matching branch, so put the highest threshold first.

   Field	Value
   Input	{{ formula.lead_score }}
   Branch A	lead_score >= 75
   Branch B	lead_score >= 50
   Branch C (default)	Everything else

5. Write the score and band on every branch

   On each Switch branch, add an Update record block targeting the matched Person. Write both the numeric score and the band, so the value is filterable and visible in the record. The band string differs per branch.

   Field	Value
   Record	{{ trigger.record }}, matched by email
   lead_score	{{ formula.lead_score }}
   lead_band	A on Branch A, B on Branch B, C on Branch C

6. Route the A band

   On Branch A only, after Update record, add a Round robin block to assign an owner from your sales pool. Then add an Enroll in sequence block to start outreach. Branch B adds the lead to a nurture sequence or list. Branch C ends after the score write.

   Field	Value
   Round robin pool	Account executives, EMEA
   Assignment target	Owner on the matched record
   Sequence	Proposal outbound, A leads
   Enrolled record	{{ trigger.record }}

Variable mapping

Variable	Source block	Field	Example value
email	Parse JSON	email	[e-mail]
fit	Parse JSON	fit	38
intent	Parse JSON	intent	32
engagement	Parse JSON	engagement	12
lead_score	Formula	lead_score	82
lead_band	Switch	Matched branch	A
assigned_owner	Round robin	Owner	Priya Sharma

Credit cost worked example

Parse JSON, Formula, and Switch are free logic blocks, so they cost nothing no matter how often they run. Only the data writes and the sequence enrollment consume credits.

Branch	Billable blocks	Credits per run
A	Update record + Round robin¹ + Enroll in sequence	3
B	Update record + Enroll in nurture sequence	2
C	Update record	1

¹ Round robin is a free workspace block. The 3 credits on the A branch come from the Update record and Enroll in sequence writes.

Say you process 2,000 events a month with a typical split of 10 percent A, 30 percent B, and 60 percent C. The monthly cost is (200 × 3) + (600 × 2) + (1,200 × 1) = 3,000 credits. Compare that against your plan’s monthly credit allowance on Credits & pricing explained. Adjust the band thresholds if the A branch enrolls more leads than your team can work. None of this involves an AI block.

If you later add a Classify record step behind a Filter, sample about ten records and read the per-run cost in the run viewer. AI cost is variable and token-based, so it is not a fixed number you can read off a table.

Test plan

1. Open the trigger in the workflow editor and click the Run test button. For the webhook path, paste a sample payload with email, fit, intent, and engagement. For the attribute path, edit the signal field on a test Person.
2. Confirm Parse JSON outputs typed values, such as fit: 38, intent: 32, and engagement: 12, and not raw strings.
3. Confirm the Formula output lead_score reads 82 for those inputs.
4. Confirm Switch routes the run down Branch A, because 82 >= 75.
5. Open the matched record and confirm Update record wrote lead_score = 82 and lead_band = A.
6. Confirm Round robin assigned an owner and Enroll in sequence shows the lead enrolled.
7. Open the run in run history and read each block’s input and output to confirm the score is explainable end to end. The Runs, debugging & publishing concept shows where each value appears.
8. Fire one deliberately failing input. Send a payload where intent is the string "high" instead of a number. Parse JSON fails the schema and stops the run before any write. Confirm in run history that no credits were spent and no record was updated.

Failure modes

Symptom	Cause	Fix
The run stops at Parse JSON	A signal field arrived as a string or was missing	Mark optional fields optional in the schema and coerce numeric fields at the source.
Every lead lands in C	Formula inputs resolve to 0 because the variable names do not match the Parse JSON output	Align the Formula variable names with the exact Parse JSON output keys.
The score looks wrong but reps cannot tell why	The component caps are missing, so one signal dominates the total	Keep the min(...) caps visible in the expression and document the scale on lead_band.
A-band leads are enrolled but not assigned	The Round robin pool is empty or its members are inactive	Confirm the pool has active members, and run Round robin before Enroll in sequence on the branch.
Duplicate enrollments appear	The trigger re-fires for the same event	Add a free Filter that exits when lead_band is already A.

Variations

Separate fit and intent scores

Output two Formula values instead of one combined number. Name them fit_score and intent_score, and write both to the record. Switch on a two-by-two matrix, where high fit and high intent together earn an A. That way, reps can see whether a lead is a good fit, actively buying, or both. Everything else stays the same.

Recompute on a schedule

Swap the trigger for Recurring schedule. Add a Find records block to pull the leads to rescore. Then wrap the Formula, Switch, and Update record steps in a Loop. The loop multiplies cost by the number of records. Gate it with a Filter that rescores only leads whose signals changed since the last run.

See also

• Webhook lead intake with safe upsert
• Variables & data passing
• Runs, debugging & publishing
• Switch
• Formula

Source: Attio Help Center, Workflows. Last reviewed against the Attio Help Center, June 2026.