Lead Scoring: Administrator Configuration Guide

Overview

Lead Scoring ranks inbound leads and existing contacts on a scale from 0 to 100. It combines a starting score with positive and negative rules based on the lead source, form answers, contact completeness, replies, deals, and other approved activity. The result is assigned to a score band such as Cold, Nurture, Warm, or Hot.

Lead Scoring is designed for company administrators who want sales teams to see which leads deserve attention first. It can also provide a pre-assignment deterministic score to Advanced Lead Routing and can optionally use an LLM to enrich a contact-stage score.

After configuration, users can see the selected current score:

• in the Lead Score column on CRM > Contacts;
• on the Contact Conversation page, directly above Lead Status; and
• in administrator-only run logs and deterministic reports.

The score is dynamic. A new inbound lead can be scored before routing and again after it becomes a contact. Later configured events—such as a customer reply, deal creation, deal stage change, win, loss, or no-response timer—can request a recalculation. The integration records the event; your published profile decides whether it changes the score and by how many points.

Lead Scoring is fail-open for normal CRM operations. If scoring is disabled, unavailable, or unable to calculate a score, lead capture, queue processing, routing fallback, assignment, forms, replies, deals, and workflows continue normally.

When to use Lead Scoring

Use Lead Scoring when you want to:

• prioritize high-intent leads consistently;
• give different weight to Facebook, IndiaMART, Justdial, Engage, contact-form, API, or other supported inbound sources;
• score company-specific form answers without hard-coding field names;
• increase a score after engagement or deal progress;
• decrease a score after no response or a lost deal;
• apply different scoring policies to different sources, integrations, or forms;
• route new leads using a deterministic pre-assignment score; or
• test AI-assisted enrichment without immediately changing the score.

Before you begin

You need all of the following:

• a company plan that includes Lead Scoring and an active CRM licence;
• a company Admin account with permission to manage Lead Scoring;
• the relevant lead sources, forms, and integrations already connected;
• at least one published default profile or one published profile with a matching active binding.

For dynamic provider and form fields, first submit a controlled test lead through the relevant source. The field appears only after the lead reaches the central lead queue and field discovery completes.

For optional LLM enrichment, you also need:

• the company LLM switch enabled;
• a published profile in SHADOW or HYBRID mode;
• at least one approved and mapped field eligible for LLM use;
• sufficient AI credits; and
• available capacity under every configured LLM limit.

Navigation and permissions

Open Admin > Features Settings > Lead Scoring.

Only company administrators with Lead Scoring management permission can open the configuration, logs, and report pages. CRM users who can read a contact may see its score according to their normal CRM visibility scope, but they cannot configure scoring.

From the Lead Scoring page, use:

• Run and Job Logs for individual runs, events, jobs, and safe error codes;
• Deterministic Report for aggregate score statistics; and
• Setup Help and FAQ for the in-product summary.

How Lead Scoring works

Lead enters the central lead queue
  -> source and form context is verified
  -> eligible incoming fields are discovered
  -> one published profile is selected
  -> event-stage deterministic score is calculated
  -> optional score-based routing uses that event-stage score
  -> lead is assigned and becomes a CRM contact
  -> contact-stage deterministic score is calculated
  -> optional LLM enrichment runs if every gate is satisfied
  -> the selected current score appears in CRM
  -> configured replies, deal events, timers, or workflows request recalculation

The deterministic calculation is:

Base Score
  + matched ungrouped rule points
  + matched grouped rule points after group caps
  = unclamped score
  -> limited to 0–100
  -> assigned to the matching score band
  -> optionally adjusted in HYBRID mode
  -> limited to 0–100 again

Each enabled rule contributes its points once when its conditions match. A rule group can limit the combined positive or negative contribution of its rules. The final deterministic score can never be lower than 0 or higher than 100.

Configuration dependency map

Lead Scoring Status
|
+-- OFF
|   +-- normal lead capture, routing, assignment, CRM, deals and workflows continue
|   +-- no new field discovery, score calculation, behavior scoring or LLM requests
|   +-- score cards are hidden; configuration, history and pending jobs are retained
|
+-- ON
    |
    +-- Published Profile
    |   |
    |   +-- Default Profile -> fallback when no active binding matches
    |   |
    |   +-- Non-default Profile -> requires at least one active binding
    |       +-- exact form/provider object
    |       +-- integration
    |       +-- marketing source
    |       +-- broad source/module/service
    |
    +-- Deterministic Rules
    |   +-- system facts
    |   +-- approved direct profile fields
    |   +-- active reusable dimensions
    |   +-- rule groups and contribution caps
    |   +-- complete 0–100 score bands
    |
    +-- CRM Score Display
    |   +-- latest current score
    |   +-- selected published profile -> requires Published profile selection
    |   +-- hidden
    |
    +-- Routing Integration
    |   +-- requires Advanced Lead Routing
    |   +-- uses only completed event-stage deterministic facts
    |
    +-- Behavior Trigger Policies
    |   +-- require matching scoring rules to change the score
    |   +-- debounce controls when recalculation starts
    |   +-- lookback controls how long the signal remains active
    |   +-- expiry recalculates when a timed signal expires
    |
    +-- Optional LLM Scoring
        +-- company LLM switch ON
        +-- published profile mode SHADOW or HYBRID
        +-- approved field allows LLM use
        +-- profile mapping enables LLM use
        +-- privacy/type policy permits the field
        +-- AI credits and all resource limits are available
        +-- SHADOW -> stores insight but does not change the effective score
        +-- HYBRID -> applies a bounded adjustment only at/above Min Confidence

Setting precedence

Profile selection precedence

When more than one published profile could match a lead, selection is resolved in this order:

1. exact form or provider-form binding;
2. integration or provider-object binding;
3. marketing lead-source binding;
4. broad source, module, service, or source-ID binding;
5. the published default profile.

Within the same level, a higher priority wins. If priorities are equal, the binding with more populated selectors wins. Publishing rejects overlapping bindings at the same specificity and priority when they could select different profiles.

Field and dimension precedence

When multiple mapped fields provide a value for the same dimension, the lower mapping_priority value is considered first. If values differ:

• INVALID_CONFLICT stops scoring safely; or
• FIRST_NON_MISSING uses the first available value by mapping priority.

Rule and score precedence

• Conditions determine whether a rule matches.
• Each matched rule contributes once.
• Rule-group minimum and maximum caps are applied before the final score is limited to 0–100.
• Score bands label the resulting score; they do not add points.
• In SHADOW, the deterministic score remains effective.
• In HYBRID, a valid LLM adjustment can change the effective score only when confidence meets Min Confidence and only within Max Adjustment.
• Initial routing always uses the event-stage deterministic score, never the contact-stage or LLM-adjusted score.

Recommended configuration order

1. Enable Lead Scoring. This allows field discovery and scoring jobs to run.
2. Send controlled test leads. Use each source or form whose fields you plan to score.
3. Review incoming fields. Approve only fields with a confirmed type, privacy classification, and scoring purpose.
4. Create reusable dimensions when needed. Use dimensions for the same concept across differently named provider fields.
5. Create the first profile as the default. Keep Apply recommended starter rules enabled unless you intentionally want an empty draft.
6. Map approved fields. Save mappings before writing rules so the page can show current profile-field IDs.
7. Review the base score, rules, group caps, triggers, and bands. Adapt starter values to your business.
8. Configure bindings. A non-default profile cannot be published without at least one active binding.
9. Run a synthetic preview. Use only non-personal test values.
10. Validate and publish. Publishing freezes that version.
11. Choose CRM display and routing behavior. Enable routing integration only after Advanced Lead Routing is ready.
12. Optionally enable LLM enrichment. Start with SHADOW, monitor cost and quality, then consider HYBRID.
13. Verify with a new test lead. Check Contacts, Contact Conversation, Run and Job Logs, and the report.

Configuration guide

1. Company-level settings

Lead Scoring Status

Enable lead scoring for this company is the master switch.

• Enabled: supported intake events can discover fields, calculate scores, create behavior events, expose event-stage facts to routing, display scores in CRM, and schedule eligible LLM requests.
• Disabled (default): normal CRM activity continues, but no new scoring work is produced or processed. Existing profiles, scores, logs, and pending jobs are retained. Pending jobs can resume after the switch is enabled.

Turning the switch off does not delete or reset historical scores. CRM score cards are hidden while the feature is off.

Routing Integration

Allow advanced inbound routing rules to use the deterministic event-stage score controls whether Advanced Lead Routing can use three read-only scoring facts: routing score, routing band, and profile key.

• Enabled: a completed pre-assignment deterministic score can be used by Advanced Lead Routing rules.
• Disabled (default): routing follows its normal configured rules and fallback without score facts.

Requires: Lead Scoring enabled and Advanced Lead Routing enabled for the company.

If event scoring fails or no profile matches, routing continues with its normal fallback. LLM results and later contact activity never alter initial assignment.

CRM Score Display

Use Score to display to choose the current score users see on the Contacts list and Contact Conversation page.

Option	Behavior	When to use
Latest current score (recommended)	Shows the most recently evaluated eligible profile for each contact. This is the default.	Most companies with one profile or source-specific profiles.
A selected published profile	Shows only the current score produced by the chosen profile. Contacts without a score for that profile show no completed score.	When one profile is the official sales-priority score.
Do not show lead scores in CRM	Hides scores from both CRM screens without disabling calculations, logs, routing, or reports.	When administrators are testing scoring privately.

Published profile appears only when A selected published profile is chosen. It is required and accepts only an active profile with a published immutable version.

Optional LLM Scoring

Enable optional LLM enrichment is off by default. Enabling it does not create a score by itself. Deterministic contact-stage scoring must complete first.

Setting	Default	Valid range	What users observe
Daily request limit	100	1–10,000	Once reached, new eligible LLM operations are skipped for the rest of the day; deterministic scoring remains available.
Daily credit limit	200.00	0–1,000,000.00, up to 2 decimals	A request that would exceed the daily configured credit budget is skipped. A value of 0 prevents chargeable LLM requests.
Maximum queued jobs	50	1–1,000	When the number of pending, waiting, or running LLM jobs reaches this limit, additional eligible requests are skipped.
Maximum concurrent jobs	2	1–50	Limits how many LLM jobs for the company may run at the same time. Excess claimed work waits and retries.

The page shows the current AI-credit balance and current per-lead price. A provider request is not sent if the balance is below the required price. Successful LLM operations are billed using the current Lead Scoring LLM price. Provider failures are not charged. Both SHADOW and HYBRID can consume credits because both call the LLM.

2. Discovered incoming fields

Fields are discovered from supported, company-owned intake contexts. Current adapters include Facebook, IndiaMART, Justdial, contact forms, CRM custom forms, Web Lead, API, Engage, TradeIndia, Google Ads, Calendly, and inbound email when those sources use a supported intake path.

Discovery stores field metadata, not a sample value for administrators. Unknown fields never affect a published profile automatically.

Review Incoming Field

Setting	Options/default	Behavior and dependencies
Type Override	Use observed type (default), TEXT, NUMBER, BOOLEAN, DATE, SINGLE_SELECT, MULTI_SELECT	Changes the approved interpretation only when compatible with observed metadata. Do not choose NUMBER for IDs, phone numbers, postcodes, or values with significant leading zeros.
Privacy Classification	PUBLIC, BUSINESS (default selection), PERSONAL, SENSITIVE	May keep or raise the system privacy floor, never lower it. PERSONAL and SENSITIVE fields cannot be approved for deterministic or LLM scoring.
Allow deterministic scoring	On by default	Allows the approved revision to be mapped for deterministic rules, provided its type and privacy policy are eligible.
Allow LLM use (policy still applies)	Off by default	Allows later LLM use only for eligible PUBLIC/BUSINESS NUMBER, BOOLEAN, DATE, SINGLE_SELECT, or MULTI_SELECT fields. This switch alone does not send the field.
Resolve an observed type conflict	Off by default	Must be enabled when the field is in TYPE_CONFLICT state and you have intentionally selected the correct compatible type. It has no purpose for fields without a conflict.

Select Approve Revision to create a versioned approval. If the provider later changes the field identity, type, or extraction shape, review the new revision and publish an updated profile.

Hide and Block

• Hide changes the field to HIDDEN so it no longer remains an approved current field. Any draft mapping to it becomes stale and cannot be published until removed or reapproved.
• Block marks the field BLOCKED and removes the Review action. Block fields that must never be used for scoring. Existing published evidence remains historical, but future drafts cannot map the blocked field.

Credentials, passwords, tokens, OTPs, payment-card data, and other prohibited technical or secret fields are excluded from discovery.

3. Reusable dimensions

A dimension gives differently named fields one stable business meaning. For example, Facebook company_size, IndiaMART employee_count, and a custom form team_size can all map to COMPANY_SIZE if their approved types are identical.

Dimensions are optional and never add points by themselves.

Setting	Required/default	Rules and effect
Dimension Name	Required; maximum 160 characters	Human-readable name, such as Company Size.
Dimension Key	Required; 1–80 characters	Must start with a letter and contain only uppercase letters, numbers, and underscores. Use a stable key because rules reference it.
Description	Optional; maximum 500 characters	Explain the business meaning and expected values. This description may accompany an eligible LLM fact.
Canonical Type	Required; TEXT, NUMBER, BOOLEAN, DATE, SINGLE_SELECT, MULTI_SELECT	Must exactly match every approved field mapped to the dimension.
Privacy Floor	BUSINESS by default; PUBLIC, BUSINESS, PERSONAL, SENSITIVE	Must be at least as restrictive as every mapped field. PUBLIC is least restrictive; SENSITIVE is most restrictive.

New dimensions are created as Active. The current page lists dimensions but does not provide an edit control. Plan the key, type, and privacy floor carefully before use. Once a dimension is used by a published version, its identity and policy are treated as immutable.

4. Create a scoring profile

Select New Profile.

Setting	Default/validation	What it controls
Profile Name	Required; maximum 160 characters	Administrator-facing name shown in configuration, logs, reports, and CRM score details.
Profile Key	Required; 1–80 characters; starts with a letter; uppercase letters, numbers, underscores; unique	Stable profile identifier used in bindings, evidence, and optional routing rules. It cannot be reused.
Description	Optional; maximum 500 characters	Explains which leads and outcome the profile covers.
Base Score	40 in the UI; integer 0–100	Starting deterministic score before rules. The default 40 begins in the starter Nurture band.
Default profile	Off	Makes this profile the fallback when no more-specific binding matches. The first profile should normally be default. Only one active published default profile is allowed.
Apply recommended starter rules	On	Adds editable bands, groups, rules, and behavior windows. It does not publish, enable LLM, or guess company-specific fields or values.

Select Create Draft. The Manage Lead Scoring Draft page opens.

The recommended template adds:

• score bands: Cold/Junk 0–19, Low Intent 20–39, Nurture 40–59, Warm 60–79, Hot 80–100;
• up to +10 for contact completeness;
• +12 for a customer reply;
• -7 for no response after 24 hours;
• +8 for deal creation;
• +15 for a won deal; and
• -20 for a lost deal, subject to the configured group caps.

Review these values before publishing. They are a starting point, not a universal definition of a qualified lead.

5. Score and LLM Policy

Base Score

Base Score is an integer from 0 to 100. It applies to every evaluation using this profile version. A neutral starting score around 40–50 leaves room for both positive and negative rules.

LLM Mode

Mode	Effect	Credits	Recommended use
OFF	Deterministic scoring only. This is the default for a new profile.	None	Initial rollout or no AI requirement.
SHADOW	Runs eligible LLM enrichment and records its proposed adjustment, confidence, reasons, and optional safe summary, but the effective score remains deterministic.	Successful requests are billed.	Recommended first AI stage for quality and cost evaluation.
HYBRID	Applies a valid adjustment when confidence is at least Min Confidence.	Successful requests are billed, including low-confidence valid results.	Use only after SHADOW results are proven reliable.

LLM adjustment controls

Setting	Recommended/default	Range	Important behavior
Max Adjustment	10	0–100; must be at least 1 for SHADOW/HYBRID	The LLM cannot suggest an adjustment outside this positive or negative bound.
Min Confidence	0.75	0–1, up to 5 decimals	HYBRID applies an adjustment only at or above this value. SHADOW records confidence but never applies an adjustment.
Rubric Version	sales-intent-v1	1–60 letters, numbers, dots, underscores, or hyphens when LLM is enabled	Identifies the server-managed evaluation rubric. Keep the supplied value unless your deployment operator provides another version.

LLM Policy JSON

Select Apply Recommended LLM Settings to load a working bounded policy. The button changes the editor only; select Save Score and LLM Policy to store it.

Policy setting	Recommended	Allowed range/options	Effect
max_input_facts	30	1–100	Maximum eligible mapped facts in one LLM request.
max_fact_value_chars	300	1–1,000	Maximum characters allowed for one eligible fact value.
max_total_input_bytes	8,192	256–16,384	Maximum total encoded request size.
max_estimated_tokens	2,000	64–4,000	Maximum estimated input-token allowance.
max_free_text_fields	0	0–10	Maximum free-text fields if free text is allowed. Current field-approval policy does not approve normal TEXT fields for LLM use, so increasing this does not bypass field eligibility.
allow_business_free_text	0	0/Off or 1/On	Permits eligible BUSINESS free text only when field policy also permits it and max_free_text_fields is at least 1. It does not permit PERSONAL or SENSITIVE data.
allow_safe_summary	1	0/Off or 1/On	Allows the validated result to contain a bounded safe summary.
safe_summary_max_chars	240	0–500; at least 1 when summaries are on	Maximum summary length.
reason_code_allowlist	STRONG_INTENT, NEEDS_REVIEW	Up to 50 unique uppercase codes	The LLM result may return only listed reason codes.

Provider, model, endpoint, prompt body, credentials, and API keys are centrally managed and cannot be added to this policy.

6. Tenant Field Mappings

Mappings connect an approved incoming-field revision to this draft. Use Approved Field and Dimension IDs on the right side of the page; IDs are company- and version-specific and must not be copied from another company or an old draft.

Mapping setting	Default/options	Behavior
inbound_field_revision_id	Required positive approved revision ID	Selects the exact approved, current field revision. A later provider change can make the mapping stale.
dimension_id	0	Use 0 for a direct field, or an active compatible dimension ID to share meaning across sources.
display_label	Required; maximum 190 characters	Human-readable snapshot for this profile mapping.
transformation_policy.trim	1/On	Removes leading and trailing whitespace before comparison.
transformation_policy.case_mode	NONE; NONE, UPPER, LOWER	Optionally normalizes text case before rules evaluate it. Comparisons are otherwise exact and case-sensitive.
transformation_policy.sort_multi_values	0/Off	Sorts a multi-select value after normalization for stable representation.
missing_value_policy	NO_MATCH; NO_MATCH or FAIL_SAFE	NO_MATCH ignores a missing/invalid value and allows other facts to score. FAIL_SAFE stops that evaluation safely.
mapping_priority	10 in the example; integer 0–1,000,000	Lower values are considered first when several fields map to one dimension.
conflict_policy	FIRST_NON_MISSING in the example; FIRST_NON_MISSING or INVALID_CONFLICT	Selects the first value or safely rejects conflicting dimension values.
deterministic_enabled	1/On	Must remain on for an active draft mapping to pass publication validation. Remove an unused mapping instead of saving it disabled.
llm_enabled	0/Off	Allows the mapping to be considered for LLM input only when the approved revision is also LLM-eligible and all privacy/type/policy gates pass.

Select Load Tenant Field Example to start from the first compatible approved field, then replace values as needed. The example is not saved automatically. Select Save Field Mappings. The page reloads and shows generated Current profile field IDs for direct-field rules.

7. Score Bands

Every integer score from 0 through 100 must belong to exactly one band. Gaps and overlaps block publication.

Band setting	Validation	Behavior
band_key	Required unique uppercase identifier, maximum 80 characters	Stable band identifier. Common keys such as HOT, WARM, and COLD also receive matching CRM colors.
band_name	Required; maximum 160 characters	User-facing label shown beside the score.
color_key	Maximum 40 characters; DEFAULT if omitted	Controls the Contact Conversation tone. HOT/HIGH/SUCCESS/GREEN are high; WARM/MEDIUM/WARNING/AMBER/YELLOW are medium; COLD/LOW/DANGER/RED are low. Unknown values fall back to score thresholds.
minimum_score	Integer 0–100	First score included in the band.
maximum_score	Integer 0–100 and not below minimum	Last score included in the band.
priority	Integer 0–1,000,000; default 100	Stable ordering metadata. Range coverage determines the matching band.

If color_key is not recognized, the Contact Conversation card uses high at 80+, medium at 50–79, and low below 50. The Contacts list has dedicated badge colors for HOT, WARM, and COLD and a neutral style for other keys.

8. Behavior Trigger Policies

Trigger policies are optional. They decide which authoritative CRM events request contact-stage recalculation and how long their signals remain active. A trigger does not add points unless a rule references its signal or behavior count.

Trigger setting	Default/validation	Behavior
trigger_key	Required; one of LEAD_ASSIGNED, MESSAGE_REPLY_RECEIVED, DEAL_CREATED, DEAL_STAGE_CHANGED, DEAL_WON, DEAL_LOST; unique per profile	Selects the authoritative event.
signal_key	Required uppercase identifier, maximum 80 characters	Tenant-defined signal exposed to rules through active signal keys.
enabled	0/1	Disabled policies ignore that event.
lookback_seconds	0–31,536,000	How long the event remains active. 0 has no timed expiry. Positive values require REEVALUATE.
debounce_seconds	0–31,536,000 and less than lookback when lookback is positive	Delays application. A newer same-type event inside the window supersedes the older delayed job. For LEAD_ASSIGNED, use this as the no-response waiting period.
expiry_policy	NONE when lookback is 0; REEVALUATE when positive	REEVALUATE schedules another score calculation when the signal expires.
maximum_occurrences	1–100	Caps how many active events of this type contribute to behavior counts. It does not multiply a rule automatically; the rule still contributes once when matched.

The starter no-response setup delays LEAD_ASSIGNED for 24 hours, keeps the waiting signal active for 7 days, and checks that reply count remains zero. A reply received before the delayed check changes the behavior facts so the no-response rule does not match.

9. Source and Form Bindings

Bindings decide which published profile applies. A default profile may have no bindings. A non-default profile needs at least one active binding.

Binding setting	Default/validation	Behavior
binding_name	Required; maximum 160 characters	Administrator-facing description.
source_key	Optional supported source key	Matches the trusted inbound source, such as FACEBOOK, INDIAMART, JUSTDIAL, CONTACT_FORM, CUSTOM_FORM, ENGAGE, API, or WEBLEAD.
module_scope	Optional supported module	Narrows the source module. Usually matches the intended source family.
service	Optional; maximum 20 characters	Narrows the accepted service context. Leave empty unless you know the exact server-owned service value.
source_id	0 or a valid non-negative source ID	Narrows the operational source record.
marketing_lead_source_id	0 or an active company/global lead-source ID	Selects a CRM marketing source.
form_type	Empty, PUBLIC_CONTACT_FORM, CRM_CUSTOM_FORM, or FACEBOOK_FORM	Must be paired with form_reference.
form_reference	Empty or valid form reference, maximum 160 characters	For public/custom forms, use the numeric company form ID. For Facebook, use the synchronized provider form reference.
integration_id	0 or valid company integration ID	Used with exact provider objects such as a Facebook form.
provider_object_type	Empty or valid provider type, maximum 40 characters	For Facebook form bindings, use FACEBOOK_FORM.
provider_object_reference	Empty or exact provider reference, maximum 160 characters	Must match the form/provider reference and requires a valid integration.
priority	100; integer 0–1,000,000	Higher priority wins between matching bindings at the same specificity.
status	ACTIVE; ACTIVE or INACTIVE	Only active bindings can select a profile. Inactive bindings remain stored in the published version but have no selection effect.

For a Facebook form, all three form/integration values must agree: the synchronized integration ID, FACEBOOK_FORM, and the exact provider form reference.

10. Rule Groups and Rules

Rule groups

Groups cap the combined effect of related rules.

Group setting	Default/validation	Behavior
group_key	Required unique uppercase identifier, maximum 80 characters	Referenced by rules.
group_name	Required; maximum 160 characters	Human-readable group name.
minimum_contribution	Default -100; integer -100 through 0	Most negative combined contribution allowed for the group.
maximum_contribution	Default 100; integer 0 through 100	Most positive combined contribution allowed for the group.
priority	100; integer 0–1,000,000	Ordering metadata for stable configuration. Rule priority controls the order in which capped contributions are applied.

Example: three matched engagement rules total +37, but the group maximum is +25. Only +25 reaches the score.

Rules

Rule setting	Default/options	Behavior
enabled	1/On	Disabled rules are ignored and do not need to pass active-rule condition checks.
rule_type	FIELD_SIGNAL; SOURCE_WEIGHT, FIELD_SIGNAL, CUSTOM_RULE	Classification for administrators and audit. Matching behavior comes from conditions and points.
reason_code	Required unique uppercase identifier, maximum 80 characters	Stable explanation code recorded with matched factors.
rule_name	Required; maximum 160 characters	Human-readable explanation shown in factor details.
rule_description	Optional; maximum 500 characters	Explains the business intent.
group_key	Optional	Places the rule under an existing group cap. Blank rules are ungrouped.
match_mode	ALL; ALL or ANY	ALL requires every condition. ANY requires at least one.
points	Required integer -100 through 100	Added once when the rule matches; negative values subtract.
priority	100; integer 0–1,000,000	Lower values are evaluated first. This matters when a group cap is reached.

Rule conditions

Each enabled rule needs 1–50 conditions.

Condition setting	Options/default	Behavior
condition_source_type	SYSTEM_FACT, PROFILE_FIELD, DIMENSION	Chooses where the value comes from.
system_fact_key	Required for SYSTEM_FACT	Uses an allowlisted source, presence, or behavior fact. Must be blank for PROFILE_FIELD.
dimension_key	Required for DIMENSION	Uses an active dimension mapped in this profile. Its comparison type must equal the dimension type.
profile_field_id	Required positive current ID for PROFILE_FIELD; otherwise 0/blank	Targets one exact mapped field in this draft. IDs can change in a new draft, so use dimensions for cross-source concepts.
operator	See operator table below	Defines the comparison.
comparison_type	TEXT, NUMBER, BOOLEAN, DATE, SINGLE_SELECT, MULTI_SELECT	Normalizes both actual and configured values. DATE uses YYYY-MM-DD.
comparison_value	Required except EXISTS/NOT_EXISTS	Main target value. IN, NOT_IN, ANY_OF, and ALL_OF require a non-empty list of up to 200 values.
comparison_value_to	Required only for BETWEEN	Inclusive upper bound.
priority	100; integer 0–1,000,000	Stable condition order; it does not change ALL/ANY logic.

Supported operators

Operator	Meaning	Compatible types
EQUALS / NOT_EQUALS	Exact normalized equality or inequality	All scalar types; use list operators for MULTI_SELECT
CONTAINS / NOT_CONTAINS	Text contains or does not contain text	TEXT, SINGLE_SELECT
STARTS_WITH / ENDS_WITH	Text prefix or suffix	TEXT, SINGLE_SELECT
IN / NOT_IN	Scalar is or is not in the configured list	TEXT, NUMBER, BOOLEAN, DATE, SINGLE_SELECT
GREATER_THAN / GREATER_THAN_OR_EQUAL	Numeric or date comparison	NUMBER, DATE
LESS_THAN / LESS_THAN_OR_EQUAL	Numeric or date comparison	NUMBER, DATE
BETWEEN	Inclusive lower and upper range	NUMBER, DATE
EXISTS / NOT_EXISTS	Value is present or absent	Any type; no comparison value required
ANY_OF	At least one configured option exists in the multi-select value	MULTI_SELECT
ALL_OF	Every configured option exists in the multi-select value	MULTI_SELECT

Text comparisons are exact and case-sensitive after mapping transformations. Use case_mode when providers vary capitalization.

Allowlisted system facts

Rules can use source key, module, service, operational source ID, source-instance type/reference, marketing source ID/name, provider form ID, page ID, campaign ID, ad set ID, ad ID, received date/time, email/phone/company presence, latest behavior event/signal, active signal keys, reply count, deal-created count, deal-won count, and deal-lost count.

Use the exact keys displayed in Rule Reference. Other submitted field keys are not accepted as system facts; company-specific fields must be approved and mapped.

11. Synthetic Deterministic Preview

Preview uses three JSON containers:

• system: allowlisted synthetic system facts;
• profile_fields: current mapped profile-field IDs and synthetic values; and
• dimensions: active mapped dimension keys and synthetic values.

Use non-personal values only. Preview reads no contact, persists no run or score, changes no routing, calls no LLM, and consumes no AI credits.

Select Run Deterministic Preview. A successful result shows the score, band, matched factors, and group-cap breakdown. Preview confirms deterministic logic only; it does not prove that a real provider will supply a mapped field.

12. Validate and publish

Select Validate Draft after saving every section. Validation checks the profile, field revisions, dimensions, mappings, rules, conditions, bands, bindings, triggers, LLM policy, and conflicts with other published profiles.

Select Publish Draft only after validation passes. Publishing:

• freezes the version for repeatable scoring and audit;
• makes it eligible for new matching events;
• retires the profile’s previous published version; and
• increments profile-selection configuration for future events.

Publishing does not retroactively recalculate every existing contact. Existing immutable runs remain tied to the version that produced them.

To change a published profile, select Create Next Draft. The new draft copies the published mappings, groups, rules, conditions, bands, bindings, triggers, and LLM policy. Edit, preview, validate, and publish the next version.

13. Workflow recalculation

When Lead Scoring management is available, Workflow actions include Recalculate Lead Score for contacts.

This action requests a contact-stage recalculation using the contact’s eligible published/frozen profile and current approved scoring facts. It does not accept a score or arbitrary point adjustment from the workflow. The recalculation action is processed before later actions in the same workflow. Duplicate causal re-entry is suppressed to prevent scoring loops.

Use it after a workflow changes data that your scoring rules depend on but that does not already emit one of the supported reply or deal events.

14. CRM score display

On CRM > Contacts, the Lead Score column displays the effective score and band. A dash means the score is pending, unavailable, not selected by the display policy, or no eligible run has completed.

On Contact Conversation, the Lead Score card appears above Lead Status. It shows:

• effective score;
• score band and profile name;
• last update time; and
• a high, medium, low, or neutral color treatment based first on the band’s color key and then on score thresholds.

Possible messages include Lead scoring has no published profile, No eligible scoring run has completed, and Score is temporarily unavailable.

15. Run and Job Logs

Open Run and Job Logs to troubleshoot individual activity. Raw lead payloads, trusted intake context, provider responses, and LLM content are intentionally not displayed.

Filter	Default/options	Use
Profile	All profiles	Limit all tables to one company profile.
Phase	All, Event, Contact	Event is pre-assignment; Contact is post-assignment or later recalculation.
Run status	All, COMPLETED, FAILED_SAFE, SKIPPED_NO_PROFILE, SKIPPED_UNCHANGED, SKIPPED_SECURITY_POLICY	Find completed, safely failed, or skipped evaluations.
From	Empty	Includes records from 00:00:00 on the selected date.
To	Empty	Includes records through 23:59:59 on the selected date.
Contact ID	Empty; positive ID	Limits to one company-owned contact.
Queue ID	Empty; positive ID	Limits to one company-owned inbound queue event.
Job type	All plus the listed discovery, evaluation, LLM, rescore, backfill, and delivery types	Filters operational jobs. Some listed future/reserved types may have no records in the current release.
Job status	All, PENDING, RUNNING, RETRY_WAIT, COMPLETED, FAILED_FINAL, CANCELLED	Shows current processing state.
Rows per table	100	1–500 rows in each table.

Select a run ID to see sanitized facts, matched-factor breakdown, and safe run metadata. Use CRM Scoring Events to confirm event/signal acceptance and Evaluation Jobs to inspect attempts and safe error codes.

16. Deterministic Report

The report includes only completed deterministic evaluations. It does not use LLM-adjusted effective scores.

Filters are Profile, Phase, From, and To, with the same phase and inclusive date behavior as logs.

The report shows:

• completed runs, profiles, event runs, contact runs, average score, and score range;
• results by profile, version, phase, and band;
• daily trend; and
• results by source and form.

Statistics are suppressed until a cohort contains at least five completed runs. Exports are not available from this report.

Which mode should I choose?

Requirement	Recommended choice
Start safely with explainable scoring	Deterministic rules, LLM Mode OFF
Evaluate AI quality without changing priority	SHADOW
Apply controlled AI adjustments after testing	HYBRID with Max Adjustment 5–10 and Min Confidence 0.75 or higher
Use one policy for most leads	One default profile
Use a different policy for one exact form	Non-default profile with an exact form binding
Reuse budget/company-size logic across sources	Reusable dimension
Ignore optional missing data	NO_MATCH
Stop scoring if required data is missing	FAIL_SAFE
Prefer one of several equivalent mapped fields	FIRST_NON_MISSING with deliberate mapping priorities
Reject inconsistent values	INVALID_CONFLICT

Complete configuration examples

Example 1: Basic deterministic starter setup

Business requirement

A small sales team wants every new lead ranked using contact completeness, replies, and deal progress.

Configuration

1. Enable Lead Scoring.
2. Create Inbound Sales with key INBOUND_SALES, Base Score 40, Default profile On, and starter rules On.
3. Keep LLM Mode OFF.
4. Review the starter bands, groups, rules, and triggers.
5. Preview, validate, and publish.
6. Keep CRM display on Latest current score.

Result

New matching leads start at 40 and gain up to 10 contact-quality points. Replies and deal events recalculate the contact-stage score according to the starter rules. Scores appear in CRM after an eligible contact run completes.

Example 2: Facebook company-size qualification

Business requirement

Facebook leads answering over_50_users should gain 20 points.

Configuration

1. Submit a test Facebook lead and approve the company-size field as SINGLE_SELECT and BUSINESS.
2. Create COMPANY_SIZE as a SINGLE_SELECT dimension.
3. Map the approved revision to that dimension with deterministic scoring enabled.
4. Add a +20 rule using DIMENSION, key COMPANY_SIZE, operator EQUALS, type SINGLE_SELECT, value over_50_users.
5. Bind the profile to the exact synchronized Facebook form.
6. Preview using dimensions.COMPANY_SIZE = over_50_users, then validate and publish.

Result

Only leads using the matching profile and exact approved answer receive the 20 points. Other Facebook answers do not match the rule.

Example 3: Engage pricing-page intent

Business requirement

A visitor who submits a lead after viewing the pricing page should gain 15 points.

Configuration

1. Send a controlled Engage lead containing the conversion-page context.
2. Approve the discovered conversion-page field as TEXT and BUSINESS for deterministic use.
3. Map it to a direct profile field or a TEXT dimension such as CONVERSION_PAGE.
4. Add a +15 rule with EQUALS or CONTAINS for your exact pricing-page URL/value.
5. Bind the profile to ENGAGE or the intended source/form context.
6. Preview, validate, and publish.

Result

The lead receives the points only when the approved normalized page value matches. The application does not hard-code which URL is a pricing page.

Example 4: Dynamic no-response scoring

Business requirement

Reduce the score by 10 if no customer reply is received within 48 hours of assignment.

Configuration

1. Add an enabled LEAD_ASSIGNED trigger with signal AWAITING_RESPONSE, debounce 172800, lookback longer than 172800, and REEVALUATE.
2. Add an enabled MESSAGE_REPLY_RECEIVED trigger with a suitable lookback.
3. Add an ALL rule worth -10 requiring active signal AWAITING_RESPONSE and reply count equal to 0.
4. Validate and publish.

Result

The first recalculation occurs after 48 hours. If no reply is active, the rule subtracts 10. A qualifying reply prevents the no-response rule from matching. The expiry job recalculates again when the configured signal window ends.

Example 5: Advanced profile selection and routing

Business requirement

Use a general default profile, but a stricter policy for one Facebook enterprise-demo form; route Hot enterprise leads to a specialist team.

Configuration

1. Publish a default general profile.
2. Create a non-default enterprise profile with an exact FACEBOOK_FORM binding, valid integration ID, and matching form/provider references.
3. Publish it and enable Routing Integration.
4. In Advanced Lead Routing, create a rule for profile key ENTERPRISE_DEMO and routing band HOT.
5. Keep a normal routing fallback.

Result

The exact form binding outranks the broad/default profile. A completed event-stage deterministic Hot score can select the specialist route. If scoring is unavailable, the standard routing fallback still assigns the lead.

Example 6: LLM limits and exceptions

Business requirement

Evaluate AI enrichment safely without changing sales priority or exceeding 50 requests and 100 credits per day.

Configuration

1. Set profile LLM Mode to SHADOW, Max Adjustment 10, Min Confidence 0.75, and apply the recommended policy.
2. Approve and map at least one eligible BUSINESS structured field with both LLM flags enabled.
3. Enable company LLM enrichment.
4. Set Daily request limit to 50 and Daily credit limit to 100.00.
5. Keep a conservative queue and concurrency limit.

Result

Eligible contact-stage runs can call the LLM and successful calls consume credits, but the CRM score remains deterministic. Requests stop when any configured resource limit or available-credit check fails; deterministic scoring continues.

What happens if?

No profile matches a lead

The system uses the published default profile. If there is no published default, the scoring attempt is skipped with no score. Normal routing and assignment continue.

Two bindings match

The more-specific binding wins. At equal specificity, higher priority wins; then the binding with more selectors wins. Publishing blocks ambiguous equal-specificity/equal-priority overlap across profiles.

A mapped field is missing

NO_MATCH ignores it. FAIL_SAFE stops the evaluation without producing a misleading score.

Two fields mapped to one dimension disagree

INVALID_CONFLICT safely rejects the evaluation. FIRST_NON_MISSING uses the value from the lowest mapping-priority number.

Rule points exceed a group maximum

The group contributes only its configured maximum. Later rules in that group may contribute zero or only the remaining allowance.

The calculated score is below 0 or above 100

It is limited to 0 or 100 before band selection. A HYBRID adjustment is also limited to the same range.

LLM is enabled but no score appears

LLM does not create the base score. Confirm Lead Scoring is on, a matching profile is published, deterministic contact scoring completed, and the CRM display policy permits that profile.

AI credits or LLM capacity are unavailable

No provider request is sent, no failed request is charged, and deterministic scoring remains effective. Check Run and Job Logs for the safe skip code.

Lead Scoring is turned off with pending jobs

Pending jobs remain stored and unprocessed. They may resume after the feature is enabled again. Normal CRM operations continue.

A published profile needs editing

Published versions cannot be edited. Create the next draft, change it, preview, validate, and publish the replacement.

Common configuration mistakes

Problem	Why it happens	How to fix it
Lead Score shows --	No matching published profile or no eligible contact run completed.	Publish a default or bound profile and inspect logs. If scoring work never progresses, ask your deployment administrator to verify background processing.
LLM is on but score remains deterministic	Profile mode is OFF/SHADOW, confidence is too low, or no eligible mapped facts exist.	Use HYBRID only after SHADOW testing; verify both field-approval and mapping LLM switches.
Draft says LLM policy is incomplete	SHADOW/HYBRID has Max Adjustment 0 or a blank Rubric Version.	Select Apply Recommended LLM Settings, choose the intended mode, and save.
A field cannot be approved for scoring	Its privacy is PERSONAL/SENSITIVE, type is ineligible, or a conflict is unresolved.	Keep/raise privacy, select a compatible type, and explicitly resolve genuine conflicts. Do not bypass privacy restrictions.
A mapping blocks publication	The approval is stale/hidden/blocked, deterministic mapping is off, or the dimension is incompatible.	Reapprove/remap the current revision, remove unused mappings, or select a matching active dimension.
A direct-field rule stops working in a new draft	It uses an old generated profile-field ID.	Save mappings first and use the current ID, or use a stable dimension key.
Text matching is inconsistent	Provider capitalization or whitespace varies.	Enable trim and choose UPPER or LOWER consistently, then use the normalized comparison value.
Bands will not save or publish	There is a gap, overlap, duplicate key, or the ranges do not end at 100.	Sort ranges from 0 upward and cover every integer exactly once.
Non-default profile will not publish	It has no active binding.	Add at least one company-owned active source/form binding.
Binding validation fails	Form type/reference are incomplete, Facebook identifiers disagree, or a selector belongs to another company.	Use the IDs shown in this company and complete every linked selector.
No-response rule applies at the wrong time	Debounce was treated as lookback, or reply trigger is missing.	Use debounce as the wait time, keep a longer lookback, and configure reply tracking.
Trigger exists but score does not change	No rule references its signal/count, or group caps absorb its points.	Add or adjust the matching rule and review group contribution limits.
Report shows Suppressed	The selected cohort has fewer than five completed deterministic runs.	Broaden the date/profile filters or wait for more completed runs.

Recommended configurations

Small team

Best for one sales process:

• one default profile;
• Base Score 40;
• starter rules reviewed and simplified;
• Latest current score display;
• LLM OFF; and
• routing integration off until the team trusts the bands.

Growing multi-source team

Best for several inbound channels:

• one default profile plus exact-form profiles only where policy truly differs;
• reusable dimensions for budget, company size, product interest, and page intent;
• active reply/deal/no-response triggers;
• strict mapping conflicts for important numeric values;
• deterministic score-based routing with a fallback; and
• SHADOW mode with conservative daily limits.

Advanced operation

Best for mature governance:

• exact bindings with documented priorities;
• tested group caps and complete bands;
• synthetic preview cases for every critical rule boundary;
• SHADOW analysis before HYBRID;
• HYBRID Max Adjustment no more than 5–10 initially;
• Min Confidence at least 0.75;
• explicit credit, queue, and concurrency limits; and
• regular review of logs, skipped jobs, score distribution, and sales outcomes.

Settings quick reference

The table below accounts for every persistent setting and customer-facing operational control in the current Lead Scoring UI.

ID	Setting/control	Purpose	Default	Depends on	Recommended use
C001	Enable lead scoring for this company	Master runtime switch	Off	Plan, CRM, admin permission	Enable after at least one draft is ready to test.
C002	Routing Integration	Expose event deterministic facts to routing	Off	C001, Advanced Lead Routing	Enable only after score bands and routing fallback are tested.
C003	Score to display	Select CRM score policy	Latest current score	C001	Use Latest for most companies; hide during private testing.
C004	Published profile	Restrict CRM display to one profile	None	C003 = selected profile	Use when one published profile is the official CRM score.
C005	Enable optional LLM enrichment	Company LLM runtime switch	Off	C001, eligible profile/facts/credits	Start only after deterministic scoring works; use SHADOW first.
C006	Daily request limit	Cap daily LLM operations	100	C005	Set from expected eligible lead volume and budget.
C007	Daily credit limit	Cap estimated daily LLM credits	200.00	C005	Use as the company’s hard daily AI-spend guardrail.
C008	Maximum queued jobs	Cap queued/running LLM work	50	C005	Keep conservative unless sustained queue growth is expected.
C009	Maximum concurrent jobs	Cap simultaneous LLM work	2	C005	Increase only when provider capacity and budget allow it.
C010	Profile Name	Name the scoring profile	Required	New Profile	Use a clear business name such as Enterprise Inbound.
C011	Profile Key	Stable unique identifier	Required	New Profile	Choose a permanent key before publishing or routing by it.
C012	Profile Description	Explain profile purpose	Blank	New Profile	Record included sources, audience, and qualification goal.
C013	New Profile Base Score	Starting deterministic score	40	New Profile	Use 40–50 for a neutral starting position.
C014	Default profile	Use as fallback	Off	Only one published default	Make the first general-purpose profile the default.
C015	Apply recommended starter rules	Add starter bands/rules/triggers	On	New Profile	Keep on for the first profile, then adapt every starter value.
C016	Dimension Name	Name shared business concept	Required	New Dimension	Use when differently named fields represent one concept.
C017	Dimension Key	Stable rule key	Required	New Dimension	Choose a permanent business key such as COMPANY_SIZE.
C018	Dimension Description	Explain meaning/values	Blank	New Dimension	Document expected source fields and normalized values.
C019	Canonical Type	Fix shared value type	TEXT shown first	Compatible mapped fields	Match the approved type of every field in the dimension.
C020	Privacy Floor	Set minimum privacy	BUSINESS	Compatible mapped fields	Keep BUSINESS unless a stricter source field requires more.
C021	Type Override	Approve effective field type	Observed type	Field Review	Override only after confirming the provider’s real data shape.
C022	Privacy Classification	Confirm/raise privacy	BUSINESS selected	System privacy floor	Classify conservatively; never lower the detected privacy floor.
C023	Allow deterministic scoring	Permit deterministic mapping	On	Eligible type/privacy	Enable for fields intended for explainable scoring rules.
C024	Allow LLM use	Permit later LLM mapping	Off	Eligible structured type/privacy	Enable only for necessary, eligible structured business facts.
C025	Resolve observed type conflict	Confirm intentional conflict resolution	Off	TYPE_CONFLICT field	Use only after reviewing conflicting samples and selecting a type.
C026	Hide field	Remove field from current approval use	Action	Discovered field	Hide irrelevant fields that may be reviewed later.
C027	Block field	Prevent future review/use	Action	Discovered field	Block prohibited or permanently unsuitable scoring data.
C028	Draft Base Score	Starting score for this version	Profile value	Draft	Tune with rule ranges so positive and negative movement is useful.
C029	LLM Mode	OFF, SHADOW, or HYBRID	OFF	Draft; C005 for runtime	Use OFF initially, SHADOW for evaluation, HYBRID after validation.
C030	Max Adjustment	Bound LLM adjustment	10 recommended	C029 SHADOW/HYBRID	Start at 5–10 to preserve deterministic influence.
C031	Min Confidence	HYBRID application threshold	0.75 recommended	C029 HYBRID	Start at 0.75 or higher and tune from SHADOW evidence.
C032	Rubric Version	Select server rubric version	sales-intent-v1	C029 SHADOW/HYBRID	Keep the supplied version unless an operator provides another.
C033	max_input_facts	Bound LLM fact count	30 recommended	C029, eligible mappings	Keep only enough facts to cover the sales rubric.
C034	max_fact_value_chars	Bound each LLM value	300 recommended	C029	Use a conservative bound for concise structured facts.
C035	max_total_input_bytes	Bound total LLM input	8,192 recommended	C029	Keep the recommendation unless validated facts require less.
C036	max_estimated_tokens	Bound estimated input tokens	2,000 recommended	C029	Use as an input-size and cost safeguard.
C037	max_free_text_fields	Bound eligible free-text facts	0 recommended	C038 and field policy	Keep at 0 unless an approved policy explicitly permits free text.
C038	allow_business_free_text	Permit eligible BUSINESS free text	Off	C037 >= 1, field policy	Keep off for privacy and predictable input control.
C039	allow_safe_summary	Permit bounded result summary	On	C040 >= 1	Enable when administrators need a short validated explanation.
C040	safe_summary_max_chars	Bound result summary	240 recommended	C039	Keep short enough for review without storing verbose content.
C041	reason_code_allowlist	Restrict returned LLM reasons	Two recommended codes	C029	List only reason codes used by the business rubric.
C042	inbound_field_revision_id	Select exact approved field revision	Required	Approved field	Copy the current approved ID shown beside the draft.
C043	dimension_id	Map to shared dimension	0	Compatible active dimension	Use 0 for an exact field or a dimension for cross-source reuse.
C044	display_label	Label mapped field	Required	C042	Use a recognizable business label for administrators.
C045	transformation trim	Trim whitespace	On	Mapping	Keep on for stable comparisons.
C046	transformation case_mode	Normalize text case	NONE	Mapping	Use UPPER or LOWER when provider capitalization varies.
C047	transformation sort_multi_values	Sort multi-select values	Off	MULTI_SELECT mapping	Enable when option order should not affect normalized evidence.
C048	missing_value_policy	Ignore or fail safely on missing value	NO_MATCH	Mapping	Use NO_MATCH for optional facts; FAIL_SAFE for mandatory facts.
C049	mapping_priority	Choose first equivalent candidate	100 if omitted	Shared dimension candidates	Give the preferred source field the lowest number.
C050	conflict_policy	Select or reject conflicting values	INVALID_CONFLICT if omitted	Shared dimension candidates	Use INVALID_CONFLICT for important facts; otherwise FIRST_NON_MISSING.
C051	deterministic_enabled	Use mapping in deterministic scoring	On	C023; required for publish	Keep on for every saved active mapping.
C052	llm_enabled	Offer mapping to LLM policy	Off	C024, C005, C029	Enable selectively after the field approval also allows LLM use.
C053	band_key	Stable score-band identifier	Required	Score Bands	Use permanent business keys such as HOT, WARM, and COLD.
C054	band_name	User-facing band label	Required	Score Bands	Use labels sales users can interpret immediately.
C055	color_key	CRM color category	DEFAULT if omitted	Score Bands	Use supported semantic colors to distinguish priority visually.
C056	minimum_score	Inclusive band minimum	Required	Complete 0–100 coverage	Start the first band at 0 and avoid gaps.
C057	maximum_score	Inclusive band maximum	Required	Complete 0–100 coverage	End the final band at 100 and avoid overlaps.
C058	band priority	Stable band order metadata	100	Score Bands	Use distinct values for predictable administrator ordering.
C059	trigger_key	Choose authoritative event	Required	Trigger policy	Add only events that should cause score reconsideration.
C060	signal_key	Name active company signal	Required	Rule behavior condition	Use a stable key that clearly describes the active behavior.
C061	trigger enabled	Activate/deactivate policy	Off if omitted	Trigger policy	Enable after a matching behavior rule is configured.
C062	lookback_seconds	Signal lifetime	0	C064 expiry consistency	Set to the period in which the activity should influence scoring.
C063	debounce_seconds	Delay/coalesce recalculation	0	Less than positive lookback	Use as the wait period for no-response logic or event coalescing.
C064	expiry_policy	NONE or REEVALUATE	Based on lookback	C062	Use REEVALUATE whenever a positive lookback can expire.
C065	maximum_occurrences	Cap behavior event count	1	Trigger policy	Increase only when rules intentionally inspect repeated activity counts.
C066	binding_name	Describe profile selector	Required	Binding	Name the source/form scope clearly for future audits.
C067	binding source_key	Match inbound source	Blank	Binding context	Use for a broad provider or intake-family match.
C068	binding module_scope	Match source module	Blank	Binding context	Add when the same source key can occur in different modules.
C069	binding service	Match service	Blank	Binding context	Leave blank unless the exact trusted service value is known.
C070	binding source_id	Match operational source	0	Company source	Use to narrow a broad source to one company source record.
C071	marketing_lead_source_id	Match CRM source	0	Active company/global source	Use when CRM lead-source classification selects the profile.
C072	form_type	Select form family	Blank	Must pair with C073	Set with an exact form reference for form-specific scoring.
C073	form_reference	Select exact form	Blank	C072; company-owned form	Use the current company form/provider reference.
C074	integration_id	Select provider integration	0	Provider object binding	Use for an exact connected-provider object such as Facebook form.
C075	provider_object_type	Select provider object family	Blank	C074/C076	Pair with the provider reference and integration.
C076	provider_object_reference	Select exact provider object	Blank	C074/C075	Use the exact synchronized provider reference.
C077	binding priority	Resolve equal-specificity matches	100	Matching bindings	Raise only to prefer one match at the same specificity.
C078	binding status	Activate profile selector	ACTIVE	Published profile	Use INACTIVE to retain a binding without allowing selection.
C079	group_key	Stable rule-group identifier	Required	Rule group	Use a permanent key for related scoring factors.
C080	group_name	Name rule group	Required	Rule group	Use a business label such as Customer Engagement.
C081	minimum_contribution	Cap negative group points	-100 if omitted	Grouped rules	Set a deliberate downside limit for related negative rules.
C082	maximum_contribution	Cap positive group points	100 if omitted	Grouped rules	Prevent several related signals from dominating the score.
C083	group priority	Stable group order	100	Rule group	Use distinct values for predictable administrator ordering.
C084	rule enabled	Activate/deactivate rule	On	Rule	Disable to retain a draft rule without applying it.
C085	rule_type	Classify rule	FIELD_SIGNAL	Rule	Choose the category that best describes the business factor.
C086	reason_code	Stable unique factor code	Required	Enabled rule	Use a permanent code that explains matched evidence.
C087	rule_name	Human-readable factor name	Required	Rule	Write the outcome users should understand in score details.
C088	rule_description	Explain rule intent	Blank	Rule	Record the business justification and expected values.
C089	rule group_key	Apply group caps	Blank/ungrouped	Existing group	Group related rules that need a shared contribution cap.
C090	match_mode	Require ALL or ANY conditions	ALL	Rule conditions	Use ALL for strict qualification and ANY for alternatives.
C091	points	Add/subtract score	Required	Matching rule	Choose an impact proportionate to the signal’s business value.
C092	rule priority	Evaluate lower values first	100	Especially group caps	Give stronger preferred rules lower values inside capped groups.
C093	condition_source_type	SYSTEM_FACT, PROFILE_FIELD, DIMENSION	Required	Condition	Prefer DIMENSION for reusable concepts and SYSTEM_FACT for trusted context.
C094	system_fact_key	Select allowlisted system fact	Blank	C093 SYSTEM_FACT	Select only a key shown in Rule Reference.
C095	dimension_key	Select stable mapped dimension	Blank	C093 DIMENSION	Use the active dimension key saved in this profile.
C096	profile_field_id	Select exact current field	0/blank	C093 PROFILE_FIELD	Use only for one exact mapped field in the current draft.
C097	operator	Choose comparison	Required	C098 type	Choose the simplest operator matching the business test.
C098	comparison_type	Normalize comparison	TEXT if omitted	Fact/dimension type	Match the source fact, approved field, or dimension type exactly.
C099	comparison_value	Main target value/list	Operator-dependent	C097	Enter the normalized value actually delivered by the source.
C100	comparison_value_to	BETWEEN upper bound	Blank	C097 BETWEEN	Supply only for the inclusive BETWEEN range.
C101	condition priority	Stable condition order	100	Condition	Use distinct values for readable, stable ordering.
C102	Preview system facts	Test allowlisted source/behavior facts	Example values	Draft preview	Test source and behavior boundaries with synthetic values.
C103	Preview profile fields	Test current mapped field IDs	Empty object	Saved mappings	Test exact-field rules with current draft IDs only.
C104	Preview dimensions	Test active mapped dimensions	Empty object	Saved mappings/dimensions	Test reusable cross-source rules with synthetic values.
C105	Apply Recommended LLM Settings	Fill bounded working LLM defaults	Action	Draft LLM editor	Use before first SHADOW/HYBRID save, then review the values.
C106	Apply Starter Template	Add starter rules to an empty draft	Action	No groups/rules/triggers	Use for an empty first draft instead of building every rule manually.
C107	Load Tenant Field Example	Load first compatible mapping example	Action	Approved field	Use as a starting structure, then replace IDs and policies deliberately.
C108	Load Group Example	Load example group JSON	Action	Rule editor	Use to learn the group format before creating business-specific caps.
C109	Load Tenant-Aware Rule Example	Load context-aware rule example	Action	Mapped field/dimension if available	Use after mappings exist to create a valid draft rule structure.
C110	Load No-response Rule	Load example negative behavior rule	Action	Matching group/triggers	Use with assignment/reply triggers, then adjust timing and points.
C111	Load Reply Example	Load reply trigger example	Action	Trigger editor	Use when customer replies should cause rescoring.
C112	Load No-response Triggers	Load assignment/reply trigger examples	Action	Trigger editor	Use as the base for delayed no-response scoring.
C113	Load Facebook Example	Load broad Facebook binding example	Action	Binding editor	Use as a template and narrow it to the intended form when possible.
C114	Run Deterministic Preview	Test draft in memory	Action	Valid synthetic facts	Run after every material rules, mapping, or band change.
C115	Validate Draft	Run publication checks	Action	All draft sections saved	Run after saving every section and before publishing.
C116	Publish Draft	Activate immutable version	Action	C115 passes	Use only after preview and validation match expected outcomes.
C117	Create Next Draft	Copy published version for editing	Action	Published profile with no draft	Use for every change to an already published profile.
C118	Logs Profile	Filter logs by profile	All profiles	Logs page	Select when troubleshooting one policy.
C119	Logs Phase	Filter event/contact runs	All phases	Logs page	Use Event for routing issues and Contact for later scoring.
C120	Logs Run status	Filter evaluation outcome	All statuses	Logs page	Filter skipped/failed-safe statuses when a score is missing.
C121	Logs From	Inclusive start date	Empty	Logs page	Set the start of the incident or test window.
C122	Logs To	Inclusive end date	Empty	Logs page	Set the end of the incident or test window.
C123	Logs Contact ID	Filter one contact	Empty	Company-owned contact	Use when investigating a Contact Conversation score.
C124	Logs Queue ID	Filter one queue event	Empty	Company-owned queue record	Use when investigating one inbound lead before assignment.
C125	Logs Job type	Filter processing category	All types	Logs page	Select the specific discovery, scoring, or LLM operation.
C126	Logs Job status	Filter processing state	All statuses	Logs page	Use to find waiting, retrying, failed, or completed work.
C127	Rows per table	Limit each log table	100	1–500	Keep 100 normally; increase for a wider troubleshooting window.
C128	Report Profile	Filter deterministic report	All profiles	Report page	Select one profile to evaluate its score distribution.
C129	Report Phase	Filter event/contact results	All phases	Report page	Separate pre-routing scores from later contact scores.
C130	Report From	Inclusive start date	Empty	Report page	Set the beginning of the reporting period.
C131	Report To	Inclusive end date	Empty	Report page	Set the end of the reporting period.
C132	Recalculate Lead Score workflow action	Request governed contact rescore	Action	Lead Scoring permission and workflow contact	Use after workflow changes relevant contact data without a native scoring event.