Reference · Scoring

How BHASM scores customers.

Every score is the sum of five inputs. No black box. This page answers one question end to end: why is this customer scored 94?

What the score means

A 0–100 number assigned to every customer in your active base. Computed on every brief run. Drives which accounts surface in your weekly dividend and in what order.

Severity bands
RangeBandWhat BHASM does
0–49WatchingContinues to monitor. No action surfaced.
50–74Developing concernEligible for the brief if room above threshold.
75–89UrgentSurfaced this week. Usually message generated.
90–100CriticalTop of the brief. Claude-crafted message. First-priority send window.

The five inputs

Every urgency score is a composite of five behavioural inputs. The composition is visible on every brief card via the Why? panel.

1. Silence duration

What it measures: days since last purchase compared to this customer's personal average purchase cycle — not a global benchmark.

Why personal: a customer who buys weekly and goes 30 days silent is more urgent than one who buys annually. A global day-30 threshold would be wrong for both.

The largest contributor to the final score. At day 47 silence against a shorter personal average, silence becomes the dominant signal in the composite.

2. Cycle elapsed

What it measures: how many personal cycles have passed beyond this customer's normal rhythm.

  • On rhythm — contribution stays low.
  • Drifting — contribution rises.
  • A full cycle overdue — significant contribution.
  • Deep silence — contribution approaches the ceiling.

The second-largest contributor. The further past the normal pattern, the more this factor shapes the score.

3. Spend trajectory

What it measures: whether recent purchases are growing or declining in value.

A customer whose last three orders went $38 → $29 → $22 carries a different signal than one whose orders held steady. Spend decline often precedes silence.

Lower contribution than silence or cycle elapsed — but often the deciding factor when other signals are borderline.

4. World context

What it measures: external context affecting this customer's likelihood to respond right now.

ContextEffect on the score
Pre-festival window (Diwali, Eid, Christmas)Amplified — buying intent elevated
Payday window (regional pattern)Amplified — aligned to detected payroll cycle
Normal contextUnchanged
Post-disruption (flood, riot, shock)Suppressed — wrong moment

World context never raises a calm signal into a panic; never silences a real one into nothing.

5. Relationship balance

What it measures: the running balance of outreach sent versus responses received from this customer.

Counter-intuitively, this input reduces urgency, not increases it. The more BHASM has reached out without a response, the less aggressive it becomes — urgency is recalibrated toward patience, not amplified.

The breakdown made visible

One real customer. Urgency 94. Every factor with its contribution.

How this score was reached 94
Silence duration
Cycles elapsed (overdue)
World context (pre-season)
Purchase velocity trend
Relationship balance

What does not affect the score

  • Customer age in the database. A 5-year customer who just went silent gets the same scoring as a new one. Past loyalty does not protect present silence.
  • Total historical spend. Past spend does not create immunity. A high-LTV customer going silent is more urgent, not less — because the value at risk is larger.
  • Your plan tier. No premium scoring for premium plans. The intelligence is the same on Seed and on Performance — only the visibility and autonomy differ.
  • Channel preference. WhatsApp vs email is a delivery decision, not a scoring input.

Score vs surfacing threshold

The score is the raw signal. The brief only surfaces customers above a threshold — and that threshold adjusts to the health of your business.

When BHASM detects elevated retention risk, the brief casts a wider net — more accounts surface so the founder can act on developing signals before they become lost customers. In a healthy state, only the clearest signals reach the brief, keeping attention on what actually needs a decision.

Bias correction

No model is perfect. BHASM monitors its own predictions.

After every brief, the system compares its urgency prediction with what actually happened to each customer over the following 30 days. If urgency was over-estimated for a particular industry or archetype, the calibration adjusts for that segment on the next run.

The first month of any new tenant is calibration mode: the scoring runs but the system is also learning your specific population. By month two, the model has tightened to your customers' actual rhythms.

Honest statement
BHASM does not claim perfect predictions. It claims traceable predictions. Every score cites its inputs; every inaccuracy feeds calibration; nothing is hidden in a model the founder cannot inspect.