Sales activity scoring: what to automate, what to leave alone

Activity scoring is the practice of assigning numeric scores to sales activities (and the people performing them) to prioritize follow-up, flag at-risk deals, and surface signal in a sea of CRM data. Done well, scoring compounds across every rep. Done poorly, it produces noise that reps ignore and dashboards nobody acts on.

Peer-reviewed research on lead scoring and vendor case studies both suggest meaningful win-rate improvements from structured scoring, with MarketingSherpa case studies reporting conversion lifts in the double digits for teams that implement scoring with discipline. This post covers the models that produce signal, the configurations that break, and the automation that keeps the surface clean.

The two scoring layers

Modern sales activity scoring runs two layers that stack:

1. Lead/contact scoring

A numeric score per contact based on behavioral signals (email opens, page visits, content downloads, meeting bookings, form submissions) and firmographic signals (industry, company size, title, region). Scores feed lead routing and lifecycle stage transitions.

2. Deal/opportunity scoring

A numeric score per deal based on deal shape (stage, age, dwell time, stakeholder count, activity recency, qualification framework completion). Scores feed forecast accuracy and risk flagging.

Both layers depend on clean underlying data. Scoring on bad data produces confidently wrong priority rankings.

What to score

The signals that actually predict outcomes:

For contact scoring

• Fit signals (firmographic): industry, employee count, revenue, title, region
• Intent signals (behavioral): page visits to pricing or product pages, content downloads of mid-funnel assets, demo requests, trial starts
• Engagement velocity: rate of recent engagement vs. historical baseline (someone going from 1 visit/month to 5 visits/week is signal)

For deal scoring

• Stage velocity: days spent at current stage vs. historical median per stage
• Activity recency: days since last meaningful interaction (not auto-logged emails)
• Stakeholder count: number of engaged contacts at the account
• Qualification depth: completion rate of framework fields (MEDDIC, SPICED)
• Bounce pattern: whether the deal has moved backward through stages

Each signal gets a weight. Weights come from historical close-rate analysis, not from guesses in a conference room.

What NOT to score

The common mistakes that produce noise:

1. Email open counts as the primary signal

Email opens are inflated by email privacy features and image-blocking tools. A high open count may indicate interest or may indicate the recipient has image loading enabled. Weight it low.

2. Call counts without context

A rep made 50 calls. A rep connected on 3. A rep had 1 substantive conversation. Scoring 50 as high and 3 as medium produces activity theater. The 1 conversation is the signal.

3. Logged notes as a scoring factor

If reps know note length drives their score, they write longer notes. The notes get less useful. Do not score on activity-system gaming surface.

4. CRM login frequency

Reps that live in the CRM are not necessarily better reps. Rep performance is in the deals, not in the audit log.

The scoring model

A reference scoring config for a mid-market B2B SaaS team:

Contact score (0-100):

• +30 for ideal customer profile firmographic match
• +20 for pricing page visit in last 30 days
• +15 for mid-funnel content download in last 60 days
• +15 for demo request submitted
• +10 for engagement velocity above baseline
• +10 for repeated engagement from multiple stakeholders at the same account

Thresholds:

• 0-39: Lead (nurture)
• 40-69: MQL (route to rep for qualification)
• 70-100: SQL (route to rep with urgency flag)

Deal score (0-100):

• +25 for qualification framework fields complete
• +20 for stage velocity within historical median
• +20 for activity in last 7 days
• +15 for 3+ engaged stakeholders at account
• +10 for champion identified and recently engaged
• +10 for deal has not bounced backward in stage history

Thresholds:

• 0-39: at-risk (flag in weekly review)
• 40-69: healthy
• 70-100: high confidence

Automation that keeps scoring clean

Four automations:

1. Scheduled recomputation. Scores refresh nightly against all open leads and deals so the surface stays current.
2. Transition triggers. When a score crosses a threshold, fire the appropriate action (route lead, flag deal, notify rep, create task).
3. Threshold calibration. Review scoring thresholds quarterly against actual close-rate data. Adjust if the thresholds have drifted from historical patterns.
4. Rep override. Reps can mark a lead or deal as “my judgment overrides the score.” The override gets logged for audit, and the system tracks over time whether rep overrides beat the score (useful calibration signal).

How Strkr handles scoring

Strkr supports both scoring layers natively on every paid tier:

• Formula fields on the Contact and Deal objects compute the score. Formulas reference any CRM data (standard fields, custom fields, related records up to three levels deep).
• The no-code flow builder fires actions on score threshold transitions (route, notify, task, flag).
• AI-assisted forecasting layers per-deal model probability on top of the formula-based score, surfacing deals the formula sees as healthy but the model sees as at-risk.
• Weekly reports show scoring distribution, threshold hits, and score accuracy over time.

The design target: a sales ops lead at Series A should be able to build the full scoring surface in a day, without an admin certification or a dedicated scoring tool.

Related reading: CRM with AI forecasting: what to look for and what to ignore covers the AI-based scoring layer.

Conclusion

Sales activity scoring works when the signals correlate with outcomes, the weights come from historical data, and the automation keeps the surface current. Scoring breaks when it rewards activity theater (email opens, call counts, note length) over substantive engagement (meaningful conversations, qualification depth, stakeholder breadth).

Build scoring on signals that predict wins. Review the model quarterly against actual close rates. The compound effect across every rep makes scoring one of the highest-leverage sales ops projects available.