An engagement scoring model is a system that assigns points to user actions to quantify how interested, active, or ready a contact is. It turns clicks, logins, event participation, or product usage into a single score you can track, compare, and act on. Teams use it to prioritise outreach, trigger campaigns, and forecast outcomes because it translates behaviour into an objective signal.
Why use an engagement scoring model?
Strong models improve conversion and retention because they focus attention on the right people at the right time. Sales teams use scores to sort leads and set follow‑up cadence. Marketers use them to segment audiences, personalise content, and time messages. Success and product teams use them to spot adoption, churn risk, and expansion opportunities. The score becomes a shared language that reduces guesswork and speeds decisions.
What outcomes should it drive?
Start with the business goal, then design the score to predict it:
If you want pipeline, predict sales‑qualified lead (SQL) creation or opportunity win.
If you want adoption, predict “activated within 14 days” or “7-day retention.”
If you want event ROI, predict meeting bookings, booth visits, or post‑event demos.
If you want donor growth, predict a first or repeat donation in the next 30 days.
Tie the score to one concrete conversion. That keeps weights, thresholds, and tests honest.
What behaviours should you score?
Score observable signals that show intent or value realisation:
Email and SMS: opens, clicks, replies, unsubscribes.
Time zones and timestamps: store in UTC; display locally to users.
GDPR/CCPA: collect consent and honour preferences because some signals (like email tracking) are regulated.
Bot and internal traffic filters: exclude staff, QA, and bots to prevent phantom lift.
How do you calculate a simple points model?
Use a repeatable formula with round numbers:
Pricing page view: +15
Demo request submitted: +35
Webinar attended (≥30 minutes): +20
Product trial sign‑up: +30
Invited a teammate: +25
Logged in ≥3 days in a week: +20
Email click: +5 (cap 2 per week)
Unsubscribe: −30
21 days inactive: −20
Score decays 15% after 14 days of no activity
Normalise to 0–100 by dividing by a defined maximum and capping the score. Publish the rubric so sales and marketing stay aligned.
How do you validate and calibrate the score?
Always test against outcomes.
Back‑test: apply the score to the last 6–12 months and plot conversion by score band. You should see monotonic lift.
AUC/ROC for binary outcomes: aim for >0.70 before you route critical workflows.
Lift charts: top decile should convert 3–5× the base rate for prioritisation to pay off.
Calibration: if 40–49 band converts at ~8%, 50–59 at ~12%, your bands are informative.
Shadow period: run the score silently for 2–4 weeks while teams work as usual. Compare performance to avoid disruption.
What thresholds should you use?
Let thresholds reflect capacity and economics, not gut feel.
Sales‑ready: pick a band where expected value × contact volume fits your team’s follow‑up capacity within 24–48 hours.
Nurture: routes to content journeys when the score is middling but trending up.
Re‑engage: triggers when the score drops by, say, 30% in two weeks.
Churn watch: flag accounts whose product engagement score falls below their own 60‑day baseline.
Review thresholds monthly at first, then quarterly as the model stabilises.
How do you combine interest and fit?
Decision first: don’t send every high‑click contact to sales. Use a two‑score system:
Fit score: firmographic/demographic match to your ideal customer profile (industry, size, tech stack, region).
Engagement score: behavioural intent.
Act by quadrant:
High fit, high engagement: immediate outreach, senior rep.
High fit, low engagement: targeted awareness nurture.
Low fit, high engagement: self‑serve or low‑touch path.
Low fit, low engagement: suppress or inexpensive automation.
What’s different for product engagement?
Product engagement measures value realisation inside your app.
Core actions: first value (created a project, shipped a campaign), depth (repeat core action across weeks), breadth (adopted multiple features), team effects (invited/activated seats).
Health composite: active days per week, time‑to‑value, feature NPS, task completion, error rate.
Milestones: onboarded in 7 days, activated by day 14, steady state by day 30, power user by day 60.
Weight actions tied to retention and expansion. For example, “invited 3 teammates” might be the strongest indicator for B2B collaboration tools.
How does engagement scoring differ by context?
B2B sales: favour meeting attended, buying‑committee activity, proposal views, and security questionnaire started. Use account‑level roll‑ups.
B2C ecommerce: focus on sessions, product views, add‑to‑carts, wishlists, and checkout starts. Time‑decay is faster.
Events and webinars: consider registration show rate, watch time, Q&A, poll responses, and booth scans. Connect to post‑event actions like demo booked.
Nonprofits: score volunteer hours, event attendance, email clicks, and recency of last gift; include negative weights for lapsed donors.
Education: count course enrolment, lesson completion, forum participation, and assessment submissions.
How do you roll up to the account level?
Buying decisions are often collective. Aggregate people to an account score:
Weighted average by persona (decision‑maker ×2, champion ×1.5, influencer ×1).
Max operator for “spike” detection: if one person hits 95/100, route now.
Diversity bonus: add points when ≥3 unique personas are active because committee engagement increases deal odds.
What workflows should the score trigger?
Make the score actionable the moment it updates.
Sales: create tasks when score crosses the sales‑ready threshold; set SLA based on band.
Marketing: switch nurture tracks when score moves up or down bands; throttle email when the score falls.
Product: in‑app guides when a key feature score is low; offer training when adoption stalls.
Success: health checks when product score drops 20% week‑over‑week; expansion plays when breadth hits a target.
RevOps: routing changes when fit × engagement surpasses a handoff threshold.
Every trigger needs an owner, due‑by time, and success metric.
How do you prevent gaming and noise?
Cap repeated actions and apply diminishing returns.
Penalise bot‑like behaviour (100 rapid clicks, zero dwell).
Ignore self‑referrals and internal IP ranges.
Require multi‑signal confirmation before major workflow changes. For example, need “pricing view + demo request” or “trial sign‑up + invite teammate.”
What do good dashboards show?
Useful dashboards fit on one screen.
Score distribution over time with bands.
Conversion rate by band and by source/channel.
Volume at or above the sales threshold versus sales capacity.
Top rising and falling accounts this week.
Feature‑level contributions for product engagement.
Alert queues with SLA compliance.
Annotate model changes on the timeline so you can separate performance shifts from scoring tweaks.
How do you keep the model accurate over time?
Models drift. Put maintenance on a calendar.
Data audits monthly: missing events, identity stitching errors, channel outages.
Performance review quarterly: check lift, AUC, thresholds, and band sizes.
Re‑weighting: retrain models or update heuristics when behaviour or messaging changes.
Seasonality: tighten decay during high‑velocity seasons; relax during holidays if cycles lengthen.
Experimentation: A/B test triggers (e.g., outreach messaging for 70–79 band) and keep what increases conversion per rep hour.
What’s a worked mini‑example?
A SaaS company wants to prioritise demo follow‑up within 24 hours.
Goal: predict demo‑to‑opportunity conversion.
Signals and weights:
Pricing page view: +15 (cap 2 per week)
Demo booked: +35
Attended demo: +25
Trial sign‑up: +30
Invited teammate: +25
Email reply to rep: +20
Unsubscribe: −30
Inactive 14 days: −15
Exponential decay with half‑life 14 days
Thresholds:
80–100: hot; call within 2 hours
60–79: warm; call within 24 hours
40–59: nurture; send case study and wait for new activity
Back‑test results:
Base demo‑to‑opportunity: 18%
80–100 band: 52% (2.9× lift)
60–79 band: 29% (1.6× lift)
40–59 band: 16% (≈base)
Sales staffs to the 80–100 band first, reaching 92% of them within SLA, and overall pipeline rises 21% quarter‑on‑quarter because effort concentrates where it pays.
How do you communicate the score to teams?
Publish a one‑page rubric with the exact weights, decay, thresholds, and actions.
Train with real examples: show a high‑scoring timeline and narrate why it matters.
Embed the score in the CRM record and task queues; don’t hide it in a distant dashboard.
Give reps a “why now” breakdown: top three contributing activities and the recency of each.
What governance do you need?
Ownership: name a cross‑functional council (marketing ops, sales ops, product analytics, data) that can change the model.
Change control: batch changes and ship on a set day; log versions and hold a two‑week shadow test.
Privacy: minimise tracking to what’s necessary, respect consent signals, and provide clear preference centres.
Fairness: check whether the score penalises certain segments without a business‑relevant reason; fix features or routing if so.
Common pitfalls and how to avoid them
Too many low‑intent signals: keep or down‑weight blog reads and generic browsing unless they correlate with your target conversion.
No decay: old activity keeps people “hot” forever; add time‑based drop‑offs.
No negative signals: unsubscribes and long inactivity matter; subtract points.
Score without actions: every band needs a playbook and SLA; otherwise the number is just decoration.
Opaque logic: if sellers can’t trust or explain it, they won’t use it; keep weights simple or provide clear “top contributors.”
How do you evaluate channel contributions without bias?
Attribute by sequences: look at the path to conversion rather than last touch only.
Use controlled experiments: e.g., holdout for a nurture series and compare score lift and conversion.
Watch for mechanical inflation: channels that create many tiny events (push notifications, chat nudges) shouldn’t dominate the score. Aggregate into meaningful macro‑events.
How do engagement and lead scoring differ?
Lead scoring usually blends fit and behaviour for net‑new leads. Engagement scoring focuses on behaviour over time and often continues post‑sale. Use both: fit to decide whether a contact is worth attention; engagement to time that attention.
How do you localise the model for different markets?
Different cadences: in some regions email engagement is low by norm; weight web or chat signals higher.
Language and time zones: compare behaviour within a market’s own baseline.
Channel mix: swap in WhatsApp interactions or local social networks where relevant.
How do you set up measurement in under two weeks?
Week 1:
Define the outcome and time window.
List 10–12 signals with simple weights.
Implement events in product analytics and marketing automation.
Build a daily job to compute 0–100 with 30‑day sliding window.
Week 2:
Add decay and negative weights.
Set thresholds from a quick back‑test.
Route actions and train teams.
Start a four‑week shadow log for version 2 improvements.
Ship quickly, then iterate based on lift and team feedback.
How do you score events and webinars well?
Treat events as high‑intent when they involve time commitment or two‑way interaction.
Registration: +10 if attended; +0 otherwise.
Live attendance: +20 for ≥30 minutes; +10 for 10–29 minutes.
Q&A or poll participation: +10.
On‑demand watch: +10 for ≥50% of content.
Booth scan: +15; meeting booked on site: +25.
Decay faster (half‑life 7–10 days) because event intent fades quickly unless followed by product actions.
How do you extend to post‑sale health and expansion?
Blend product and relationship signals.
Health up: active days, breadth, team adoption, feature depth, positive NPS, time‑to‑value achieved.
Risk down: unresolved P1 tickets, executive sponsor churn, usage drop versus baseline, late invoices.
Expansion triggers: seat saturation, feature limits hit, new team invited, successful use cases expanded into new departments.
Route risks to success managers within 24 hours and expansion triggers to account managers with a tailored offer.
A quick glossary of related terms
Activity: a single event such as a click, login, or view.
Signal: a grouped set of activities intended to reflect intent or value (e.g., “evaluation intent”).
Weight: numeric importance assigned to a signal.
Decay: time‑based reduction in score when no new activity occurs.
Threshold: score value that triggers a workflow.
Band: a range of scores grouped for interpretation or routing.
Fit score: non‑behavioural score based on firmographics/demographics.
Account score: engagement aggregated across people in the same organisation.
Calibration: ensuring score bands map to observed conversion probabilities.
Drift: performance degradation over time due to behaviour or system changes.
Checklist to launch with confidence
Named outcome and time window.
10–15 clear signals, each with a reason and weight.
Negative signals and time decay configured.
0–100 scale with published thresholds and SLAs.
Back‑test demonstrating lift by band.
Dashboard with distribution, conversion by band, and top movers.
Documented governance and change log.
Training with real examples and “why now” explanations.
Final take
A good engagement scoring model is specific to your goal, simple enough to trust, and strict about recency and negative signals. Ship the smallest useful version, tie it to real actions, and keep tuning it against outcomes. When the score consistently helps teams focus and move faster, you’ve built the right one.
