The original HubSpot lead score was a single additive number. Lead Scoring v2 introduces multiple scores (Engagement, Fit, Intent), property-level decay, and rule grouping. The migration looks routine and isn’t. Here’s the cutover playbook.
Don’t migrate; run side by side first
For 30 days, build the v2 model in parallel and write its output to lead_score_v2_engagement and lead_score_v2_fit, separate from the legacy hubspotscore. Your routing keeps using the old score. You’re collecting evidence.
Compare distributions before cutover
Pull the daily score distribution for both models. If v2 produces a meaningfully different shape (e.g., the legacy bell curve becomes bimodal in v2), your downstream thresholds break. Adjust the v2 thresholds before flipping routing, never after.
Decompose Fit and Intent
Fit is who they are: company size, industry, job title. Intent is what they’re doing: pricing visits, demo requests, repeat sessions. Score them as separate axes. A contact with high Intent and low Fit is a lead-magnet trap; high Fit and low Intent is a target account ABM signal. Routing logic should consider both.
Set decay deliberately
v2 lets you decay points over time. Default decay (50% over 30 days) is too aggressive for enterprise cycles and too soft for SMB. Configure per segment:
Engagement decay: SMB segment = 30 days half-life
Engagement decay: Enterprise segment = 90 days half-life
Fit score: no decay (firmographics don't change weekly)Re-route in shadow mode
Build the new routing workflow but set its action to notify_only for two weeks. Sales leaders see what would have routed where, without disrupting actual rep load. Compare to the live routing and tune.
Watch for the “score inflation” trap
v2 lets you add many more rules. Teams add 40+ rules in week one and every contact ends up scoring 80+. Cap rule count at 12 per axis. Force a quarterly prune.
Communicate the cutover date
Reps trust the score they’ve used for two years. Switching it overnight without warning destroys trust. Announce, document the new thresholds, and run a 30-minute training before flipping.
What to do this week
Spin up v2 in parallel against the legacy score. Pull the distribution comparison after seven days. Don’t touch routing yet.
