Skip to main content

Calibration Overview

Calibration helps you set scoring weights based on what your data actually shows, not just what you guess. kenbun looks at how each rule correlates with real conversions in your account, then suggests a weight that reflects how much that signal lifts a lead's chance of converting.

What Calibration Does​

A scoring rule says "give this lead some points when they do X". Calibration answers a sharper question: how much do leads who do X actually convert, compared to leads who don't? That ratio (the "lift") is what really tells you whether the rule is pulling its weight.

For each rule in your engagement, profile, account, and deal dimensions, kenbun tracks:

  • Sample size: how many leads have triggered this rule
  • Baseline conversion rate: how often a typical lead in your account converts
  • Observed lift: how many times more often the rule's leads convert vs. baseline
  • Confidence: how trustworthy the estimate is, given the sample size

It then suggests a weight that's proportional to the lift. High-lift rules (e.g., "requested a demo" — leads convert 6x baseline) get bigger weights; low-lift rules (e.g., "viewed a blog post" — barely above baseline) get smaller ones.

Why It Matters​

Most teams set scoring weights by intuition — a workshop, a whiteboard, a guess. Six months later, nobody can remember why "Pricing Page Visit" is worth 15 and "Webinar Attended" is worth 25. Worse, the numbers might be wrong: maybe webinar attendees never close, and pricing visitors close all the time. Without data, you'd never know.

Other platforms make you export your data, build a pivot table, compute lift in a spreadsheet, and translate the result back into weights yourself. kenbun does the math for you and puts the result inline next to every rule, with a one-click Apply.

The kenbun Approach​

Calibration runs automatically on a schedule (weekly, by default). For each rule, you'll see:

  • The current weight you've set
  • A suggested weight based on observed lift
  • The data behind the suggestion: sample size, conversion counts, lift, and confidence

You decide whether to apply. Nothing changes until you click Apply suggested.

You Stay in Control​

kenbun never auto-applies suggested weights. Every change goes through a confirmation dialog that shows you the full rationale: the conversion event being measured, the lookback window, the sample size, the lift, and the confidence band. Every Apply is recorded in your rule audit log so you (and your team) can trace exactly when and why a weight changed.

This matters for two reasons. First, calibration is a recommendation, not a rule — your business judgement still has the final word. Second, when something goes sideways three months from now, you'll want to know who changed what and on what basis.

Quick Tour​

Calibration is documented across five short pages:

  • Conversion Events — choose what counts as a "conversion" for your account. This is the input that drives every suggestion.
  • Cold Start — what to expect in your first weeks before calibration has enough data.
  • Applying Suggested Weights — how to review and apply suggestions in bulk.
  • Co-occurrence Warnings — what the amber triangle on a rule means and when to act on it.

Where to Find It​

  • Inline panel: open any scoring rule (engagement, profile, account, or deal) and the calibration panel appears in the slideout, right next to the weight field.
  • Bulk view: Configure > Scoring > Calibration Review shows every calibrated rule across dimensions in one sortable table.
  • Conversion event setup: Govern > Org Units — set or change your conversion definition.

Best Practices​

  1. Set your conversion event early. The sooner kenbun knows what "good" looks like in your account, the sooner suggestions become useful. See Conversion Events.
  2. Don't apply blindly. A suggestion is only as good as the data behind it. Check the confidence band and sample size before clicking Apply.
  3. Watch for confounded rules. If two rules trigger on largely the same leads, the lift estimate for each is inflated. Look for the amber triangle — see Co-occurrence Warnings.
  4. Re-calibrate after big changes. New ICP, new product line, new pricing? Your old weights probably no longer reflect reality. Calibration will catch up automatically over a few weeks of new data.

Troubleshooting​

"I don't see any suggestions." You may be in the cold-start period, or your conversion definition may not be set. See Cold Start.

"The suggestion is way off from my intuition." Trust the data, but verify it. Check the sample size and confidence. If both are high, your intuition may be the thing that needs updating. If sample size is small, wait for more data before acting.

"Suggestions stopped updating." Calibration runs weekly. If your conversion event was changed recently, the next refresh will recompute everything from the new ground truth.

How kenbun Keeps Suggestions Trustworthy​

kenbun ranks your conversion events from strongest to weakest signal of real buying intent, and weights stronger outcomes more heavily when learning what behavior predicts a deal. To avoid overreacting to a small number of conversions, kenbun blends each suggested weight toward a sensible default until enough data accumulates. The more conversions kenbun observes, the more the suggestions reflect your actual data rather than the default. This is why early suggestions are conservative and become sharper as your history grows.