Operator Preset

Pick the scenario closest to your platform. Each preset rebalances dimension weights for the operating context.

Dimension Weights

Each dimension contributes 0 to 100 to the weighted score. Total weights should sum to 100. Adjust to match what drives supply quality in your model.

Flagship feature

Tenure-based bias correction

Drivers under 30 days on the platform get score smoothing toward the portfolio median. Avoids penalizing new drivers with thin data while still surfacing genuinely weak performers. Most fleet scorecards skip this step.

How dimension scoring works

Each driver is scored 0 to 100 on each dimension. The weighted total is the sum of (weight × dimension score) divided by total weight. Scoring rules:

Acceptance Rate %: raw % becomes score directly. 85% acceptance = 85 points.
On-Time Delivery %: raw % becomes score directly. 92% on-time = 92 points.
Customer Rating (1 to 5): normalized to 0 to 100. 4.8 stars = 96 points. Anything below 3.5 = 0 points.
Completion Rate %: raw % becomes score directly. Share of accepted orders that completed without refusal or failure.
Response Time min: normalized against best driver. Best = 100, each added minute = 5 point penalty.

When bias correction is ON, drivers with under 30 days tenure get a 50/50 blend between their raw weighted score and the portfolio median. The shorter the tenure, the more weight on the median. Drivers with 30+ days = full raw score.

Drivers

Enter driver-level performance from the last 30 days. Tenure column drives bias correction. Drivers with low tenure show with NEW flag in the ranking.

Driver	Tenure days	Volume deliveries	Accept %	On-Time %	Rating 1 to 5	Complete %	Response min	remove

Fleet Snapshot

Tier Distribution

Top = top 20%, Reliable = next 40%, Watch = next 20%, At-Risk = bottom 20%. Bar length shows delivery volume share within each tier.

Ranked Drivers

Sorted by weighted score. NEW = under 30 days tenure, bias-corrected when toggle is on.

#	Driver	Accept	On-Time	Rating	Complete	Response	Score	Tier

Action Recommendations

Generated from tier classification + tenure + score. Actions for new drivers favor support and onboarding completion. For tenured drivers, accountability and replacement.

Use cases I've seen this work

Patterns from courier and dispatch ops I've run before. Adjust to your specific platform. The operating logic transfers.

Case 1: Instant delivery driver onboarding leak

You're onboarding 500 drivers a month. The dashboard says 500 onboardings. The P&L says 220 of them never delivered a second order. That gap is where the money disappears.

How this maps: use the "Instant delivery" preset. Run the scorecard with tenure-based bias correction ON. New drivers get a chance to ramp, but the ones with poor signal in the first 30 days surface in the Watch and At-Risk tiers. The onboarding funnel becomes measurable instead of vibes-based.

The play: targeted support intervention on Watch-tier new drivers. At-Risk new drivers either get a structured remediation or get cut before they cost more onboarding investment.

Case 2: Gig courier supply during peak event

Black Friday or game-day surge. You have 8,000 active couriers but only 1,500 are reliable. Routing high-value orders to the wrong driver during peak = refunds, churn, and a P&L hit.

How this maps: use the "Gig courier platform" preset. On-time delivery + completion rate weighted heaviest. Run the best-gets-first-call simulation to confirm what share of peak volume your top drivers can absorb.

The play I ran at Rappi: tier-based dispatch during peak events. Top tier drivers got first call on high-value orders, At-Risk drivers got low-stakes orders or were paused. Refund volume during peak dropped because the right driver got the right order.

Case 3: Driver retention through earnings stability

Top drivers earn more because they get more orders. But when dispatch is opaque, even top drivers feel inconsistency, and they churn for the next platform.

How this maps: use the scorecard to make tier-based dispatch transparent to drivers. Top tier sees their tier publicly. Reliable tier sees what to improve to graduate. Visible meritocracy is a retention tool.

The reframe: retention through earnings stability isn't just "pay them more." It's "let them see why they earn what they earn, and how to earn more." The scorecard becomes the conversation.

Case 4: Bias correction for new drivers (the structural insight)

Most fleet scorecards penalize new drivers with thin data. A driver with 5 deliveries and one bad rating looks worse than a driver with 500 deliveries and 4 bad ratings. That's a math artifact, not a quality signal.

How this maps: bias correction blends new-driver scores toward the portfolio median. The driver needs to PROVE they're below median, not just have low confidence from thin data. You stop firing drivers your scorecard couldn't measure yet.

The structural insight: the same applies to vendor scorecards, content moderator QA, fraud analyst ranking. Anywhere you score people on early data, bias correction is the line between fair decisions and structural unfairness.

Driver / Courier Performance Scorecard