Skip to content
LexDataLexData
PlatformIndustriesCustomers
DocsThe Field GuideBlogWhy models drift
AboutCareersSecurityContact
Log inStart now

// Chapter 04 · Day 90

What a confidence score actually means

The number beside every detection is a ranking signal, not a promise. Reading it correctly is the difference between a queue you trust and a threshold you fight.

7 min read·The field guide

In this lesson
  1. 01A ranking, not a probability
  2. 02Modern models are overconfident
  3. 03A threshold is an operating point
  4. 04What the score is actually for

In this lesson

  1. 01A ranking, not a probability
  2. 02Modern models are overconfident
  3. 03A threshold is an operating point
  4. 04What the score is actually for
01

A ranking, not a probability

A detection tagged 0.9 has not told you there is a ninety percent chance it is right. The score is what remains after training losses and post-processing have shaped it: a number that orders this detection against that one, on data that looks like the training set. As an ordering it is genuinely useful. As a probability it is a misreading.

The distinction earns its keep the first time someone sets a business rule on it. “Auto-file anything above 0.8” treats the score as a chance of being correct. What 0.8 means depends on the class, the camera, and the season - the same score can be near-certain for one class and coin-flip for another.

02

Modern models are overconfident

This is measured, not folklore. As networks got deeper and more accurate, their scores got less honest - a landmark 2017 study found modern architectures systematically overstate their certainty, where the shallower networks of decades earlier were roughly trustworthy. Accuracy went up; calibration went down.

It gets worse exactly when you need it most. Under distribution shift - the drifted conditions the previous lesson maps - models do not reliably become hesitant. They are often confidently wrong on inputs unlike anything they trained on. A score calibrated against your validation set is not calibrated against next winter.

This is why the drift lesson watches the confidence distribution for movement rather than trusting any single score. Shape change is signal; the number alone is not.

03

A threshold is an operating point

Every threshold you set chooses a position on a curve: catch more and flag more falsely, or flag less and miss more. Neither side is free, so the choice belongs to the acceptance sentence from chapter one - how many false alarms a shift can absorb, and what a miss costs.

Base rates do the surprising work here. A detector that is right 95 times in 100 sounds finished - point it at a defect that occurs once in a thousand frames and most of what it flags will still be false alarms. Nothing about the model got worse; the arithmetic of rare events is just unforgiving. Set thresholds against the real frequency of the thing you are hunting, not against the score distribution alone.

04

What the score is actually for

Spend it as an ordering. Sort the review queue by confidence and human attention lands where the model is least sure, which is where it earns the most. Watch per-class score distributions over time and you have the cheapest early drift signal there is.

What the score is never for: skipping verification on the classes that matter. A high score on a safety-critical class is a strong hint, not a verdict - the expensive failures in this field are the confident ones.

// sources

  • Guo et al., On Calibration of Modern Neural Networks - ICML 2017 ↗
  • Ovadia et al., Can You Trust Your Model's Uncertainty? - NeurIPS 2019 ↗

Next in the guide

How to catch model drift before your customers do7 min→

Start with your footage.

Start with your footageBack to the guide →
LexData
LexData

Product

  • Platform
  • Industries
  • Use cases

Resources

  • Docs
  • The Field Guide
  • Blog
  • Why models drift

Industries

  • Energy & utilities
  • Oil & gas
  • Agriculture
  • Manufacturing
  • Insurance
  • Retail
  • Robotics

Company

  • About
  • Customers
  • Careers
  • Contact

Trust

  • Security
  • Privacy
  • Terms

Stay updated

What we learn running vision models in production.

See everything.
Miss nothing.

Stay updated

What we learn running vision models in production.

Terms of use & Privacy policy

© 2026 LexData Labs · All rights reserved