Any metric · across versions

Know if a version is ready before you ship it.

Build time, binary size, latency, accuracy, test counts: any number you can measure, pushed in one call and compared across versions, products and targets.

Start the free trial →Read the docs

30 days free, no card needed. Flat price per org. We never see your source or your binary.

Interactive · sample data

One metric, everywhere

How does it compare across products and components?

binary_size_mb · MBcamera-app4.54 MB-9.5%

Click a point for its commit and pull request.

How it works

One HTTP call, from CI or anywhere else. No agents, no infra.
CI job or scriptany number you measurePOST /v1/pushvalue · version · tagsDriftmetricsruns · tags · historyDashboardtrend · split · comparePush responseeach value vs. previous runShip itor fix it firstCI job or scriptany number you measurePOST /v1/pushvalue · version · tagsDriftmetricsruns · tags · historyDashboardtrend · split · comparePush responseeach value vs. previous runShip itor fix it first
01 · push
Push the numbers

Any number you measure, from CI or a script, tagged with version and target.

02 · see
See the drift

Every metric across versions, split by the tags you push.

03 · act
Act before you release

Compare any two versions and see what moved, and in which artifact.

Use cases

Whatever you ship, push the numbers that drive your decisions.
push
model size, peak RAM, latency per inference, accuracy
slice by
model, quantization, runtime, device
decide
Pick the model that fits the device, and see which release made it slower.
Interactive example · sample data

One product, every metric

Is anything drifting in this release?

  • binary_size_mb · MB5.02 MB+11.8%
  • peak_ram_mb · MB237.6 MB+14.1%
  • latency_ms · ms34.04 ms+1.2%
  • accuracy · %89.51%-1.3%
  • ci_build_time_s · s490 s+12.4%
  • coverage · %75.9%-0.3%
  • unit_tests529+2.3%
  • integration_tests50+2.0%

v2.6.0 against v2.5.0, every platform together.

Two versions, one target

What moved between these releases, and in which artifact?

metric · runv2.5.0v2.6.0Δ
peak_ram_mbinference-engine325.5 MB409.7 MB+25.9%
binary_size_mblibvision.so6.62 MB8.11 MB+22.5%
ci_build_time_swhole app436 s490 s+12.4%
accuracyinference-engine90.5%86.9%-4.0%
unit_testswhole app517529+2.3%

One row per run of camera-app on android-arm64, biggest moves first. The same table the compare view shows for any two versions.

We never see your source or your binary.

Driftmetrics takes numbers (a metric name, a value, a few tags) and your CI keeps everything else.

  • no source
  • no binaries
  • no artifacts
  • no build logs

Flat price per org. Never per seat.

Everyone reads the numbers, so we never charge by the head.
One plan
$40per month, per org
  • Unlimited seats, projects and pushes
  • Trends, tag splits and two-version diffs

30 days free, no card needed. When the trial ends the org pauses until it's enabled; nothing is deleted. Need invoicing, a DPA or self-hosting? Talk to us.

Start the free trial →

Talk to us

Questions about your setup or a plan for your team: write to us.

Email info@driftmetrics.watchLinkedIn