· scanning teams at

The bone scan. For your engineering org.

See what's amber before it goes ember.
Marrow fuses repo, on-call, and team signal into one diagnostic — so engineering leadership reads the drift before it ships.

payments-api — amber drift, 14mo since last refactor

You already suspect it.

An engineer who used to push twice a day now pushes twice a week. A service hasn't been refactored in fourteen months and three of the four people who'd know how are gone. Two engineers carry seventy-one percent of the on-call rotation and don't say it out loud. The #payments-platform channel has gone quiet in a way that isn't peaceful.

None of this shows up in Datadog. None of it shows up in Lattice. It will show up — in a quit, in an outage, or in a rewrite.

Marrow shows it now.

Read your scan →

T7 · push cadence ↓ 61% T7 · on-call load 71% (2 of 14) T7 · refactor +14mo

One instrument. Three signals. One reading.

Repo signal. Refactor cadence, ownership distribution, silent SLO drift, the modules everyone writes around.

On-call signal. Page distribution, incident clustering, the engineers carrying the rotation and the engineers who don't.

Team signal. The shape of communication — response cadence, channel quietness, the patterns that precede a departure. Patterns only. Never message content.

Fused per service into a single reading: bone, amber, or ember.

L2 · billing-svc · refactor cadence steady · on-call balanced · channel responsive

The leader who saw it first.

Engineering-org drift used to be invisible until the quit, the outage, or the rewrite. The signals existed — repo cadence, on-call distribution, the silence in #payments-platform — but no instrument fused them.

Marrow does.

The CTOs running scans this quarter are the ones walking into the board meeting with a slide labelled intervention plan instead of post-mortem. The dignity of being the leader who saw it first is the product.

See a sample report →

One-page redacted Marrow scan report: a service marked AMBER with three contributing signals and an intervention note.

Currently scanning.

127

engineering orgs already scanned

*plausibility marks, not customer claims. The real customer wall ships when the real customers do.

What the instrument reads. And what it doesn't.

Do you read my team's messages?

No. Marrow reads the shape of your team's signal — cadence, distribution, response latency, channel quietness. Never message content. Never any individual person's words. The instrument reports patterns; it does not transcribe.

Where does the data come from?

GitHub or GitLab for the commit graph and ownership. PagerDuty or Opsgenie for rotation and incident distribution. Slack or Teams for channel metadata only — message bodies stay on your servers and are not transmitted to Marrow.

Who sees the scan?

Only the engineering-leadership seats you license. No individual contributor is ever named in any report. Service-level and team-level aggregates only. The scan is for the people who can act on it.

What does the report look like?

One page. Three to seven vertebrae per service. A bone, amber, or ember reading and a single sentence per finding. Designed to print as a board slide, not a dashboard.

How fast is the first scan?

First reading ships within 24 hours of connection. Subsequent scans run weekly. The instrument is meant to be read on Monday morning, not stared at on Friday night.

Request your bone scan.

One repo. One on-call rotation. One channel id.
First reading in 24 hours.

The instrument runs in the dim room whether you're watching or not — the scan happens; the question is whether you're the one reading it.