Knowledge risk brief

Generated 27 August 2026, 14:13 UTC across 220 people, indexed from codeowners, confluence, git, github, jira, org-csv, pagerduty, slack.

Summary

Where this company's knowledge rests on too few people. Each figure below jumps to the detail behind it.

Concentrated subjects (all 25)

Where an expert is marked not lately, they know the subject but have stopped working on it. That is a sharper finding than concentration alone: measured against a real project, a lead who had already stopped was less than half as likely to still hold the subject six months on.

A subject is concentrated when most of its expertise sits with one or two people. Bus factor is how few people together hold the majority of it, so a bus factor of one means a single departure takes the topic with it.

Topic Level Bus factor Top expert holds Who holds it Also covers
kafka-lag critical 1 83% Stanley Lewis (83%), Dinesh White (17%)
billing-retries critical 1 98% Stanley Lewis (98%), Dwight Luciano (2%)
sso-login critical 1 100% Stanley Lewis (100%)
search-indexing critical 1 100% Lalo Goodman (100%)
data-warehouse critical 1 100% Walter Schrute (100%)
payroll-taxes critical 1 100% Michelle Schrute (100%)
contract-review critical 1 100% Carlton Halpert (100%)
certificate-renewal elevated 2 75% Stanley Lewis (75%), Hector Wexler (12%), Chalky Rosetti (12%)
vacation elevated 2 75% Stephanie McGill (75%), Holly Dunn (25%)
cdn-caching elevated 2 69% Stanley Varga (69%), Chalky Gilfoyle (15%), Holly Fring (15%)
api-ratelimits elevated 2 69% Lalo Vance (69%), Kevin Lewis (15%), Joey Butler (15%)
health-benefits elevated 2 47% Jared Capone (47%), Stephanie McGill (37%), Gale Malone (16%)
laptop-hardware elevated 2 67% Monica Hanneman (67%), Andy Malone (15%), Will Lewis (15%), Saul Varga (2%)
kubernetes-deploys elevated 2 79% Walter Bream (79%), Jim Schroeder (21%)
oncall-paging elevated 2 76% Steve Rosetti (76%), Will Varga (21%), Kevin Salamanca (2%)
hiring-interviews elevated 2 78% Laurie Schroeder (78%), Russ McGill (22%)
terraform-state elevated 2 72% Erin Hale (72%), Stanley Varga (28%)
image-uploads elevated 2 69% Meyer Wexler (69%), Lalo Vance (28%), Oscar Dunn (3%)
onboarding-paperwork elevated 2 72% Will Salamanca (72%), Monica Hanneman (28%)
database-migrations ok 3 45% Gale Malone (45%), Walter Bream (23%), Michelle Pinkman (15%), Gavin Hamlin (15%), Angela Thompson (2%)
mobile-releases ok 3 46% Richard Belson (46%), Steve Rosetti (23%), Erlich Hale (15%), Mike Bighetti (15%)
office-facilities ok 3 47% Todd Boetticher (47%), Laurie Schroeder (22%), Meyer Wexler (16%), Jimmy White (15%)
payment-webhooks ok 3 56% Meyer Bratton (56%), Lalo Goodman (24%), Kim Capone (20%)
feature-flags ok 3 55% Gale Schrader (55%), Walter Schrute (24%), Owen Fring (21%)
expense-reports ok 3 55% Gavin Flenderson (55%), Michelle Schrute (24%), Owen Dunn (21%)

Who is exposed (16)

The same finding read by person rather than by topic. Sole knowledge has no cover behind it at all. Led knowledge has others contributing, so it degrades rather than disappears.

Stanley Lewis — 1 sole, 3 led

Would leave with them

sso-login

Would lose its lead

billing-retries, certificate-renewal, kafka-lag

Carlton Halpert — 1 sole, 0 led

Would leave with them

contract-review

Lalo Goodman — 1 sole, 0 led

Would leave with them

search-indexing

Michelle Schrute — 1 sole, 0 led

Would leave with them

payroll-taxes

Walter Schrute — 1 sole, 0 led

Would leave with them

data-warehouse

Erin Hale — 0 sole, 1 led

Would lose its lead

terraform-state

Jared Capone — 0 sole, 1 led

Would lose its lead

health-benefits

Lalo Vance — 0 sole, 1 led

Would lose its lead

api-ratelimits

Laurie Schroeder — 0 sole, 1 led

Would lose its lead

hiring-interviews

Meyer Wexler — 0 sole, 1 led

Would lose its lead

image-uploads

Monica Hanneman — 0 sole, 1 led

Would lose its lead

laptop-hardware

Stanley Varga — 0 sole, 1 led

Would lose its lead

cdn-caching

Stephanie McGill — 0 sole, 1 led

Would lose its lead

vacation

Steve Rosetti — 0 sole, 1 led

Would lose its lead

oncall-paging

Walter Bream — 0 sole, 1 led

Would lose its lead

kubernetes-deploys

Will Salamanca — 0 sole, 1 led

Would lose its lead

onboarding-paperwork

One-person connections (8)

Subjects that only one person has ever done work across. They have their own experts, so nothing that counts experts per subject shows this: what rests on one person is not any of the subjects but the knowledge that they belong together, and whoever picks one up afterwards has no reason to look at the others. Where the crossings join up they are listed as the one body of work they are, since that is what would leave.

SubjectsOnlyPeople holding themWorked on together
billing-retries + kafka-lag Stanley Lewis 3 40%
certificate-renewal + sso-login Stanley Lewis 3 40%
api-ratelimits + image-uploads Lalo Vance 5 33%
cdn-caching + terraform-state Stanley Varga 4 33%
data-warehouse + feature-flags Walter Schrute 3 33%
database-migrations + kubernetes-deploys Walter Bream 6 33%
expense-reports + payroll-taxes Michelle Schrute 3 33%
health-benefits + vacation Stephanie McGill 4 33%

Joined work (1)

Subjects that get worked on together, where the same person leads every one of them. This is a heavier finding than a concentrated subject on its own: whoever takes the work over has to learn the whole of it at once, and a list of subjects one at a time never shows that.

Stanley Lewis — 4 joined subjects

billing-retries, certificate-renewal, kafka-lag, sso-login

How this was measured

Every figure here is arithmetic over the indexed graph, with no model involved and no judgement applied. A person's share of a topic is their weight in it against everyone else's. Bus factor counts the fewest people whose combined share passes a majority. A topic is critical when one person holds most of it, and elevated when two do.

The finding only sees what was indexed, which here was codeowners, confluence, git, github, jira, org-csv, pagerduty, slack. Knowledge that lives somewhere whodar was not pointed at does not appear, so this is a floor on concentration rather than a complete picture.