Ten numbers nobody computes

Ten decisions an operations team settles every quarter, and the number each one actually turns on. Put your own volumes in below: the arithmetic runs in your browser, and nothing is sent anywhere.

Arslane Chaouche Ramdane, business operations and AI transformation, open to US-remote roles.

Your figures

Detection

The first two fields are enough for a demonstration. The next two (your alerts for the year, and the share that turned out to be founded) refit the model to your system instead of this one's.

Have an export of scored cases? Two columns are enough: the score, and whether the case turned out to be founded. Nothing is uploaded: the file is read here, in this page.
Process time

Your volume and your hourly cost are enough. An event log (case, activity, timestamp) replaces this repository's process with your own.

An event log? Three columns: case, activity, timestamp. A fourth (minutes actually worked) separates work from waiting.
Onboarding triage

What a human currently spends opening files a rule could settle.

The funnel

Your raw volumes, step by step. Everything that comes out is computed on those alone, including the confidence interval that says whether your weakest step can really be named.

The verdict

The ten tools

01 The economics of a detection threshold

7 of 8 analysts are paid to handle nothing, and the next true positive is free down to threshold 0.45.

source 7/8analysts idle
02 Onboarding triage

58 % of files decided without a human, and 0 uncontrolled onboardings: the agent stops where it is not sure.

source 0breach in 400
03 Where the funnel leaks, and what that is worth

The biggest leak is not the best place to spend: 31.14× against 2.1× per dollar, a factor of 15.

source 15×between best and worst lever
04 Growth versus controls, priced instead of argued

The A/B test settles the lift (2.1 % [0.15 – 4.04]) and not the decision. The sign flips at an undetected-risk share of 1.33 %, inside the range both sides defend.

source 1.33 %where the verdict flips
05 Where should the next dollar go?

Sending every field to the large model reaches 78.8 % for $800. Routing field by field reaches 94.4 % for $191: better and 4.2× cheaper, because 3 of the 5 fields are carried by regexes.

source 4.2×cheaper, and more accurate
06 The order is the plan

Taking the worst finding first misses 3 of 8 deadlines and costs $695,000. The identical work sorted by deadline misses none. Same team, same effort, only the order changes.

source $695,000the reflex costs
07 The drift threshold sits above the signal

Every note says to alarm at 0.2. A 0.3σ shift moves the index to 0.09: the alarm is above the signal it exists to see. Below 350 observations a check, no threshold separates noise from that shift at all.

source 0.09signal, against a 0.2 alarm
08 Eleven days, and nobody worked on it for nine

6.4 days end to end, of which 2.3 hours of actual work. 95 % of the time nobody is touching the file.

source 95 %of the delay is waiting
09 Regression bench

Across 3 deterministic versions, 13 → 19 of 22 cases. A rising pass rate can still hide a case that just broke.

source 13→19cases out of 22
10 Document search

On 25 questions: 10 right, 9 wrong, and 6 times it said nothing, every time no answer existed in the corpus.

source 6justified silences

Two apps, built end to end

The tools above are analysis. These are products: designed, built, and taken to the point where the remaining work is Apple's review queue rather than mine.
Atlas, a strength-training companion. Flutter, English and French, everything on the device. at 1.0
Hisho, an on-device habit tracker organised around needs rather than streaks. No server, no account. in build

Neither is on the App Store yet, and neither is open source, so neither carries a link here. They are named because ten measurement tools do not say whether the person can ship a product.

Arslane Chaouche Ramdane, business operations and AI transformation, open to US-remote roles. Written in TypeScript with no runtime dependencies: clone any repository, run npm test, and every figure regenerates. The demonstration data is synthetic and seeded; the regulation cited is real (31 CFR Chapter X).