One graduate · one month · n = 50

Instrumented Job Hunt

2026-07-08 to 2026-07-28 · 19 days of coverageAs of 2026-07-21The full writeup

This page is descriptive, not inferential. At n = 50 applications, 65 operations and 19 days of coverage, nothing here supports a significance test, a confidence interval, or a causal claim - a single person's one-month sample doesn't carry that kind of weight. Every figure below comes from the committed, anonymised CSVs in results/, cell for cell.

§1 Headline

Tracked

50

Submitted

43

Rejected

15

No response yet

28

Interviews

0

"No response yet" is censored, not ghosted: these applications are still open as of 2026-07-21, not confirmed dead.

§2 Funnel

Of 50 tracked applications: 43 submitted, 5 closed by the employer before ever being submitted, 1 skipped, and 1 reached an outcome the export couldn't classify cleanly. Of the 43 submitted: 15 rejected, 28 with no response yet. ‘Employer closed’ conflates a posting withdrawn by the employer with one filled by someone else - the export doesn't distinguish the two.

§3 Time to rejection

10 of 15 rejections had both an applied date and a status date and could be timed; 5 are excluded from the chart for missing one of those dates - excluded, not dropped: they are still counted in the 15 above. Timed rejections ranged 0-5 days.

§4 Channels

Applications sorted into 6 channels: other 22, company_portal 8, linkedin 8, indeed 5, workday 4, jooble 3. LinkedIn shows a 0.0% rejection rate across 8 submitted applications (0 of 8 rejected). That is not a good result: nothing came back at all. Workday, by contrast, shows 3 of 4 rejected.

Per-channel outcomes, from results/channel_outcomes.csv - jooble has no submissions, so its rejection rate is undefined (shown as —, never 0%)
ChannelTotalSubmittedRejectedRejection rate
other22201050.0%
company_portal87114.3%
linkedin8800.0%
indeed54125.0%
workday44375.0%
jooble300

§5 Tiers

Entry: 3 of 11 rejected. Stretch: 1 of 6 rejected. Unspecified: 11 of 33 rejected - unspecified carries most of the volume because tier is derived from role titles, and most role titles don't state a level.

§6 Operations

65 operations across 6 categories: 41 closed, 22 open, 2 snoozed. Of the closed ones, 24 carry a machine-readable close date (17 do not); those that do took 0-6 days.

§7 Finance

Buffer moved from 66.0% of target on 2026-07-10 to 88.0% on 2026-07-24, with income-change markers on 2 distinct dates. No monetary amount exists anywhere in this repository, by design - only the percentage of a private target.

§8 What this cannot tell you

§9 Reproduce it

  1. export/*.csv - the anonymised, gate-scanned export.
  2. sql/*.sql - the analyses run over that export.
  3. results/*.csv - the outputs, gate-scanned again before being committed.
  4. charts/*.html - one standalone plotly document per chart, rendered from results/ only.
  5. This page - generated from results/ only by scripts/build_dashboard.py, re-rendering nothing.

Coverage: 19 days (2026-07-08 to 2026-07-28). Inbox activity was logged on 8 of those days (191 messages seen, 7 suppressed as sensitive before export).