Plate № 55 · Operations · evidence census
A pattern from the gf.cx specimen book
The evidence-pipeline audit
A read-only page that classifies every item in a corpus on two axes — how far it got down the pipeline, and what it is — then shows the flow as stage cards and the outcomes as a funnel where colour carries meaning. A census, not a sample: it names what fell out, and moves nothing.
"Prove a corpus is whole and correctly processed by classifying every item on two axes — pipeline progress, and what it is or where it landed. Render the pipeline as plain stage cards and the outcomes as a colour-coded funnel where green→red is the meaning, not decoration. Surface the fallout openly; the page counts and names what's missing and moves nothing. Emit the data and the page from one generator, so the census is reproducible and can never go stale-green."
Worked example → amazon evidence pipeline Worked example → liked reels, audited
The two axes on one page — the pipeline reads left-to-right in plain cards; the outcome funnel carries the semantic colour (green landed · amber in-flight · blue set-aside · red lost). Nothing here has been moved or deleted.
The problem
You have a corpus you need to trust — every Amazon order you ever placed, every liked reel, every backed-up file. The question is never "roughly how many?" but "which ones aren't here, and why?". A spot-check or a single headline count can't answer that: a clean "876 routed" hides the 33 that stalled, the 80 that were too thin to card, and — worse — the items that never entered the pipeline at all. A sample proves nothing about wholeness; a total that doesn't name its own fallout is a number you can't act on.
The shape
- Two axes, every item. Classify the whole corpus — not a sample — on two independent axes: how far down the pipeline each item got (stage reached), and what it is / where it landed (category or outcome). Every item appears on both.
- The pipeline reads as plain stage cards. A horizontal row of white cards joined by arrows (
① → ② → ③ → ④), each a stage name, a count, and one line of what that stage means. This axis is a flow, so it stays monochrome — the eye follows the arrows, not a colour. - The outcomes read as a colour-coded funnel. A grid of buckets where colour is the meaning: green landed · blue set-aside/informational · amber in-flight/warning · red lost. Each bucket's top-bar matches its numeral, so the status is legible before you read a word.
- Read-only — it names the fallout, moves nothing. The audit surfaces what fell out (the stalled, the thin, the duplicate, the missing) rather than hiding it behind a clean figure. It never deletes or relocates; a "plan only" dedup list is a proposal, not an action. The footer says so: "nothing here has been moved or deleted."
- One generator, both outputs. A single script emits the machine census (
audit.json) and renders the page from it — with--render, a--checkdrift-guard, and a--selftest. The page can never disagree with the data, and a zero/empty walk aborts rather than printing a false "clean". - Managed type, on the almanac base. Link
almanac.cssfor the warm-paper tokens and prose spec, then hold the type discipline: serif-italic 400 title with an accentem, 700 uppercase kickers, serif-600 numerals, sans body. The header sits in the content column (nav → kicker → title → lede), never a full-bleed band, so it aligns with the pipeline and funnel by construction.
Why it works
The two axes answer two different questions at a glance — did it get processed? and what became of it? — without a table to read. Reserving colour for the outcome funnel makes status pre-attentive: red and amber are the items that need you, and they're visible before any number is parsed. Because the page is a census rather than a sample, the fallout buckets are the finding; and because one generator owns both the data and the render, the audit stays honest build-over-build — there's no hand-maintained figure to drift away from the truth.
When it breaks
- Colouring the pipeline. Giving each flow stage its own hue is decoration, not meaning — it competes with the funnel, where colour actually signals good-to-lost. Keep the flow monochrome.
- A band and an in-column header. Running a full-bleed masthead band above an in-column nav/kicker puts the title in a different column than the body — they visibly fail to line up. Pick the in-column header; it aligns for free.
- Sampling instead of censusing. "I checked a hundred and they were fine" can't prove wholeness. If you can't classify every item, you don't have an audit — you have an anecdote.
- Hiding the fallout. A page that shows only the clean count has thrown away its reason to exist. The stalled, thin, and missing buckets are the point; surface them.
- A hand-edited number. The moment a count is typed into the HTML instead of emitted by the generator, the census can lie. Emit everything; let
--checkfail the build on drift. (See edit the generator, not its output.)
Where it applies
- Use for any corpus you need to prove is whole and correctly processed: order-evidence archives, an enrichment or ingest feed, a media library, a backup or sync set, a document pipeline.
- Use when the valuable answer is the exceptions — what stalled, what's missing, what's duplicated — not the headline total.
- Skip for a live operational dashboard with moving real-time numbers (that's monitoring, not a census), and for a corpus small enough to eyeball in a list.
Reusable elements
- The two-axis frame — pipeline progress × what-it-is/outcome, every item on both. The single idea that makes "where did everything go?" answerable at a glance.
- Pipeline stage cards —
.pipe/.pstage: a monochrome left-to-right flow of white cards and arrows; stage name, count, one-line meaning. - The semantic outcome funnel —
.fbucketwith a top-bar colour that matches its numeral (t-okgreen ·t-infoblue ·t-warnamber ·t-badred): status before words. - Read-only census discipline — surface the fallout, move nothing, "plan only" for any proposed change, and a loud footer that says nothing was touched.
- One-generator reproducibility — emit
audit.jsonand the page from the same script;--render/--check/--selftest; abort on an empty walk rather than print a false clean. - In-column editorial header on the almanac base —
almanac.csslinked,.pageoverridden in one line; serif-italic title, 700 kickers, serif-600 numerals, header in the content column.
Reference
- Worked examples
- home.gf.cx/amazon/audit — "Evidence pipeline" (orders: present / reunited / missing / lost) · ebay/audit · growth.gf.cx/likes/audit — "Liked reels, audited" (pipeline + where-clips-land + tagging quality).
- Generators
pa_purchases_dashboard.py(render_audit) for the amazon / ebay evidence pipelines;likes_audit.py(--render/--check/--selftest) for the liked-reels census.- Origin
- 2026-10-31 — the order-evidence audit ("where do 3,864 orders fall out?") and the liked-reels audit converged on the same two-axis, colour-funnel, read-only shape; promoted to a pattern so the next audit surface inherits it instead of reinventing it.
- Kin
- edit the generator, not its output (the census is emitted, never hand-typed) · one base layer, linked, never copied (the warm-paper base this page is built on) · keep the probes out of the reader's numbers (verification that doesn't contaminate the thing measured).