---
title: Scoring Sinks The Plumbing
type: anti-pattern
level: L3
status: live
revision: 1
updated: 2026-08-14
systemVersion: 4.2
tags: [prioritisation, scoring, planning]
rating: 8.10
ratingAxes: useful 7 · evidence 9 · pull 8 · original 9 · form 8
ratingKind: derived
source: pattern log PAT-061, 2026-08-14
---

# Scoring Sinks The Plumbing

_Written 2026-08-14 · last verified 2026-08-14 · system v4.2 · live_

**TL;DR** — An idea-scoring pass ranked distribution surfaces first through third and the publishing pipeline ninety-third. The scoring axis rewarded reader-visible impact, and infrastructure is invisible by definition. Ordering by score would have built the shop window before the shop.

## Pattern

You generate a large field of ideas and score them on something reasonable — impact, reach, pull. The ranking comes back and you start at the top.

In one run of 120 ideas, the reader-facing surfaces took positions one, two and three. The publishing pipeline came ninety-third. The safety gate that has to run before anything is published came eighty-fourth. The scheduled generator whose absence had already killed the previous version of the project came seventy-third.

## Why it looks right

The scores are not wrong. A reader really does experience the surface and never experiences the pipeline. Any axis that asks *"how much will this matter to the audience"* will rank infrastructure last, correctly, every time.

The ranking is also the one artefact everyone trusts, because it looks like the output of a method rather than an opinion.

## Why it fails

The bias audit that caught it recorded the finding in one line:

> `Distribution surfaces ranked 1-3; publishing pipeline 93rd, safety gate 84th, scheduled generator 73rd.`

**Execution order is set by dependencies, not by score.** The window cannot precede the shop. Worse, the items that score lowest are systematically the ones whose absence is fatal rather than merely disappointing — a missing safety gate does not produce a weaker launch, it produces no launch at all.

An impact axis cannot see this, because the cost of a missing dependency lands on items other than itself.

## Instead

Score to find what is worth doing. **Sequence separately, by dependency.** Two passes, written down as two different lists, so nobody mistakes one for the other.

Then run a deliberate check on the bottom of the ranking before discarding it: *which of these low scorers is a precondition for a high scorer?* In the run above, that question moved four items from the bottom quartile into the first phase. A bias audit caught it — but only after the ranking already existed and had started to feel authoritative.