The lineage
Sixty years of measuring engineers.
Every era invented a number to capture developer productivity. Every number that got pointed at individuals was gamed — and abandoned. Here is the pattern, and where TIER fits.
The record is remarkably consistent: a metric appears, it's useful as an aggregate diagnostic, someone attaches it to individual evaluation, and within a cycle or two it inflates, distorts, and is discarded. This is Goodhart's Law, on schedule.
Lines of code
The first industrial productivity unit. KLOC drove progress reports, contractor billing, estimation models like COCOMO — and, in some shops, individual quotas. More lines meant more "output."
"If we wish to count lines of code, we should not regard them as 'lines produced' but as 'lines spent.'"Edsger Dijkstra, EWD1036, 1988
In 1982 the Lisa team asked programmers to report weekly lines of code. The week Bill Atkinson rewrote QuickDraw's region engine to be six times faster and 2,000 lines shorter, he wrote −2000 on the form. They stopped asking him to fill it out.
Failure mode — rewards volume, punishes deletion and reuseFunction points
Allan Albrecht at IBM fixed the obvious flaw — LOC wasn't comparable across languages — by sizing software by delivered functionality: inputs, outputs, inquiries, files, interfaces. Standardized by IFPUG, adopted in contracts and government outsourcing.
But counting was laborious and subjective, it fit transaction systems better than modern work, and it never became a clean per-person measure. A better sizing unit — not an individual scoreboard.
Failure mode — too subjective and costly to run per developerStory points & velocity
Agile's relative-estimation unit. Velocity — points completed per sprint — was designed for one thing: a team forecasting its own throughput. Then managers noticed it was a number that trends up, and started comparing people with it.
The result is textbook: estimates inflate, work is split to book more points, hard-but-valuable problems get avoided, and cross-person comparison corrodes trust. The closest historical analogue to any single AI-yield metric.
Failure mode — inflates the moment it's tied to evaluationDORA · SPACE · DX Core 4
The modern frameworks learned the lesson. DORA's four keys, SPACE's five dimensions, and DX Core 4 were built by their own authors (Forsgren, Storey, Zimmermann, Noda, Tacho and others) to be deliberately team- and system-level.
Their guidance is explicit and unusual: do not use these to rank or pay individual developers. SPACE is multidimensional specifically to resist single-number misuse; DX Core 4 says measure per-engineer-flavored metrics at the team level, never the individual. The guardrail is the point.
Design choice — aggregate by construction, to prevent the gamingGoodhart & Campbell
Why the pattern never breaks. Charles Goodhart (1975), in Marilyn Strathern's phrasing:
"When a measure becomes a target, it ceases to be a good measure."Goodhart's Law · Strathern, 1997
Campbell's Law says the same for social indicators: the more a quantitative measure drives high-stakes decisions, the more it is corrupted, and the more it distorts what it was meant to monitor. LOC, function points, velocity — all of it, on schedule.
TIER — the yield of AI spend
AI put a new, real, measurable cost input next to the work: tokens, in dollars. TIER names what none of the others could — outcome per dollar of AI spend — and it is deliberately built on the responsible side of this history.
In its anonymized modes, team and quarter is the served unit and a k-anonymity floor hides small cohorts; and TIER ships its own 28-scenario failure-mode analysis, including the Goodhart warning, with the code. It grades the work — so people can raise their own yield — not the worker.
The lesson, applied — the metric that learned from the ones before itThe honest read
The history is a moat, not a liability.
Every metric that got attached to individual evaluation was gamed and buried. Every framework that stayed aggregate — DORA, SPACE, DX Core 4 — endured and became a standard. TIER is the next chapter in that second lineage: the AI-era yield metric, extended to the new cost dimension, carrying the guardrails the first lineage learned the hard way.
Where TIER stands
Every prior metric graded the developer — and got gamed. TIER grades the work, and ships its own failure-mode analysis to prove it.
Sourcing note: Dijkstra (EWD1036) and the Atkinson "−2000 lines" account (folklore.org) are primary-verified; DORA/SPACE/DX Core 4 authorship and their team-level guidance are from the primary papers; Goodhart's phrasing is Strathern (1997). The "aircraft weight" line often attributed to Bill Gates is widely circulated but its original source is untraced — omitted here rather than misattributed.