First-Time-Right Rate

An operations manager reports that the team completes 1,000 cases a day and needs headcount for 1,100 next quarter. The underlying system shows 1,350 case-touches a day. Nobody is lying: 1,000 cases complete, and 350 of them pass through a stage twice. The capacity plan has been built on the wrong number for two years, and every hiring round has been slightly short.

First-time-right rate is the share of cases completing with no re-entry, and its usefulness comes from how it compounds. Six handoffs each rejecting 1 in 20 gives 0.95 to the sixth power, about 74%, so roughly a quarter of cases are reworked even though every individual stage looks excellent.

Why it matters

For an architect the metric does three things. It sets the capacity multiplier: at 74% first-time-right the system must be sized for about 1.35 times nominal volume, which changes worker pool sizing, queue depth limits and database write volume. It locates where to put validation, because a defect caught at handoff one costs one re-touch and the same defect caught at handoff six costs six. And it predicts lead time better than throughput does, since rework returns late to a queue already sized for new work.

The compounding is the part teams miss. Arguing for a better stage is usually worth less than removing a stage: going from six handoffs to four at the same 5% rejection rate lifts first-time-right from 74% to 81% and drops the multiplier from 1.35 to 1.23.

Implementation patterns

• Count re-entries, not completions. Instrument each stage with an arrival event carrying the case identifier and an attempt number. Without the attempt number the rework is invisible in throughput figures.
• Publish the per-stage rejection rate alongside the compounded rate, so the conversation is about which handoff to fix rather than about the aggregate.
• Validate as early as the information exists. A field that can be checked at intake should never be checked at approval. This is the cheapest single intervention and it is usually blocked by the intake system being owned by a different team.
• Give rework its own queue and its own service-level target. Mixed into the new-work queue it is invisible and it ages badly, because the people able to handle it are the busiest.
• Set a rejection-rate threshold that triggers design work, for example any stage above 8%, rather than reviewing the process annually.
• Model the catch-up explicitly. A stage at 80% utilisation absorbing a 26% rework stream has far less slack than its utilisation figure suggests, and queue times rise sharply as it approaches capacity.

Industry example

Order-to-cash flows in manufacturing and distribution show the pattern plainly: credit checks, pricing approval, stock allocation, pick confirmation, invoice and settlement each reject a small share, and the compounded figure is routinely between 60% and 80% in businesses that believe each stage is fine. The arithmetic is not new — it is the rolled-throughput-yield logic that manufacturing quality programmes have used since the Six Sigma methods of 1986, applied to a process whose stages happen to be software and people rather than machines. The visible symptom is not capacity but a long tail in lead time, with a small group of experienced people resolving everything that falls out.

The same arithmetic governs automated processes. A pipeline of six services each failing 1 in 20 calls and retrying produces the same 26% of work done more than once, paid in compute and in tail latency rather than in people.

Failure scenarios

• Capacity planned from completions, so every forecast is short by the rework multiplier and the shortfall is attributed to productivity.
• Rework hidden inside a stage. A team that quietly fixes and resubmits rather than rejecting formally keeps the measured rate high and the real one unknown.
• Automating the stage with the highest volume rather than the one with the highest rejection rate, which leaves the compounding untouched.
• A validation added at the end to catch defects, which raises measured quality and lengthens lead time, because defects are now found at the most expensive point.
• Key-person dependency. Rework concentrates on the most experienced handler, so the metric also predicts a staffing risk nobody has named.

Trade-offs

Driving the rate up is not free. Early validation means collecting more information at intake, which raises abandonment: a form that rejects fewer cases later rejects more applicants at the start. For a sales funnel that trade can be a loss even when the process metric improves.

Pushing the rate towards 100% also hits the exception tail, where the cost per case automated rises steeply. The economically correct end state is usually a smaller manual path rather than none.

When not to use it

The metric is uninformative for processes with one stage and no handoff, and misleading for genuinely exploratory work — a design review or an investigation is supposed to iterate, and counting iterations as rework rewards shallow passes.

It is also the wrong focus where volume is low. At 40 cases a month, a 26% rework rate is ten cases, and the fix is a checklist rather than instrumentation. Spend the effort on the process that runs thousands of times, where the compounding is worth money.

Interview question

Q: A business process has six handoffs, each rejecting about 5% of cases. The operations lead asks for 10% more headcount to meet growth. What do you tell them, and what would you change first?

What a strong answer covers: computing the compounded rate at about 74% and the 1.35 capacity multiplier; showing that removing a handoff beats improving one; moving validation to the earliest point the information exists; separating the rework queue so it is visible and does not age; naming the abandonment trade-off of earlier validation; and pointing out that the headcount request is sizing the wrong number rather than being too small.

Quick check

Quiz: Six handoffs at 5% rejection each. What share of cases completes first time and what capacity multiplier does that imply? About 74% and roughly 1.35 times nominal volume, before any volume growth.

Flashcard: Why does removing a handoff beat improving one? — Because rejection compounds multiplicatively. Six stages at 5% gives 74% first-time-right; four stages at the same 5% gives 81%, cutting the capacity multiplier from 1.35 to 1.23.