# The cost of poor planning

What a product organisation spends today on specs that arrive ambiguous, and what it costs to stop. Four cost centres against the Plansmith seats, with your own team and salaries, in dollars or euros.

The interactive calculator is at https://www.plansmith.co/roi. This is an illustrative model, not a customer result: every figure below is an input you can replace with your own.

**Planner output: 4.3x.** Each planner writes build-ready specs and prototypes 4.3x faster: 13 planner hours a spec and prototype today, 3 with Plansmith. The conservative case gives 2.6x.

Measure Plansmith against your payroll, not your tooling budget. A team of 100 engineers and 40 planners takes 60 planner seats (every planner and 20% of engineers) plus 80 comment-only seats for the other engineers, at $124,000 a year. That is 0.49% of a $25,500,000 product payroll.

## Where the money goes today

Ambiguous specs never show up as a line item. They cost capacity in four places.

| Cost centre | What happens today | Why it costs what it costs |
|---|---|---|
| Writing specs & prototyping | A planner hand writes the spec, structures it, and breaks it into tickets. | Skilled product time goes on formatting and decomposition instead of judgement. |
| Questions back and forth | An engineer hits an ambiguity, asks, waits, switches context, resumes. | You pay twice: the engineer's blocked time and the planner's interrupted time. |
| Rework | The feature ships, then is reworked because it met the ticket but not the intent. | The most expensive defect. It consumes full delivery capacity before anyone notices. |
| Wasted agent runs | A coding agent takes a vague ticket down the wrong path and the output is thrown away. | The only metered cost. Vague input turns straight into billed tokens you discard. |

## Worked example

Profile: 100 engineers and 40 planners, writing 1920 specs a year at 4 per planner per month. Fully loaded salaries of $185,000 a year per engineer and $175,000 per planner. Every engineer on a coding agent at $200 a month. Team plan, billed yearly. All figures in US dollars; the calculator also works in euros and per month.

| | Conservative | Base case |
|---|---|---|
| Annual recovery | $1,984,846 | $4,554,462 |
| Plansmith seats per year | $124,000 | $124,000 |
| Net benefit | $1,860,846 | $4,430,462 |
| Return | 15.0x | 35.7x |
| Payback | 4 weeks | 2 weeks |
| Planner output (build-ready specs and prototypes per planner hour) | 2.6x | 4.3x |

| Cost centre | Basis, base case | Conservative | Base case |
|---|---|---|---|
| Writing specs & prototyping | 8 hrs down to 2 hrs per spec | $646,154 | $969,231 |
| Questions back and forth | 10.0 hrs down to 2.0 hrs per spec | $1,098,692 | $1,329,231 |
| Rework | 30% of engineering capacity, cut by 40% | $222,000 | $2,220,000 |
| Wasted agent runs | 30% of 100 agent seats wasted, 50% of it recovered | $18,000 | $36,000 |
| Total annual recovery | | $1,984,846 | $4,554,462 |

## Why the number is hard to argue with

Most vendor ROI models need a large behavioural change to clear a large licence fee. Here the seats are a fraction of a percent of payroll, so even the conservative case clears them many times over. Every input is yours to change.

## Sensitivity on the two contestable inputs

Two assumptions carry the model: how much engineering capacity rework eats today, and how much of it Plansmith removes. Here is the recovery across their plausible range, with everything else held at your inputs.

| Capacity lost to rework | Cut by 20% | Cut by 30% | Cut by 40% |
|---|---|---|---|
| 10%, a disciplined team | $2,704,000 | $2,889,000 | $3,074,000 |
| 20% | $3,074,000 | $3,444,000 | $3,814,000 |
| 30%, base case | $3,444,000 | $3,999,000 | $4,554,000 |
| 40% | $3,814,000 | $4,554,000 | $5,294,000 |

The weakest corner of this grid, a disciplined team with only a 20% improvement, still returns $2,704,000 against $124,000 of seats. The case does not depend on the optimistic corner being true.

## What this model deliberately excludes

Four real benefits are left out. They are harder to defend in a finance review, and the case does not need them.

- Cycle time. Shipping the same roadmap sooner has revenue value we have not attempted to price.
- Retention. Rework is among the most commonly cited sources of engineering frustration.
- Escaped defects. Spec errors that reach production cost materially more than those caught in delivery.
- Token efficiency. Better structured specs should cut agent token use. That is for a pilot to measure, not a spreadsheet to assert.

## Assumptions, stated plainly

Salaries are fully loaded: base pay plus employer costs, usually about 1.3x base. The model spreads them over 2,080 working hours a year to price the per-spec time. Seats are the Team plan, billed yearly. Spec volume of 1920 a year assumes 4 per planner per month. The per-spec hours are estimates to replace with your own during a pilot. Rework percentages are ranges and not measurements, which is why the stress test is there. No figure here is an industry benchmark presented as fact.