Measuring ROI — credible Claude Cowork numbers in 6 weeks

TL;DR. A defensible ROI calculation a CFO will sign off on. Not "we save 30% of time" — actual hours, actual cost, actual outputs. Pick exactly two metrics, capture a Week-1 baseline, re-measure at Week 6, monetise honestly. Two metrics with proper baselines beat five metrics measured loosely.

The two metrics that matter#

Pick exactly two for the pilot:

• Hours saved per role per week, measured against a Week-1 baseline.
• Output throughput — number of artifacts produced per role per week, or quality-adjusted equivalent.

Two. Not five. Not zero.

The Week-1 baseline#

Before the pilot starts, capture:

• For each operator in scope: hours per week on the targeted tasks.
• For each task: number produced per week, average rework rate.
• Sponsor signs off on the baseline numbers.

Without a baseline, ROI is a story.

The Week-6 measurement#

• Same operators, same tasks, same week of the month if cyclical.
• Count hours and outputs in the same way as the baseline.
• Calculate delta: hours saved × loaded cost per hour = monetised saving.
• Throughput delta: outputs × value per output (revenue, deals, decisions, etc.).

The point of "the same way as the baseline" is non-negotiable. ROI math that compares apples to oranges loses the CFO conversation in 30 seconds.

The cost side#

Honest accounting:

• Cowork seats × monthly cost.
• Tinkso engagement cost (or internal team time, valued honestly).
• Any custom MCP or skill build cost.
• The pilot owner's time.

Total it. Compare to monetised savings. State payback in months.

A worked example (illustrative)#

Realistic numbers, anonymised:

• 5 operators in scope, 4 hours saved each per week.
• Loaded cost per hour: £75.
• Weekly saving: 5 × 4 × £75 = £1,500.
• Annualised: ~£75k.
• Pilot cost: 5 × Pro seat × 12 = £1,200/year + Tinkso 6-week engagement at £18k = £19.2k year one.
• Payback: ~3 months.

This is illustrative — your numbers will vary.

The throughput story#

Hours saved is the floor. Throughput is the ceiling.

• Same hours, more outputs — e.g., proposals shipped per week.
• Same outputs, higher quality — e.g., win rate on proposals.
• Same hours and outputs, faster turnaround — deadline savings.

Throughput often beats hours saved by 2–5×, but it requires more careful measurement. Worth the extra effort if your function has a clear unit of output.

What not to claim#

• "Cowork saved us X% of headcount." Not credible without months of data, and politically explosive even when it is.
• "AI productivity gains" without naming the function.
• Hours saved that weren't measured against a baseline.
• Throughput without a denominator.

A conservative, defensible ROI memo wins more arguments than an aggressive one.

The ROI memo#

A one-page artifact with:

• Two metrics, baseline → current.
• Cost side.
• Payback period.
• Three caveats (where the numbers are softest).
• Recommendation: scale, hold, or stop.

Goes to the sponsor, finance partner, and the steering group.

Tinkso's take#

The two-metric discipline is the most under-used governance trick. Sponsors love a single dashboard; two metrics with proper baselines beat five metrics measured loosely. The CFO cares about the gap between baseline and current, not about the dashboard.

Try this#

Take your pilot. Write the Week-6 ROI memo today, with the numbers you would expect. The exercise reveals which metrics you can defend and which you would hand-wave. Fix the hand-wave ones before Week 1.