Count the hours before we touch them.
The Audit process inventory sets Hours_now: the measured time your team spends on the work in scope today. Every downstream number references this baseline, so the math starts from evidence, not a pitch.
How we measure ROI
We don't publish an ROI. We measure yours.
Most firms are sold "hours saved" off an industry average. We report hours that stay saved, and what each one cost to reclaim, measured in your own system off the Audit baseline and the Build's eval suite. Not an average. Your number.
See your number, free
Free · 30–60 min · leave with a plan
1number that matters
your systemwhere we measure it
quarterlyre-checked in Run
$0to find out
One honest number
The number a partner should ask for is payback, not a percentage.
The metric we report first is the payback period in months: how long the engagement takes to return its own cost. A non-technical partner already thinks in payback and in partner hours. A "312% ROI" headline does not survive contact with a real budget meeting; a "this pays for itself in month eleven" does.
Two numbers support it. Cost per reclaimed hour tells you what an hour of capacity costs to buy back through the system, so you can compare it to the hour you are paying for today. And stay-back retention, the share of reclaimed hours still verified at six and twelve months, is the reliability moat: a saved hour that quietly returns as rework was never saved.
Primary
Payback period
Months until cumulative value equals total cost. The number a partner can act on without a translation layer.
Supporting
Cost per reclaimed hour
Total spend divided by the hours the system actually reclaims. Compare it to the hour you buy today.
Supporting
Stay-back retention
Share of reclaimed hours still verified at 6 and 12 months. The moat, and the number we refuse to guess.
The formula, so you can reproduce it
Every input is read from your system, not assumed.
01
Baseline the hours
Hours_now comes straight from the Audit's process inventory: the time your team spends today on the specific work in scope. Not a survey guess, a counted baseline.
Hours_now = counted from Audit process inventory
02
Reclaim only what the eval suite proves
A is the reliable automation rate read from the Build's eval suite: the pass rate on your own work, not an assumed percentage. Gross reclaim is what A can carry.
Gross_reclaim = Hours_now × A A = eval pass rate on the client's own work
03
Subtract the review the system adds back
Every system adds review hours: someone checks, approves, corrects. We subtract those, then annualize on 46 to 48 working weeks, never a full 52, so holidays and slack are not counted as capacity.
Net_reclaim = Gross_reclaim - review_hours_added Annual_reclaim = Net_reclaim/wk × 46–48 wks
04
Price the hour with the right lens
The cost lens values a reclaimed hour at the fully-loaded staff cost you stop spending. The capacity lens values it at the realized billable or advisory rate, and it only applies if the firm has demand to redeploy those hours into.
Cost lens → fully-loaded staff cost / hr Capacity lens → realized billable rate / hr (only if redeployable demand exists)
05
Divide, then read the payback
Cost per reclaimed hour is spend over annual reclaim. Payback is Year-1 cost over the monthly value the reclaim produces under the chosen lens.
Cost_per_hour = spend / Annual_reclaim Payback_months = Year1_cost / (Annual_value / 12)
How we instrument it
Audit · baseline
Eval suite · proof
Decision tracing · add-back
Run · keeps it from decaying
The baseline, the proof, and the guard against decay.
Prove the reclaim on your own work.
The reliable automation rate A is the eval pass rate on your own documents and cases, not a vendor benchmark. If the suite says the system is right 70% of the time, A is 70%, and the reclaim is sized to it.
Count the review the system creates.
Decision tracing records what the system did and where a human had to step in. Those review hours are the add-back we subtract from gross, so Net_reclaim reflects the checking the system genuinely requires.
Re-check when the model changes.
Run re-runs the eval suite on model upgrades and reports cost per hour and payback each quarter. A reclaim that decays after a model swap is caught here, not discovered in a year-end surprise.
The Audit sets the baseline, the eval suite proves the reclaim, and Run keeps it from decaying.
Two lenses, one honest fork
A reclaimed hour is worth what you can do with it.
The same reclaimed hour can pay back on two very different timelines. We show both and tell you which one applies, rather than quoting the richer one as if it were automatic.
Cost lens
The hour you stop paying for.
Value each reclaimed hour at the fully-loaded cost of the person no longer doing it by hand. Conservative, always valid, and the number we lead with when demand to redeploy is uncertain.
Applies: always. Never overstates.
Capacity lens
The hour you resell.
Value each reclaimed hour at the realized billable or advisory rate it can be redeployed into. Larger, but only real if the firm has the demand to fill that freed capacity with paid work.
Applies: only if billable demand exists to absorb the freed hours.
Two worked examples
Illustrative, not a measured result
Illustrative, not a measured result
Illustrative, not case studies. The math is the point.
Kindl is early-stage. The figures below are worked illustrations to show how the formula behaves, not measured results from a client. Every number is labeled as such.
Law firm · first-pass contract and NDA review
4 fee-earners · contract triage- Net reclaimed~15 hrs/wk · ~700 hrs/yr
- Loaded staff cost~$110 / hr
- Billable rate~$325 / hr
- Build (one-time)$55K
- Run$7K / mo
Cost / hr · year 1~$199
Cost / hr · steady~$120
Payback · capacity lens~12–13 mo
On the cost lens ($110/hr loaded), year one does not fully pay back: the reclaimed hours are worth ~$77K against ~$139K of year-one spend. It clears only on the capacity lens: redeploy 60% of the freed hours into billable work at $325 (~$136K/yr) and payback lands around month 12 to 13. The honest fork: this only holds if the firm actually has billable demand to fill those hours. No demand, no capacity value.
Accounting firm · tax-season intake and data entry
busy-season capacity- Net reclaimed~620 hrs/yr
- Junior wage (loaded)~$65 / hr
- Advisory rate~$250 / hr
- Build (one-time)$45K
- Run$6K / mo
Cost / hr · year 1~$189
Cost / hr · steady~$116
Payback · capacity lens~14–16 mo
On the junior wage alone (~$65/hr), this loses: 620 reclaimed hours are worth ~$40K against ~$117K of year-one spend. It pays through busy-season capacity: avoid a seasonal hire (~$25K) and redeploy ~200 senior hours into advisory at ~$250 (~$50K), for ~$75K/yr of realized value, and payback lands around month 14 to 16. The honest point: the wage math does not carry it. The seasonal-capacity math does, and only if that redeployment is real.
What we will not do
The evolution ladder
The discipline is in what we refuse to claim.
/ We won't publish a fabricated average.
There is no "Kindl clients see X%" headline, and there will not be one until real engagements produce real numbers. An average measured on someone else's business is not a forecast for yours; it is a marketing number wearing a lab coat.
/ We won't quote the capacity lens as automatic.
The billable-rate value only exists if you have demand to redeploy freed hours into. We lead with the cost lens, name the capacity lens as conditional, and never blend the richer number into the headline as if it always applied.
/ We won't count gross hours or a full year.
We subtract the review the system adds back before we call an hour reclaimed, and we annualize on 46 to 48 working weeks, not 52. Counting gross hours across a full calendar year is the easiest way to inflate a number, so we do neither.
/ We won't put a number on retention until Run has one.
Stay-back retention is measured, not projected. Until Run has produced a real 6 and 12-month figure, we report the metric as pending rather than inventing a placeholder. A moat you have not measured is not a moat you get to claim.
Now
Illustrative worked examples
The formula, run on labeled illustrative figures. Every number carries its caveat. This is where we are today, and we say so.
Next
One consented case
A single client, with permission, real inputs, real payback. Named as one data point, not a trend.
Then
Measured median, N stated
A median across engagements, always published with the sample size behind it. No N, no claim.
Later
Measured 6/12-month retention
Stay-back retention from Run's own re-checks. The reliability moat, reported only once it is measured.
See your number, on your numbers. For free.
A working session with our team · 30–60 min · your baseline, your math, a plan you keep