Home → How we measure

Our method

Every number on this site tells you where it came from.

Promata labels each published figure with how it was produced: measured from system logs, baselined by timing the work with the client's team, estimated from volume with the arithmetic shown, client-reported, or drawn from a third-party source. We do this because we sell the ability to count what nobody in a business has counted. If we are not rigorous about our own numbers, there is no reason to believe we would be rigorous about yours.

A figure from our logistics case study97%of orders processed without manual entryEstimated← every figure carries thisShow the arithmeticNo number without a method
Inbound orders per working day384
Arriving machine-readable341
Routed to a human by confidence gate29
Processed with no human touch312

Why it says 97% and not 81%. The published figure counts the machine-readable channel only. Across all intake, including fax and handwritten, it is 81%. Both are true. We show both.

Why this page exists

The same number, published two ways.

One of these is a marketing claim. The other is something a CFO can take into a meeting. They contain identical data.

What almost every agency publishes

97%

Reduction in manual entry

No denominator. No method. No way to tell whether it was measured, modeled, or chosen. A careful buyer discounts it on sight, and they are right to.

What we publish

97%

Of machine-readable orders, processed without manual entry

· Estimated from channel volumes, Q3· 81% across all intake channels· Arithmetic shown, assumptions listed

Smaller claim. Far more useful. This is the number you can plan with.

The five tags

Ordered by how hard the number is.

Not all evidence is equal, and pretending otherwise is how this market lost its credibility. These run from strongest to weakest, and we tell you which one you are looking at.

MeasuredStrongest · rarest

Logged by a system we built

The number comes out of production logs. We can produce the query and the raw output on request. It only exists after go-live, and only where we instrumented the thing being counted.

ExampleMedian order-to-TMS: 8 seconds
Median of arrival-to-write timestamps
12 weeks post-launch, n = 41,600
BaselinedTimed, not modeled

Timed with the client's team

Someone sat with the people doing the work and timed it, before go-live and again after. We hold the timings and the method. This is how the Blueprint produces a defensible number in ten days without waiting for a system to exist.

ExampleManual order entry: 14 minutes
Timed across 12 orders, 3 coordinators
Discovery shadowing, week 1
EstimatedMost of this site

Modeled from volume, method published

Observed time per unit multiplied by verified volume, with the assumptions listed. Most published figures here carry this tag. That is not a weakness. It is what honest modelling looks like when the raw data sits under NDA.

ExampleStraight-through rate: 97%
Clean orders ÷ total inbound
Modeled from channel volumes, Q3
Client-reportedTheir data, not ours

A figure the client produced

Their number, their method, quoted with permission. We do not restate it as ours and we do not model on top of it. Headcount changes, turnaround times and capacity multiples usually land here, because the client owns that data.

ExampleIntake roles: 6 → 2
Client's own headcount figure
Confirmed in writing
Third-partyNever ours to claim

Published data from a named source

Industry statistics, benchmarks and research we did not produce. Always linked to the original. We will never present someone else’s data as our result, which is a more common practice in this market than it should be.

ExampleMissed-call rate in veterinary practice
Source named and linked inline
Never restated as a Promata finding
← Hardest evidenceSoftest evidence →
The honest part

Most figures in our case studies
are estimates. Here is why.

Automation work happens inside client systems under NDA. We rarely get to export a year of production logs and publish them, and where we could, a client's competitors would be able to reverse-engineer their volumes from the result.

So for published cases we model: observed time per unit multiplied by verified volume, with the arithmetic shown and the assumptions listed. That is an estimate. We label it as one. On a call we will show you the model, the assumptions, and where it could be wrong.

A worked example

What an estimate actually looks like.

Logistics intake · straight-through rateEstimated
Straight-through rate, derived from daily inbound order volume
Inbound orders per working day (verified from TMS)384
Scanned and handwritten, always human-reviewed43
=Arriving in a machine-readable channel341
Routed to a human by the confidence gate29
=Processed with no human touch312
Straight-through rate · 312 ÷ 38481%
97%Published figure
Machine-readable channel only
81%The number to plan with
All intake channels

Where this could be wrong. The 384 is a working-day average across one quarter, and peak weeks run roughly 20% higher, which pushes the human-reviewed share up rather than down. Both numbers above are true. Only one of them is the number a buyer should plan with, and it is the smaller one. So we show both.

The figures we quote about your operation will not be estimates.

They will be timed, with your team, in the first ten days. Someone will sit next to the person doing the work and start a stopwatch. That is not a nice extra. It is the entire point of the Blueprint, and it is why a $1,099 diagnosis exists before a $4,500–$9,500 build.

You should not take our word for what your manual work costs. You should take the timings.

Get Your Execution Blueprint →
Definitions & limits

The two terms we use most, and the claims we refuse.

Hours returned

The measurement of manual work eliminated by an automation: baselined by timing the process with the people doing it before go-live, then measured monthly against system logs after.

  • Reported monthly, in writing, for every build
  • Counts operational task time only, never meetings or strategy
  • If it drops, we investigate at our cost

Execution Blueprint™

A fixed-price operational diagnosis that maps every workflow in a company, scores each manual task by hours lost and cost of error, and returns a sequenced, costed automation roadmap in ten business days.

  • $1,099 fixed, 4 to 6 hours of your time in total
  • Yours to keep and yours to execute, with us or without us
  • Has told clients not to build. In writing.

What we will never claim.

  • Revenue growth attributed to an automation
  • Compounding multi-year projections
  • “Up to” figures, in any context
  • A percentage without a stated denominator
  • Someone else’s benchmark presented as our result
  • A number we could not defend on a call tomorrow