Baseline
Last updated
A baseline is the measured record of how a process ran before anything was changed, used as the reference point for whether a deployment improved it. A baseline that stops updating stops being a reference point.
What Baseline actually means
Every claim about improvement is a comparison, and the baseline is the other half of it. Without one, a post-deployment number is a number rather than a result: forty minutes of handling time is neither good nor bad until it is set against what the handling time was before, measured the same way, across the same population.
The distinction that matters is between a measured baseline and an estimated one. Most programs have an estimate, produced in a workshop or from a subject-matter expert recollection, and it is usually wrong in the flattering direction. A reduction measured against an estimate cannot be separated from the estimate being inaccurate, which is why those numbers do not survive scrutiny from finance.
The second failure is staleness. A baseline captured once at pilot stage describes a business that no longer exists within a quarter or two, because teams reorganize, systems get replaced, and volumes shift. Comparing a current number against a frozen baseline attributes ordinary business change to the deployment, in whichever direction happens to be convenient.
This is why a baseline is better understood as infrastructure than as a document. If it is captured continuously from observed execution, it reflects the current business, it covers the variants rather than an idealized path, and the same evidence base serves the next deployment as well as the last one. A snapshot depreciates from the day it is taken.
Examples
Measured against an estimate
A program reports handling time down from ninety minutes to fifty. The ninety came from a workshop estimate. Observed pre-deployment handling time was fifty-five, so the reported improvement is mostly the estimate being wrong.
A baseline that moved on its own
A close process is compared against a baseline captured eighteen months earlier. Two entities have been divested and a system has been replaced since. The comparison credits the deployment for structural change it had nothing to do with.
A baseline covering the variants
Rather than one average, the baseline records the distribution: the median case, the exception path, and the volume of each. Post-deployment, the exception rate can be checked separately from the median, which is where regressions usually hide.
Frequently asked questions
Related terms
Incrementality
Incrementality is the measure of whether outcomes actually improved against a true baseline, as distinct from whether a tool was used. Applied to AI, it asks whether the work got faster, cheaper, or better because of the deployment.
Work Intelligence
Work intelligence is the operational layer that provides a current, evidence-based picture of how work actually moves through an enterprise. It is used to target AI deployment, verify post-deployment outcomes, and continuously update the baseline as the business changes.
Adoption Metrics
Adoption metrics measure whether an AI tool is being used: seat counts, logins, prompts per user, satisfaction scores. They are frequently reported as ROI and are not ROI, because usage says nothing about whether the work improved.
