Glossary
What LumaTrack's terms mean, defined to match what the engine does. Every term below is linked from the places it appears in the product.
- Run event
- The JSON record one automation execution sends to the ingest API: status (success, failure, skipped, cancelled), timing, and optionally units of work. Every dollar in the ledger traces back to run events.
- Builder seat
- A seat for someone who creates or edits: automations, baselines, costs, settings. Plans meter builder seats. Viewers are unlimited on every plan and see every report, dashboard, and export.
- Candidate
- A task you have not automated yet. LumaTrack projects its annual net savings and payback from your own inputs, so the business case exists before you build.
- Manual baseline
- What the task costs a human: minutes per execution at a loaded hourly rate, plus expected volume. The savings math is the gap between this baseline and what the automation actually does.
- Oversight minutes
- Human time an automated run still consumes: review, approval, spot checks. Deducted from every run's saved minutes, so a run someone babysits saves less than a run nobody touches.
- Savings class
- The lane a dollar reports in. Hard savings are budget money (headcount avoided, licenses cancelled). Cost avoidance is spend that would have happened without the automation. Productivity is time returned to people, priced with a conservatism discount and kept in its own lane so it can't inflate the savings total.
- Hard savings
- The strictest savings class: money that leaves a budget line, such as avoided headcount or cancelled spend. LumaTrack reports it at full value, with no conservatism discount.
- Cost avoidance
- Spend that would have happened without the automation: an incident that never escalated, an audit finding nobody had to remediate by hand. It is real money, and it reports in its own lane apart from hard savings.
- Conservatism
- A multiplier (0.50 by default) on the dollar value of productivity-lane time savings, because saved minutes do not all convert to productive work. Hard savings and cost avoidance are reported at full value.
- Evidence grade
- How the baseline minutes were established: measured (timed runs), sampled (a timed sample), or declared (someone's estimate). The weaker the evidence, the more the projection is discounted before anything reports it.
- Payback
- The month the automation's cumulative net savings pass its build cost. Computed from trailing actuals for live automations and projected from inputs for candidates.
- Closed period
- A month that has frozen. After the close lag, the period's figures stop moving, and a nightly checksum verifies the frozen rows stayed byte-identical. Evidence that arrives late books as a prior-period adjustment, so the month's original figure survives.
- As reported vs restated
- A closed month keeps its as-reported figure forever. Corrections that arrive later book separately, and as-restated equals as-reported plus corrections, so you can always see both what was claimed at the time and what is known now.
- Backfill
- Importing historical runs after the fact. Backfilled runs carry their provenance on the ledger, so a number built on imported history says so.
- Realization factor
- A 0-to-1 haircut on modeled value, applied because finance discounts claims. A realization factor of 0.8 books 80 cents of every modeled dollar.
- Value stream
- Value beyond replaced labor, declared as its own entry: a deflected ticket queue, an avoided incident, recovered revenue. Each stream reports in its own savings class with its own realization factor.
- A cost that serves several automations (a platform license, a shared runner) and is allocated across them by a stated rule, so no automation looks free just because its bill lands elsewhere.
- Initiative
- A grouping of automations pursued as one effort, so a program of several workflows can show one net figure.
- Run cost
- What one execution costs to run: compute, API calls, tokens. Charged per run, including failed runs, which cost money and save nothing.