Know which automations pay for themselves.
Some of what you run earns its keep. Some of it quietly costs more than it saves, and the thing you have not built yet may beat both. LumaTrack prices every run from every tool at labor rates you can cite, counts the failures against the total, and tells you what to keep, what to retire, and what to build next. When someone asks you to show your work, the report is already written.
Explore the live demo, no signup →
Free tier: 5 automations, 25,000 run events/month, unlimited viewers. Forever.
| 1,608 h returned × $85.02 blended loaded rate | $136,708 |
| build, license, and per-run costs | −$33,081 |
| Net savings to date | $103,627 |
of SOCs still assemble their metrics by hand, and nearly half call it too time-consuming.
of CFOs are confident they can deliver enterprise AI impact. The rest are asking you for evidence.
the ROI your vendors claim for themselves, across commissioned studies with incompatible methodologies. Finance discounts all of them.
Forrester TEI (Microsoft) · Forrester TEI (UiPath) · IDC (Red Hat) · do your own math instead
For years I automated network changes: Ansible, Python, REST calls against switches. The playbooks worked. Every few months someone would ask what they were worth, and the best I had was a count of successful runs and a graph. That is not a number finance accepts. The failed pushes, the ones that caused an outage and cost more than doing the change by hand would have, never appeared anywhere at all. I built LumaTrack because I wanted an answer to that question.
See what a single automation returns in a year.
Describe the task this automation takes off someone's plate and it works out the labor value it gives back over a year, net of what the automation costs to run.
This is a rough estimate: labor value minus running cost, before the build cost and the adoption ramp.
Three steps to a number that survives the meeting.
Set the baseline
What does the task cost when a human does it? Minutes times a fully loaded rate, seeded from Bureau of Labor Statistics medians and cited to the dollar. Override anything with your own numbers.
Stream the evidence
Anything that can speak HTTP reports each run with one call. Failures count too. A failed run costs money and saves nothing, and the ledger books it that way.
curl -X POST https://lumatrack.io/api/v1/runs \
-H "Authorization: Bearer lmt_..." \
-d '{"automation": "os-patching", "status": "success"}'
The ledger does the reporting
Net savings and hours returned, per automation and across the whole portfolio, with the payback date. It lands in your inbox or Slack on the schedule you set, ready to forward to whoever signs the budget, so the quarterly spreadsheet stops being someone's job.
Keep it, retire it, or build the next one.
One view of everything you run, ranked by what it actually returned, with the work you have not built yet ranked by what it would. Realized value comes from your own runs, and projections are labeled as projections.
Decide what to automate next
A candidate pipeline ranks the work you have not automated yet by projected annual value and payback period, so the backlog is ranked on what each item returns and how fast.
Retire what stopped paying off
When an automation's upkeep outgrows the value it returns, LumaTrack flags it for retirement with the net-to-date in plain dollars, before it quietly costs you for another year.
Test the decision before you commit
Reprice the whole portfolio against a different rate card, or against half the volume, and see which automations survive it. The scenario runs on the same ledger, so the comparison is like for like.
Built like a ledger.
Vendor dashboards grade their own homework: one made-up rate, failures ignored, history that rewrites itself when settings change. LumaTrack is built the way finance already works.
History that can't rewrite itself
Rates and assumptions carry their effective dates, so March's savings are priced at March's rates, forever. A closed month doesn't change: corrections arrive as their own visible entries, and a nightly job re-checks every closed month against the day it closed and re-derives the open figures from the raw runs, to the cent.
Savings classes stay separate
Hard savings, cost avoidance, and productivity are reported side by side, so hours returned never get counted as budget cuts. That's the first distinction a finance reviewer checks.
Your data leaves when you do
Every ledger row and run event downloads as CSV from inside the app, on any plan. The same records read over the API. There is no exit fee and no ticket to open. Plain CSV and JSON, still readable the day after you cancel. What you can export.
Every decision traces back to the runs behind it.
Retiring an automation is an argument with someone, so the number had better survive it. Every figure on the dashboard and in the ledger expands to the runs it came from, the rate in force when it happened, and the assumption version it was priced on. If your analyst cannot re-derive it on a calculator, we do not show it. Change an input and the correction posts as its own entry; the months you already reported stay exactly as you reported them.
Losses count too. Track the incidents an automation is meant to prevent, and the same ledger turns them into a board business case, then into the automation that closes them.
Reports that send themselves
A scheduled Exec Value Pack to email and Slack, and share links that stay live after you send them, so the figure the reader opens next week is the current one. Nobody assembles a deck the night before the review.
Which report are you on the hook for?
MSPs: the QBR page
"We returned $7,206 to your account this quarter, here is the math." The page that sentence comes from, generated for you.
LumaTrack for MSPs →AI teams: the agent P&L
Langfuse tracks what an agent costs to run. What the work was worth is the other column: runs × human baseline − tokens, on one shareable page for the perennial "is it worth it?" And when someone asks what happens if your provider reprices, the scenarios page already has the answer.
LumaTrack for AI teams →SOC: the CISO-pack section
Every Tines story and SOAR playbook reports its runs, and the automation-value section of your monthly pack assembles itself. Enrichment API costs are netted out.
Get the CISO-pack template →Network and infrastructure: the change you cannot price
Ansible playbooks, Python scripts, REST calls against switches and firewalls. Value here gets argued in avoided outages and hours of CLI work, and a failed change costs more than the manual one would have. LumaTrack prices that side: a failed run books its cost with no value against it, and downtime reduction is a value-stream type of its own.
See a network estate priced →One ledger, from ingest to audit.
Runs come in from anything that speaks HTTP, get priced against cited rates, and become reports and exports your finance team can audit. Closed periods reconcile themselves nightly.
The value report your QBR deck is missing.
One ingest pipe feeds every client, each one isolated, with closed quarters that don't change and its own evidence grades. White-label, so it goes out under your brand. Priced by the size of your client book.
LumaTrack for MSPs →Some of your automations are paying for themselves.
Find out which.
Free forever tier · cancel anytime · Team $59/mo flat · Business $499/mo · full pricing.