How it works
From export to answer. Nothing in between is a guess.
MIA runs one loop: read your work order export, ask about anything it can’t classify, and answer your questions with the backing work orders attached. Here is each step — including the one where your data is messy, because it usually is.
01 · The loop
Four steps, and you can see inside every one.
Upload the export
The file your CMMS already produces — CSV or Excel, whatever the “export” button gives you. It is read in the browser: no integration, no IT ticket, no software on your laptop.
You don’t clean it first. Messy is the expected starting point.
Received
workorders.csv
1,204 rows · 34 assets · 13 months of history
Columns detected
WO_NUM · STATUS · WO_TYPE · ASSET_ID · COMPLETED_DATE · DOWNTIME_HRS · LABOR_HRS · OT_HRS · DESCRIPTION
Read in the browser — nothing to install
MIA reads it — in your plant's words
Every plant names things its own way. MIA works out what each column is and reads the values in your spelling — the statuses somebody invented years ago, the asset IDs only your site uses.
What it can read, it counts. What it can’t, it doesn’t guess at — that goes to the next step.
Reading workorders.csv
- STATUS — work order statusreadable — 5 terms
- ASSET_ID — machine34 assets
- DOWNTIME_HRS — downtime hoursrecorded on 96% of stoppages
- WO_TYPE — work type2 terms MIA can’t read
A column MIA can't read is asked about, not assumed
It flags what it can't classify — and asks
This is the step most tools skip. A word MIA has never seen never becomes a number: it stops, quotes your plant’s own spelling back to you with the row count, and asks what it means.
You answer once. The answer is remembered, and those rows join every count from then on.
MIA asks — before any number is computed
“EMERG.” — a work type on 168 work orders
What does this mean at your plant?
- Emergency corrective work
- Something else — tell MIA what
Answered once · remembered · 168 rows join the counts
Ask. The answer arrives with its evidence.
Plain English in, plain English out — and under every number, the exact work orders it was counted from. When someone in the meeting challenges a figure, you open the list instead of defending a spreadsheet.
That is the whole product: not a dashboard to interpret, an answer you can hand to the person who asked.
You ask
“How much downtime did we have last month?”
MIA answers
142.7 hours across 38 stoppages. Three machines account for just over half — PACKLINE-A07 leads with 41.2 hours.
Evidence — 38 work orders
- WO-24118PACKLINE-A07 — gearbox replaced6.5 HRS
- WO-24107CONV-03 — belt tracking fault2.8 HRS
- …and 36 more
02 · Three kinds of answer
Every answer says what kind of answer it is.
A number counted from your work orders, a piece of general reliability experience, and a question your data can’t answer are three different things. Most tools serve all three in one voice. MIA labels them — so nothing general ever passes for a finding from your plant.
Computed from your data
“PACKLINE-A07 has had 21 gearbox work orders in 12 months.”
Always carries its evidence — 21 work orders attached
General knowledge — labeled
“Repeat gearbox failures on a wrapper often trace to alignment or loading rather than the gearbox itself.”
Marked as general reliability experience — never presented as a finding from your data
The data can't say
“Whether those failures cluster on one shift isn’t answerable — this export carries no shift column.”
Said plainly, with what would unlock it — not papered over
03 · When the data is messy
The verdict comes first, not after you’ve presented.
Before any analysis, MIA states what your export will support. Every question lands in one of three states — and only the first produces a number.
Supported
The columns are there and the words are readable. Counted now, evidence attached.
Downtime hours — 1,204 rows · hours recorded on 96% of stoppages
Needs your answer
The column is there but some of your plant's words are new to MIA. It quotes them back with row counts; one answer per term, and those rows join the counts. Until then, they are excluded and it says so — never silently miscounted.
Work type — “EMERG.” · 168 rows / “CALL OUT” · 46 rows — tell MIA what they mean
Not in this export
The data simply isn't there. MIA names it plainly and tells you what adding it would unlock. No number is produced in the meantime — a missing column never becomes a zero.
Shift attribution — no shift column · add one and per-shift questions unlock
A number MIA can’t back never reaches the page.
Book a demo
See the loop run on your own export.
Thirty minutes, in the browser. Bring the file your CMMS exports today — messy is fine, that’s the point — and watch what MIA answers and what it flags.
Nothing to install · runs in the browser