Manufacturing analytics
OEE and schedule performance for one production week, measured against the plan that was live at the time.
The manufacturing analytics dashboard reports two things: how effectively the plant ran (OEE), and how closely it ran to the plan (schedule performance). Every figure on the page is measured from what the floor scanned and logged — none of it is typed in.
The period control
One production week at a time, compared with the week before. Weeks start on Monday, matching how schedule weeks are bucketed on the server, and arrows step backwards and forwards through them.
The current week is measured up to now, and its comparison week to the same point in the week. Wednesday-so-far against a whole previous week would read as a collapse every Monday.
Alongside the title, a chip shows which published schedule version is currently active, or warns that none is. With no published schedule, OEE still measures — schedule performance simply has nothing to compare against.
OEE
OEE is Availability × Performance × Quality. Four cards give the plant-wide figure and its three terms, each rolled up across departments from the underlying seconds rather than by averaging percentages — otherwise a department that ran for an hour would weigh as heavily as one that ran all week.
| Card | How it is calculated |
|---|---|
| OEE | Availability × Performance × Quality |
| Availability | Run time / Scheduled time. Scheduled time is machine-hours derived from the shift pattern in the production schedule settings — the same assumption the solver sizes capacity with — less any time logged as not scheduled; run time is that figure less the availability losses in the downtime log |
| Performance | Ideal time / Actual time, measured from the gaps between consecutive batch-ticket scans on each machine |
| Quality | Good units / Total units, where total is good plus waste plus seconds |
Hovering any figure shows the arithmetic behind it, built from the same inputs the number came from.
Measured versus estimated
Anything that could not be measured says so rather than being passed off as a measurement:
- A department with no downtime logged in the period has its availability inferred from runtime instead. It carries an
estimatedchip, and the Availability card reads Availability (estimated) while any department is in that state. This matters because an unmeasured department computes availability as 100%, so OEE would drop — apparently — the day someone starts logging downtime. - When no batch-ticket scan gap qualifies in the period, performance falls back to standard time earned over run time, and the card reads Performance (estimated).
- A performance figure above 100% is reported raw rather than clamped, with a warning: it means a stale ideal cycle time or a missed scan that collapsed two tickets into one interval, not a machine that beat physics.
By department
A table breaks the same figures out per department: Total units, Good units, Run time (h), Availability, Performance, Quality and OEE. A period with nothing scanned says so rather than showing zeros.
Downtime by reason
A Pareto chart of lost minutes by downtime reason, largest first, with a cumulative percentage line — the point being which one or two reasons are worth attacking. It also reports how many downtime events were logged and how much of the lost time was changeover.
With no downtime logged in the period, this card explains that availability above is estimated rather than measured.
Schedule performance
Measured against the schedule version that was published before each week began, so republishing cannot rewrite a week the floor has already worked.
| Metric | What it shows |
|---|---|
| Schedule attainment | Planned units actually produced, capped per campaign at what was asked — over-building one item cannot mask a miss on another |
| Output vs target | Total output as a share of the plan, uncapped. Above 100% means the plant built more than was asked for, which is not the same as hitting the plan |
| Frozen-week adherence | The share of campaigns frozen at publish that survived the week untouched, with counts of how many were changed and added |
| Unplanned output | Units produced that no published campaign called for |
| Waste | Units scanned as scrap in the period |
Every ratio can be empty, and an empty one renders as — rather than 0%. A week with no published plan has no attainment; showing it as a total miss would make the first weeks after publishing look like a catastrophe.
If nothing was published covering the period, the section says so directly: the numbers start accumulating from the first published week onward.
Planned versus actual
A grouped bar chart of planned against actual, with one control — Week, Machine, Department or Item — that drives the whole section and lives in the URL, so a drilldown is a link you can send rather than a thing you describe over the phone.
Frozen-week adherence by version
When commitments have been made in the period, a table shows what happened to each one: the schedule version, the date it was frozen through, how many campaigns were frozen, how many were changed and added afterwards, and adherence expressed both by campaign and by units.
Related
- Machine status and downtime: the downtime log behind availability
- Publishing and releasing: what freezing a week commits to
- Production schedules: the plans performance is measured against
Next: Material analytics