8 Production Dashboard Mistakes That Send Planners the Wrong Way

Most production dashboard mistakes are not bad data. They are correct numbers read against the wrong assumption, then acted on quickly because a dashboard invites quick action. The EDGEBIC by User Solutions cockpit is unusually explicit about provenance (every tile carries a chip saying whether its number is a real event, a projection, or a source that does not exist yet), and even so, these eight readings come up again and again. Here is what each one actually means.

For what the tabs are and who they serve, see the dashboard overview; for configuration, the setup guide.

Mistake 1: Trusting a Tile You Have Not Refreshed

Symptom. A number in a meeting that nobody on the floor recognizes.

Cause. The cockpit does not auto-refresh by default. Each tab loads its snapshot when you switch to it, and Refresh now reloads on demand. That is deliberate: continuously re-querying a large schedule in the background costs more than it returns.

Fix. Click Refresh now as the meeting starts, every time. If you want the capacity tab to refresh itself, set an auto-refresh interval in seconds in its settings dialog, where zero disables it. Make it a ritual rather than a rule and the argument disappears.

Mistake 2: Reading a Red Andon Board as a Broken Plant

Symptom. Most work center tiles show red by ten in the morning, every morning.

Cause. Andon state is inferred from schedule actuals rather than machine sensors. A station goes red when it is more than ten minutes past its planned start with no actual start recorded, and yellow when it is running more than 25% over its planned duration. If nobody punches in, everything eventually reads red.

Fix. Enforce a punch at operation start on the kiosk. The board then becomes a real signal within a week, and the same punches light up schedule adherence, the operator counts on the shift production report, and the quality factor in OEE. One habit makes four measurements real.

Mistake 3: Confusing Late With Overdue

Symptom. A job sits in the Late slice with weeks left before its due date, and somebody calls it a bug.

Cause. The two slices measure different failures. Late means the plan itself misses the due date, or the job finished after it, which is structural and visible in advance. Overdue means the plan met the date but the date has passed with the job unfinished, which is execution.

Fix. Treat them as separate queues, because the fixes differ. Late needs capacity, an alternate routing, or a renegotiated date. Overdue needs someone to find out why the work stalled. And remember that starting a job is a badge on the active jobs list rather than a slice, because starting says nothing about whether it will be on time.

Mistake 4: Quoting a Plan Number as a Fact

Symptom. A promise made to a customer from a tile, which then moves.

Cause. Chips exist for exactly this. Live means a persisted real event. Plan means the scheduler's projection, which improves as actuals arrive. Estimate means a heuristic. On-time in full carries a Plan chip because it reads projected end dates until actuals fully replace them; backlog carries Live because open order counts are persisted facts regardless of any schedule.

Fix. Read the chip before you read the number. A Plan figure is fine for internal steering and wrong for a commitment. When you need a commitment, run the quote simulation against finite capacity rather than quoting a dashboard tile.

Mistake 5: Treating an Offline Dash as an Error

Symptom. Safety, quality, or margin tiles show a dash and someone files a defect.

Cause. Those tiles have correct formulas and no source data yet: safety needs an incident log, quality needs a defect log, margin needs revenue. The dash is the honest display.

Fix. Read the dash as "not captured", not "zero". This is the same design choice that makes OEE read "n/a" without kiosk punches instead of inventing a 100% quality factor, and it is what makes the tiles that do carry numbers worth trusting. If a tile matters to you, the fix is capturing its source, not filling the space.

Mistake 6: Averaging Two Views That Measure Different Windows

Symptom. The heatmap shows a red day and the utilization report grades the same station GOOD.

Cause. The heatmap is per day. The utilization report averages a range. One oversold Tuesday inside a comfortable month is genuinely both.

Fix. Use the per-day capacity dialog for day-level truth and the utilization report for the trend, and never resolve the difference by splitting it. The response to the red day itself is a four-step playbook covered in reading a red day. If the station is red every week rather than once, work center overload causes and fixes is the deeper triage.

Mistake 7: Not Clicking Through the Number

Symptom. A disputed load figure that nobody can decompose, or a job detail popup whose rows do not add up.

Cause. Two related habits. First, the numbers on the resource calendar are clickable: any non-zero hours-assigned or backlog cell opens a job detail popup listing every allocation behind it, and most arguments about a cell end the moment somebody clicks it. Second, that popup honors the same shift and job filters as the grid, so with a shift filter active the rows legitimately sum to less than the unfiltered cell.

Fix. Click first, argue second, and check the filter bar before doubting the arithmetic. A cell that will not open is either zero or an identity column rather than a value.

Mistake 8: Reading a Missing Configuration as a Missing Number

Symptom. The bottleneck tile shows a dash, or a department's throughput on the exec tab looks impossibly low.

Cause. Both are master data. The bottleneck tile and every bottleneck panel read the work center bottleneck flag; with no station flagged, there is nothing to show. The multi-plant roll-up pegs each job to the department whose work centers consumed the most of its scheduled hours, and skips work centers with no department, so a job that ran entirely on unassigned stations contributes to nobody.

Fix. Flag your genuine constraint, and only your genuine constraint: marking a station that is not the real bottleneck causes the engine to protect it deliberately at other stations' expense. Then assign every work center to a department and re-check the roll-up. Both are five-minute fixes that make three tabs honest at once.

The Habit Underneath All Eight

Every tile carries an info icon whose tooltip gives its formula and its data source in one sentence, and every chip declares its provenance. Between them, a planner can answer "where does this number come from and how much should I trust it?" without leaving the screen. Most of the mistakes above are one hover away from being impossible.

Then, when a tile says something is wrong, go down a level rather than sideways. Dashboards are the glanceable layer; the reports carry the authoritative formulas and the built-in column glossary, and Scheduler Anomalies runs 22 automated checks against the current plan before anyone escalates a scheduling fault. The equivalent traps on the report side are collected in report reading mistakes, and symptom-first triage lives in the troubleshooting guide.

For the metrics behind the tiles, our manufacturing KPI guide and schedule adherence KPI cover the generic definitions, and the complete EDGEBIC guide covers the system producing them. Want to see how your plant reads on this cockpit? Bring your data to a demo.