Knowledge Base

Why Dashboards Don't Fix Factories, and What Does

DBR77 Team7 min read

On this page

Dashboards don't fix factories because a screen can show a problem but cannot give it an owner, a rule for when to act or a check that the fix worked, and it often shows the problem too late and without a cause. Plants improve when every signal that matters becomes a task: one person, one time limit, one verification step. The dashboard is the first step of that loop.

Many plants have more screens than five years ago and much the same losses.

What dashboards do well, and where they stop

Dashboards do three things well: they show the state of many lines at once, they show trends that are hard to see shift by shift, and they give the office and the floor the same numbers, as long as the definitions match.

A dashboard cannot act: it cannot decide which red tile matters most, send a technician to line 3 or check an hour later whether scrap went back down. Those steps happen in people's heads, phones and notebooks, or they do not happen at all.

Siemens estimates that unplanned downtime costs the world's 500 largest companies 11% of their revenue, about $1.4 trillion a year. Every minute between seeing a stop and acting on it is part of that bill.

At 10:40, scrap on machine 7 rises to 4%. The dashboard tile turns red.

Six-step response loop for scrap above 3% on machine 7: the dashboard shows the signal, the plant adds rule, owner, time limit, check and learning

Five gaps between the screen and the floor

  1. No owner. A red OEE tile is seen by the plant manager, the shift lead, maintenance and quality. Each one assumes someone else is on it.
  2. No rule for when to act. If nobody has written down who acts and how fast, a red tile turns into wallpaper.
  3. Data that arrives too late. Many plant dashboards still refresh from yesterday's spreadsheet. A loss you see the next morning can only be discussed.
  4. No cause. The screen shows that performance fell, not why. In a Chalmers University study of automatic OEE measurement, almost half of recorded losses could not be classified because loss categories were missing or unclear.
  5. No closure. When nobody checks the fix, the same problem comes back next week and looks new on the screen.

What andon got right

Toyota's andon is still a good model. The Lean Enterprise Institute describes andon as a visual tool that signals whenever an abnormality occurs, where the lit number "summons a quick response from the team leader".

In andon, the signal was designed together with the response. Every andon call has a known responder and a known rule: if the problem is not solved before the product reaches a fixed position, the line stops. A dashboard is designed for viewers, an andon for a responder.

QuestionTypical dashboardResponse loop
Who sees the signal?Everyone with a loginEveryone, plus one named owner
When is it seen?When someone happens to lookSent to the owner when a rule fires
What triggers action?Someone's judgmentA written threshold, e.g. scrap over 3% for 20 minutes
How fast?Not definedA response time, e.g. 15 minutes
How is the fix checked?It is not, or the next dayA measurement, photo or checklist
What if nobody acts?NothingEscalation to the shift lead, then the plant manager
What is learned?A trend lineCause, action and result stored with the task

An illustrative example from a molding shop

Illustrative example: a plant runs 12 injection molding machines on three shifts. At 10:40, scrap on machine 7 rises from 1.5% to 4%.

Dashboard only. The tile turns red. The shift lead notices it at 11:15 and asks the setter to have a look. The setter raises the holding pressure. Nobody writes down what changed. At the next morning's meeting, the weekly scrap chart shows a spike, and the team spends ten minutes reconstructing what happened. Two weeks later the same spike returns on the night shift.

Response loop. A rule says that scrap above 3% for 20 minutes on any machine opens a task for the shift's process technician, with 15 minutes to respond. The task carries the machine, the mold, the material lot and the last parameter change. The technician adjusts the holding pressure, notes it in the task, and the system checks scrap on the next 200 shots. If it is still above 3%, the task goes to the shift lead. The cause and the fix are stored, so the night shift finds them next time.

How to close the gaps without throwing the dashboards away

  1. Pick the five signals that cost the most. For example long stops, scrap, late material, missed checks and near misses. Not 40 tiles.
  2. Write one rule per signal: threshold, owner by role and shift, response time, verification step, escalation path.
  3. Send the signal to the owner, not only to the screen. A notification on a phone or terminal beats a tile nobody watches.
  4. Review open and late tasks in the daily meeting instead of yesterday's charts. The meeting gets shorter when the list is already there.
  5. Measure the loop itself: time to respond, time to close, and how often the same problem returns.
  6. Remove tiles nobody acts on. If a number has had no owner for a month, give it one or take it off the screen.

All of this is easier when each KPI has one definition across departments. How to Build a Real-Time KPI System for Your Factory shows how to agree one.

How this works in DBR77 IRIS

DBR77 IRIS keeps dashboards and tasks in the same system. IRIS KPI shows the numbers. When a rule fires, IRIS Tasks creates a task with its context attached: line, machine, work order, material lot. Routing rules send the task to a role, and it moves through open, in progress, review, done and verified. A task can require verification by a named role before it closes, and overdue tasks are flagged.

IRIS Gemba brings open issues onto tiered boards, and alerts and tasks are routed by role, so operators, shift leads and managers each see what they own. Data & AI proposes recommendations from cross-module data, and a person decides what happens to each one. From Signal to Task to Closure: The Execution Loop follows the full path, and What Is a Plant Operating System? explains the wider idea.

FAQ

Should we stop building dashboards?

No. Keep them for overview and trends. Make sure every number that needs action has a rule and an owner behind it.

How many signals should become tasks?

Start with five or fewer per line. Add more only when the first ones are closed on time most of the time.

Won't this flood people with alerts?

Only if every tile becomes an alert. Send a signal only when it has an owner and a defined action. If owners often close tasks with "no action needed", the threshold is too low.

Do we need AI for this?

No. Rules, owners and response times work without AI. AI helps later, when there are more signals than people can rank by hand. See What AI Can Really Do in Factory Operations Today.

Conclusion

Start with the five signals that cost the most and write one rule for each: threshold, owner, response time, verification step and escalation path. After a month, look at how fast those tasks close and how often the same problem returns. Any tile that still has no owner by then gets one or comes off the screen.

Sources