One Dashboard, Five Versions of Factory Truth.
A dashboard cannot resolve conflicting definitions. Management visibility starts by agreeing what a job, quantity, milestone, delay and margin actually mean.
The short answer
A printing dashboard is trustworthy only when every metric has an agreed business definition, source event, unit, owner and refresh rule. If sales, planning, production, quality and finance define “complete” differently, combining their data creates a more attractive disagreement—not one factory truth.
The meeting with five answers
Management asks how many jobs finished yesterday.
Sales says twelve because twelve orders are ready to invoice. Production says fifteen because press and finishing operations closed. Quality says nine because six batches await release. Warehouse says eight are ready to dispatch. Finance says eleven jobs have complete cost.
Every answer is defensible. The question was not defined.
Start with decisions
Before designing a dashboard, identify the decisions it should support: protect today’s deliveries, resolve capacity constraints, contain quality risk, improve margin or manage customer service.
Each decision needs a small set of measures and drill-down evidence. A display filled with every available number can obscure the one exception requiring action.
Define audience and time horizon. A shift supervisor, planner and managing director should not receive identical views simply because the data exists.
Write a metric contract
For every KPI, record:
- business question;
- exact definition and exclusions;
- unit and time basis;
- source event and system;
- owner and data steward;
- refresh timing;
- target or threshold;
- drill-down path;
- known limitations.
This contract prevents a label such as “on time” from changing meaning between reports.
Separate milestones
Order approved, artwork ready, material ready, production complete, quality released, dispatch ready, shipped and delivered are different events.
Do not collapse them into one status. The dashboard can present a customer-facing headline while preserving the operational chain underneath.
The traceability pillar provides the identities and evidence that allow each milestone to be reconstructed.
Make quantity states explicit
Planned, started, good, rejected, held, reworked, packed, allocated and shipped quantities answer different questions.
Specify units and conversions. A sheet, blank, carton, bundle and pallet cannot be added without context.
When a KPI uses “completed quantity,” state whether held units are excluded and whether partial batches count. Consistency matters more than a convenient large number.
Define delay against a reference
A job can be late to the original customer promise, current revised promise, internal operation plan or carrier cut-off.
Show which reference the dashboard uses. Preserve original and revised dates so schedule change does not erase performance history.
Reason codes should distinguish customer change, material, artwork, capacity, breakdown, quality and logistics, while allowing accountable investigation.
Connect quality with business effect
Defect count alone does not show affected quantity, containment time, rework, delivery risk or complaint severity.
Combine first-pass acceptance, holds beyond deadline, rework quantity, recurrence and customer impact. Allow drill-down to the specific batch, material lot and decision.
This keeps quality visible without turning every minor correction into the same red signal.
Show margin at the right maturity
Estimated margin, current expected margin and operationally closed actual margin are different values.
Label provisional cost. Include outstanding material, subcontracting, freight and quality accruals. Do not present an invoiced job as financially final when cost is still moving.
Management can then distinguish commercial price performance from incomplete posting.
Build from events, not manual summaries
Dashboards become fragile when departments retype totals into spreadsheets. Use operational events—receipt, release, start, completion, hold, shipment and cost posting—where reliable.
Human judgment remains necessary for causes, concessions and decisions. Capture it once against the relevant job or event.
Automating an ambiguous field only distributes ambiguity faster.
Design exception-first views
Show jobs where action can still change the result:
- material not ready before its operation deadline;
- artwork or tooling unresolved;
- machine event invalidating the schedule;
- held quantity approaching a decision deadline;
- production complete but shipment blocked;
- complaint scope not established;
- waste or cost beyond expected range.
Leaders can drill from portfolio to job without listening to a verbal reconstruction.
Preserve history when truth changes
Current truth evolves. A job moves, quantity is released, cause is corrected and cost arrives.
Keep timestamps, source and previous meaningful states. Do not overwrite the original promise or first diagnosis.
History supports learning and prevents a dashboard from looking perfect only because yesterday’s exceptions disappeared.
Assign data ownership
IT can integrate systems but cannot decide what “material ready” means. Operational leaders own definitions; process users own timely evidence; data stewards monitor quality; technology enforces structure.
Resolve disputes in a governance forum tied to decisions, not as endless report reconciliation.
Publish changes to definitions so trend breaks are understood.
Test the number backward
Select a headline KPI and trace several records to source evidence. Can the team explain why each job is included or excluded?
Then compare the same definition across departments. Differences may reveal timing, unit, identity or ownership problems.
The best dashboard QA is not whether the total matches another spreadsheet. It is whether the total can be reconstructed from agreed events.
Use a layered management rhythm
Shift views focus on immediate safety, quality and flow. Daily views protect material, capacity and delivery. Weekly views examine recurring exceptions. Monthly views connect customer, operational and financial performance.
Reuse definitions across layers while changing aggregation and decision horizon.
This prevents the monthly meeting from rediscovering issues that could have been contained during the shift.
Frequently asked questions
Create a small governance council
Bring together owners from sales, planning, production, quality, supply chain and finance. Meet around disputed decisions, not abstract data architecture.
For each issue, agree the business question, definition, source and owner. Record changes and effective dates. Retire duplicate measures when possible.
The council should be small enough to decide and close enough to operations to test whether definitions work on real jobs.
Display confidence and freshness
Not every number has equal maturity. Show when data was updated and whether it is confirmed, estimated or incomplete.
A current expected margin may use accrued subcontracting; a closed margin uses posted actuals. A supplier date may be confirmed but not dispatched. Users can act intelligently when the confidence state is visible.
Do not turn uncertainty into a hidden decimal. Explain it briefly and name the evidence that will mature the metric.
Design drill-down as part of the KPI
A red total without a route to action creates another meeting. Define how users move from portfolio to product, job, batch, event and owner.
Preserve filters and time reference so the detailed view explains the same population as the headline. Allow authorised users to inspect evidence and reason code without exporting another spreadsheet.
The drill-down is where trust is earned. If users cannot reproduce representative records, they will build private reports.
Review behaviour, not screen usage
Measure whether decisions improve: earlier exception resolution, fewer schedule surprises, faster containment, better quantity reconciliation and more accurate margin.
High dashboard visits do not prove value. Ask which meeting or manual report disappeared and which action now occurs sooner.
Remove metrics that nobody owns. Improve source processes where users repeatedly override data. The objective is a better operating rhythm, with the dashboard serving as shared evidence.
Should there be one dashboard for everyone?
Use one governed data language with role-specific views. Different decisions need different emphasis, not different definitions.
Is real-time data always better?
Only when the source event is reliable and action benefits from speed. Some financial and analytical measures mature later.
Who resolves conflicting numbers?
The metric owner and relevant process leaders should trace definitions and source evidence. IT supports retrieval but should not invent the business rule.
How many KPIs should leaders see?
As few as needed for the decisions, with drill-down for explanation. Add a measure only when someone owns the response.
Make one truth explainable
Start with one recurring management conversation, such as the daily delivery review. Replace its disputed spreadsheet totals with agreed milestones and exception definitions. Keep the first view small enough that every red item can be traced during the meeting.
For four weeks, record questions the dashboard cannot answer, manual corrections and decisions delayed by missing evidence. Improve source capture or definitions rather than adding decorative charts.
Once the review runs reliably, reuse the governed identities and events for weekly capacity, quality and margin views. This staged approach creates trust through operational usefulness.
The strongest sign of success is not a larger screen. It is a shorter meeting with clearer owners, fewer arguments about numbers and earlier action on jobs whose outcome can still change.
One factory truth does not mean pretending every department asks the same question. It means their answers use compatible identities, milestones, quantities and history.
When a leader sees a red number, they should be able to reach the job, batch, event and owner behind it. That is when the dashboard stops being presentation and becomes operational control.