The Real Cost of Siloed Systems and Spreadsheets in a Plant
DBR77 Team7 min read
On this page
Siloed systems cost a plant in five places: hours spent typing the same data twice, time spent reconciling numbers that disagree, problems seen too late to prevent, defects found after they leave the line, and decisions made on the wrong version of a figure. None of these shows up as a budget line, so it is rarely fixed. You can measure it in two weeks by following a few event types across departments.
The cost is real but scattered across maintenance, quality, planning and the meetings where people compare spreadsheets.
Why the cost is hard to see
No one buys silos on purpose. Each tool was chosen for a good reason: a CMMS for maintenance, a quality database for audits, a spreadsheet for the plan. Each works for its own team; the cost appears between them, where nobody owns it.
Most of that cost is salaried time, so it never appears as an invoice. A planner who spends 40 minutes a day rebuilding the schedule from three sources is paid either way.
The wider business world has tried to put a number on bad data. Gartner research from 2020 found that poor data quality costs organizations at least $12.9 million a year on average. The figure is not specific to manufacturing and is not a plant estimate, but it shows that the cost is large enough to be worth measuring.
- Who pays
- Operators, technicians, shift leads
- Two-week check
- Count re-entries per event type
- Who pays
- Managers, controlling, planners
- Two-week check
- Log minutes spent on whose number is right
- Who pays
- Production, customers
- Two-week check
- Time from event to first action
- Who pays
- Quality, sales, finance
- Two-week check
- Lots on hold found after release
- Who pays
- Planning, purchasing
- Two-week check
- Plan changes caused by stale data
Siloed systems cost a plant in five places, and none is a budget line.
Five places silos cost money
| Cost | Where it shows up | Who pays | How to see it in two weeks |
|---|---|---|---|
| Double entry | The same stop, defect or move typed into two or three systems | Operators, technicians, shift leads | Count re-entries per event type |
| Reconciliation | Meetings and emails to agree on one number | Managers, controlling, planners | Log minutes spent on "whose number is right" |
| Late signals | Problems seen next day, from a report | Production, customers | Time from event to first action |
| Quality escapes | Defects found after the lot moved on or shipped | Quality, sales, finance | Lots on hold found after release |
| Wrong decisions | Plans and purchases based on an old figure | Planning, purchasing | Plan changes caused by stale data |
The fourth row is usually the most expensive. ASQ separates internal failure costs, for defects found before the customer receives the product, from external failure costs, for defects found after. When quality learns about a machine problem two shifts late, defects move from the first group to the second: rework turns into returns, sorting at the customer and complaints.
The other rows add up through repetition. Five minutes of retyping per stop is small. Across 30 stops a day, it is a part-time job.
Spreadsheets: the glue that holds silos together
Spreadsheets fill every gap between systems. They are fast, flexible and owned by the person who needs them, so they survive every software project and become the real record for planning, KPIs and open issues.
Raymond Panko's review of fifteen years of research concluded that spreadsheet errors are both common and non-trivial, and that only one technique, cell-by-cell code inspection, had been shown to be effective. Few plants inspect their planning workbook cell by cell.
On the shop floor it gets worse. Versions multiply: the plan exists as "plan_v7_final" on one laptop and "plan_v8" in an email. The KPI workbook depends on one controller's macros, so when that person is away, the numbers stop. And a spreadsheet can list 25 open issues but cannot assign them or check that they closed.
An illustrative example: one moist resin lot
Illustrative example: a plastics plant receives a resin lot that is slightly too moist. Here is how the event travels through four separate tools.
- Receiving (06:30). The warehouse books the lot in the ERP add-on and files the certificate as a PDF.
- Production (08:15). Splay marks appear on press 4. The setter raises the dryer temperature and notes it on the shift sheet.
- Quality (13:40). An inspector finds the defect, opens a record in the quality database and emails the warehouse to ask which presses got the lot.
- Warehouse (15:10). The reply lists two more presses. By then 1,800 parts from those presses have been packed.
- Planning (next morning). The planner learns about the hold in the daily meeting and moves three orders by hand.
The event took more than a day to reach the plan, five people searched for or retyped the lot number, and the cause sat on a paper shift sheet. In one shared record, the quality hold would have reached the warehouse and the plan in the same minute.
How to measure the cost in your own plant
Count for two weeks:
- Pick three event types that cross departments, for example a machine stop over 15 minutes, a quality hold and a material shortage.
- Follow ten of each. Note each system it was typed into, each person who touched it and the time to first action.
- Time the reconciliation. In meetings, note the minutes spent agreeing on numbers rather than acting on them.
- List the spreadsheets that feed a decision, with their owner and how often they are updated.
- Count KPI disagreements. Compare OEE, scrap and on-time delivery for the same week from production and from finance.
The result is a list of handoffs ranked by hours and delay, which is the starting point for a business case. How to Build a Business Case for a Plant Operating System shows how to turn it into numbers without inventing savings. If the KPI counts disagree, start with How to Build a Real-Time KPI System for Your Factory.
How this works in DBR77 IRIS
DBR77 IRIS runs production, maintenance, quality, warehouse and planning on one shared data model, so an event is entered once. Starting an order in IRIS MES reserves its material in IRIS WMS, and finished output is booked into stock. A failed incoming inspection in IRIS QMS puts the lot on hold in IRIS WMS, so it cannot be picked. A machine failure reported in MES opens a work order in IRIS CMMS. IRIS KPI pulls its numbers from the modules, with no manual exports.
Modules can be added one at a time while existing tools stay connected; How to Unify MES, WMS, QMS and CMMS Without Replacing Everything explains how. Pricing is per module, so a plant can start with the handoff that costs most.
When a KPI crosses its alert rule, IRIS raises an alert. Data & AI looks for correlations across modules and proposes recommendations, each with the basis for its score, so the person who decides can see why it was suggested. The IRIS assistant answers follow-up questions from the same data.
FAQ
Is this an IT problem or an operations problem?
Both, but the cost lands in operations and finance. IT connects tools; operations decides which record everyone uses.
Can't a data warehouse or BI tool fix this?
It can fix reporting, but it does not stop double entry or give an event an owner.
How big is the cost in a typical plant?
There is no reliable general figure for a mid-size plant. Measure your own handoffs: a number from your own events convinces more than any benchmark.
Conclusion
Start the two-week count with three event types that cross departments, ten of each. Rank the handoffs by hours and delay, fix the most expensive one first, and judge the result by how fast events reach the right person. If production and finance report a different OEE or scrap for the same week, fix the KPI definitions first.
Sources
- Gartner, Data Quality: Why It Matters and How to Achieve It, 2020 research
- ASQ, Cost of Quality (COQ)
- Raymond R. Panko, Spreadsheet Errors: What We Know. What We Think We Can Do, 2008