The transformation graveyard
Every container glass plant we walk into has a story. The ERP that was supposed to be the single source of truth. The MES that never finished rolling out. The IoT dashboards from a pilot three years ago that nobody opens. Ambitious, expensive, and still — here we are — surrounded by spreadsheets.
The pattern is consistent enough that it is worth naming. Transformations fail when they try to replace the plant's existing brain rather than augment it. The spreadsheets exist because they are flexible, fast, and owned by people close to the work. Replacing them requires meeting those virtues, not just the function.
Three principles that keep us sane
- Role before technology. Start with who needs what, when, on what device. The platform falls out of that — it is rarely the starting point.
- Existing data, existing systems. MES, SCADA, ERP, quality systems are already there. Integrate, do not replace. Rip-and-replace doubles the timeline and halves the buy-in.
- Measure in months. First role live in 4–8 weeks. If the first dashboard is not in an operator's hand within two months, the political clock is already against you.
The four roles that actually matter
We have built dashboards for almost every role in a container glass plant. Four matter most:
- Operator (handheld): current product, current setpoints, live defects, next actions, one-tap issue escalation.
- Team leader (tablet / screen): shift scoreboard, jam-up pareto, swab compliance, changeover progress.
- Area manager (browser): line-by-line OEE, top defects, open escalations, standard work adherence.
- Plant leader (browser + email): daily pack, weekly trend, monthly KPI tree vs. plan.
Build these four and 70% of the spreadsheets retire themselves.
Digital audits — the unsexy win
Safety, quality and housekeeping audits are usually on clipboards or three-year-old Excel templates. Digitising them is the least glamorous and most valuable single thing we do. Photos as evidence, escalation in real time, trend analysis by shift and crew — and, crucially, the ability to prove compliance to a customer or regulator without a two-week document hunt.
What not to do
- Do not wait for perfect data. Ship dashboards on 80% data, iterate. The first version reveals what data is actually missing.
- Do not build everything yourself. The plant has vendors, partners, internal IT. Pick your build-vs-buy battles. Our own dashboarding stack is a toolkit, not a religion.
- Do not ignore the floor's existing systems. If a shift leader has a whiteboard with the three numbers that matter, digitise that board — do not replace it with a feature-richer worse version.
A realistic 90-day plan
- Weeks 1–3: Role interviews, existing-data inventory, one-page vision per role.
- Weeks 3–6: First role dashboard prototyped in the tool, tested on one shift.
- Weeks 6–10: Go-live on one line, iterate with the crew, escalation rules wired up.
- Weeks 10–12: Second role dashboard prototyped, third role roadmapped.
By day 90, the plant has two live dashboards, two spreadsheets retired, and a queue of the next ten. The momentum compounds from there.