A dashboard shows what you already knew to look at.
The dimensions a record needs before analysis is possible, the four questions worth answering, and why comparison is where analytics earns its keep.
There is a difference between a dashboard and analytics, and most manufacturing software sells the first while using the language of the second. A dashboard shows you the current value of things you already knew to look at. Analytics answers questions you had not thought to ask — which machine-operator-part combinations produce the most rejections, whether last month's improvement survived, whether two plants differ because of their machines or their practices.
Getting to the second requires a data model that was designed for questions rather than for screens.
The joins that make analysis possible.
Machine, cell and plant
So a figure can be aggregated upward and drilled downward without recomputation, and so two sites can be compared using one definition.
Part, job and route card
So performance can be attributed to work rather than to time. Without this, a slow month and a difficult product mix look identical.
Operator and shift
So a practice problem can be separated from a machine problem — carefully, because this data is easy to misuse and should inform training rather than blame.
Time, at second resolution
So short stops survive aggregation. Data stored at shift granularity has already discarded the losses most worth finding.
Four questions, in rising order of value.
What happened. OEE, downtime, output, energy — by machine, shift, part and period. This is table stakes and every vendor provides it.
Why it happened. Losses attributed to causes, rejections correlated with the machine state at the moment they were produced, tool age set against cycle-time drift. This requires the joins above and is where most platforms thin out.
Whether it changed. Before-and-after around a specific intervention, with the same definitions on both sides. Plants improve things constantly and rarely confirm the improvement held, which is how a fix that quietly stopped working goes unnoticed for a year.
What is different. Two machines running the same part, two shifts on the same machine, two plants in the same group. Comparison is where analytics earns its keep, and it only works if every unit computes the KPI identically — which is an argument for one platform rather than three spreadsheets.
Analytics that stays inside a vendor's interface is half a product.
Every view exports to Excel and PDF, because that is what a customer audit, a management review and an internal quality pack actually consume. Scheduled reports arrive by email or WhatsApp so nobody has to remember to open a dashboard.
Where a plant has its own BI stack, the underlying data is queryable rather than trapped: it is your record, on your server, in an accessible schema. And where a standard view does not answer a specific question, our team builds the report — the platform includes custom dashboards and reports rather than charging per report, which matters because per-report pricing quietly becomes the largest line in year three.
The question a dashboard could not answer.
A plant's rejection rate rises for six weeks with no obvious cause. The quality dashboard shows the rate; it does not show why. Filtering rejections by machine shows they are spread across the floor, which rules out a single machine. Adding the part dimension shows they concentrate on two part numbers. Adding tool age shows both parts use the same insert grade, and the rejections cluster in the last third of each tool's life.
The correction is a tool-life threshold change on two part numbers, and it took four questions to reach. None of them were answerable from a dashboard alone; all of them were answerable from a record with the right dimensions in it.