Operational Reporting

Seco is moving from selling tools to selling manufacturing outcomes, delivered through Innovation Hubs in the UK, Germany and the US. Platform 1 replaces the Smartsheet project tracker; this report reads what Platform 1 captures.

It answers exactly seven questions, and no eighth:

1. Who is working on what?
2. What stage is each project at?
3. What are the project leads working on next week?
4. What is blocked, and by what?
5. Where is capacity?
6. How long from request to review, and are we delivering on time?
7. How many projects by region, segment and month?

Platforms 4 and 5 are two reports rather than one because the business chose not to build low-level security and split by audience instead. Honouring that at the report layer only would be cosmetic: one careless visual reintroduces the problem. So the boundary is drawn in the semantic model.

No cost, budget, rate, margin, value, invoice status or actual hours. Not hidden by row-level security, not hidden by object-level security — absent. This model uses planned assignment hours and available capacity hours, which is what every forward-looking operational question needs.

The test a BI developer can apply without asking anyone: if a visual needs a financial field, it is a Platform 5 visual.

The audience are engineers and engineering managers. They do not think in percentages against a colored delta; they think in tolerances — a value, a target, and the band inside which the value is acceptable. So the default KPI form here is a measured value plotted against its band, not a 40pt number with a green arrow.

A band implies somebody agreed the tolerance. Today, of fourteen bands, none has an agreed value: four have a named owner, ten have neither owner nor value. Every one therefore renders as a plain number labelled no target set. That is correct, not unfinished — a fabricated band gets read as an agreed one and quoted back in a review meeting. The list of unowned targets is one of the more useful outputs of this exercise.

1. A true bullet-with-band visual. There is no native one, and AppSource is closed in a Sandvik tenant. It gets approximated with a line-and-clustered-column combo plus constant lines — good, not exact, and the band fill is the compromise.
2. A composed variance label — glyph, number and word together, sized and colored independently within one line. Power BI will not compose it, so it becomes a single DAX string measure and loses per-part typography.
3. Reflow. A canvas that adapts from a wall display to a laptop. Power BI is absolute-positioned at a fixed size, and mobile layouts are authored separately per page. That is why this instrument does not resize either: pretending otherwise sets an expectation the built report cannot meet.