BI Reporting & Insights

No, and the boundary is deliberate. I do not sell dashboards, implement platforms or write data pipelines. Those are engineering jobs, and a company that needs them should hire engineers. What I fix sits upstream of the build, which is where most reporting programmes go wrong: deciding which decisions the reporting must serve, choosing the five to nine metrics that serve them, writing one defensible definition for each, naming who owns it, and wiring the whole thing to a review with real decision rights. Where genuine engineering work is required, I will specify what should be built and to what standard, then hand that specification to your team or your vendor. The distinction matters commercially too. You are paying for judgement about what to measure, not for build hours.

Whichever ones you already own. I am tool-agnostic by conviction rather than diplomacy. In 19 years I have not seen a management reporting problem primarily caused by the choice of software, and I have seen several made worse by a purchase that postponed the real conversation for a year. Spreadsheets, a reporting module inside an ERP, a modern BI suite, or some combination of all three can carry a well-designed scorecard, and none of them can rescue a badly designed one. So I work inside your stack and recommend no purchase as part of this engagement. If, once the design is settled, your current setup genuinely cannot produce a metric on the cadence you need, that becomes a specific, scoped requirement you can take to a vendor with your eyes open.

They share a spine and differ in the entry point. Operations governance starts from how the business is steered — cadence, ownership, escalation — and builds measurement to serve that. This engagement starts from the reporting itself, which is usually how the problem is felt: the pack is too long, the numbers do not agree, the dashboards go unread. In practice the two converge, because a scorecard without decision rights is decoration and decision rights without trusted numbers are guesswork. If your operating rhythm is broadly sound and the reporting is the weak link, start here. If leadership has no working cadence at all, the governance sequence is the better one, and I will say so rather than sell you the page you happened to land on.

Five to nine, and the hard part is the exclusion rather than the selection. Below five you are usually blind to part of the value chain. Above nine, attention thins until none of the numbers is genuinely governed and the scorecard becomes a report by another name. Each metric must earn its place by answering a decision leadership actually makes, and each arrives with a definition precise enough that two people compute it identically, a named owner, a source, a threshold, and an agreed response when it breaches. Anything that fails that test is not abolished; it moves one layer down to the functional review. Leaders usually find the removal harder than the design, because the number being cut is somebody's favourite and its owner is in the room.

Usually the demand side. A capable BI team can build almost anything asked of it, which is precisely the problem when the asking is undisciplined. Every request is legitimate in isolation, and the accumulated estate is one nobody can hold in their head. I do not do the team's job. I give them a defensible specification and a way to say no: an agreed scorecard, written definitions, a named owner per metric, and a change-control rule for adding or retiring one. Most BI teams welcome that, because the alternative is being measured on ticket throughput rather than on whether the business makes better decisions. Where the team is stretched, deciding what not to build is the cheapest capacity available to you.

With the rules, not the database. When two teams produce different figures for one metric, the cause is rarely a missing table. It is that each applied a defensible but different rule — a different cut-off, a different treatment of cancellations, a different exclusion for internal work. Both are honest, which is why the argument recurs monthly and never resolves. So the first move is to place the two calculations side by side and isolate exactly which rule diverges. Then get a ruling on which one the business will use, write it into a definitions register, attach a named owner empowered to change it, and declare one system of record final even for those who lost the argument. Reconciliation stops being a monthly event once it becomes a single governance decision.

With a common spine and honest local detail. A small set of metrics is defined identically across every entity so the centre can compare and consolidate without translating, while each entity retains the operating detail its own management needs. Forcing total uniformity produces numbers that are comparable and meaningless; allowing total autonomy produces a reconciliation exercise every month and figures that arrive stale. The other target in a group is lag, which is usually procedural. Multi-entity reporting is slow far more often because approvals queue across entities than because the data is hard. I have taken a billing approval cycle from roughly two months to 15 days across 75 entities, and most of that came from redesigning where approvals and reporting sat rather than from installing anything.

It opens with a fixed-fee diagnostic: a bounded piece of work producing a written read on what your reporting is currently for, what it should contain, and where the definitions are in conflict. That is deliberately low-commitment, and some companies take the document and act on it themselves. If we continue, the work moves to a monthly retainer scaled to scope and never billed by the hour, because hourly billing rewards slowness and this work should be finite. Most reporting engagements run as a defined block of weeks rather than open-ended. Designing the scorecard is not the long part; running the cadence until the discipline holds without me is. The retainer ends when a named person inside your business chairs the review and maintains the definitions.

Two things, and neither is technological. The first is a change-control rule: adding a metric to the leadership scorecard requires naming the decision it serves and, ordinarily, retiring one that no longer earns its place. That single constraint prevents most accretion, because it forces the trade-off to be made out loud rather than absorbed quietly. The second is ownership with standing. If the definitions register and the review belong to someone junior, the discipline erodes the first time a senior stakeholder wants a special cut. So developing a credible internal owner is part of the work, not an afterthought: they chair alongside me, then instead of me, and they hold the pen before I go. Expect to revisit the scorecard once a year, because the numbers that matter change as the business does.