
Averages, bar charts and percentage change are not evidence. Three actions that stop an ERP, Copilot or analytics programme being built on misleading numbers.
Summarised from the Go2Cab business insight series
Organisations reach for a tool as the answer to business transformation, business insight and operational problems — a habit accelerated by how easily an app can now be built and deployed on a phone. What gets skipped is the root cause. The result is investment in band-aid solutions justified by the idea that doing something beats doing nothing.
The evidence that follows is usually averages, tallies and percentage change over time. That is data, but in most cases it is the wrong evidence — and it pushes people, at board level or on the front line, toward the wrong decision exactly where operational dynamics and measurable objectives are out of sight.
The first reaction is to launch surveys, appoint change and programme managers, and run brainstorming workshops. Budget burns quickly and the payback is poor. A small team with the right skill set achieves far more.
That team selects the vital data to gather, designs the best mechanism to collect it, and — just as importantly — explains how the analysis will be used, by whom, when and against which measurable business objective.
Most organisations have plenty of data and are unsure what to do with it. The usual conclusion, after a few workshops, is that the problem is volume — so spend begins on big data and machine learning before anyone has proven the limits of simpler, adequate tools.
Modern desktop statistical and system-dynamics tools already handle millions of rows and include genuinely sophisticated algorithms at a reasonable price. The constraint is almost never the tool. It is the absence of a defined decision the analysis was meant to support.
Averages hide the dynamics that decide outcomes. Managers ask for one magic number for cycle time, delivery, or defects; senior managers then average the averages their teams reported, and by the time the figure reaches a decision-maker the data is so aggregated that it means nothing.
Three-point estimates — worst case, base case, best case — are slightly better than one magic number, but they still do not carry the variation, sequence or behaviour that actually drives the result.
Bring bad evidence into Microsoft Dynamics 365 or Business Central and configuration locks the error in place: the workflow is automated, nobody questions the number, and the average becomes policy. Lean Six Sigma value stream analysis exists to establish a measured baseline first, so process, capability and analytics design start from how work actually flows.
Microsoft Copilot raises the stakes. An AI assistant grounded in averaged, aggregated measures answers confidently and quickly — and quickly wrong. Copilot and any advanced analytics layer are only as trustworthy as the measures behind them, which is why the measurement model is engineered before the assistant is switched on.
The takeaway
Pick one measurable objective, define the evidence that proves it, then let Dynamics 365, Business Central or Copilot carry that measure — not a bar chart of last quarter.