Start with the decisions the owner actually makes

A weekly dashboard should begin with recurring decisions, not with whatever a software package can graph. Ask what the owner genuinely needs to decide each week: whether demand is sufficient, whether enquiries are converting, whether delivery capacity is tightening, whether cash needs attention, or whether a quality problem is emerging. Then choose only the signals that help make those decisions.

The Australian Bureau of Statistics reported that in 2024–25 only 10% of all Australian businesses actively collected or analysed data to make informed decisions, while 7% measured the contribution of digital activities to overall business performance. Those figures cover all businesses, not a small-business-only sample, and they do not prove that a dashboard will improve performance. They are useful context for a simpler point: collecting more numbers is not the same as building a decision habit.

Use a minimum operating view before a management pack

For many small businesses, a useful starting view can fit on one page: one demand measure, one conversion or response measure, one delivery or capacity measure, one cash measure, and one quality or customer signal. The exact measures depend on the business model. Every measure should answer a recurring question and have a clear owner.

This is deliberately smaller than a traditional KPI pack. If a number does not change a decision, trigger an investigation or expose an exception, it is a candidate for removal. business.gov.au similarly recommends starting with goals before choosing metrics and targets, then using the performance data to decide what to change.

Define each number before you automate it

Write down the definition, source and owner for every measure. Suitable enquiries is not useful if one person counts every form submission while another excludes poor-fit leads. Work in progress is not useful if the sales system and job system disagree about what has actually started. Cash due this week is not useful if invoice status is reconstructed manually from several places.

A business does not need one giant database to solve this. It needs an authoritative home for each important fact and a clear rule for what wins when copies disagree. If the weekly view takes hours of spreadsheet repair or argument before anyone trusts it, the reporting process itself is part of the operating constraint.

Add exceptions and a short operating narrative

Numbers rarely explain themselves. Beside the measures, record the exceptions that need attention and the decisions that need an owner. A useful weekly note can be three sentences: what changed, what might explain it, and what needs a decision. Keep observation separate from interpretation. “Suitable enquiries fell from 18 to 11” is observable; “demand is weakening” is a hypothesis that still needs context.

This is where a small operating view becomes more useful than a passive dashboard. The aim is not to read numbers aloud. It is to notice where reality differs from the plan, decide what deserves attention, assign the next action and return the following week to see what happened.

Review the view as aggressively as you review the business

Every few weeks, ask whether each measure changed a decision or helped expose a constraint. Remove vanity measures, tighten definitions and add a metric only when a recurring decision cannot be made from the current view. business.gov.au recommends regularly reviewing digital strategy and monitoring whether tools are still helping the business save time or money; the same discipline should apply to the report itself.

Do not automate a fragile report merely to make it arrive faster. First stabilise the definitions and source ownership. Then automate collection where the boundary is clear, while leaving interpretation and judgement visible. A polished dashboard built on disputed inputs can make a weak operating system look more certain than it is.

The practical weekly test

For the next four weeks, write down five things before opening a dashboard product: the recurring decision, the measure that informs it, the authoritative source, the person who owns the next action, and what would count as a meaningful exception. Use a spreadsheet or document if that is enough.

If the view becomes trusted and repeatedly useful, then improve the delivery mechanism. If the data cannot be assembled without reconstruction, follow the one-source-of-truth problem. If the view is useful but takes too much manual work, inspect reporting automation only after the definitions are stable. If the owner still has to narrate everything from memory, use the weekly owner brief to separate the numbers from the judgement.