Five Numbers for the Boardroom, Three for the Bin: Cloud Reporting That Lands
Photo: akhenatenator, CC0, via Wikimedia Commons
There is a particular kind of silence that follows slide fourteen of a cloud review. The charts are accurate, the axes are labelled, the legend is present. And the board has stopped listening, because none of it answers the question they actually came with.
Boards do not want dashboards. They want three things: are we spending sensibly, are we exposed, and would we recover if the worst happened on a Tuesday. Everything else is either evidence for one of those, or decoration.
Five numbers that earn their place
1. Cost per customer. Or per order, per claim, per tenant — whichever unit your business actually sells. Total spend tells a board nothing, because spend should rise when the business grows. Unit cost tells them whether growth is profitable. It also has a useful side effect: it forces your team to allocate shared platform costs, which is where the genuinely wasteful line items are hiding.
2. Forecast accuracy. Not the forecast itself — the variance against it. A team that predicts within five per cent and explains the gap is trusted with a bigger budget than one that reports a surprise every quarter. Present it as a percentage and a short reason, not an apology.
3. Recovery time, evidenced. Not the RTO in the contract. The RTO you last demonstrated in a test, with a date attached. Boards have become noticeably sharper about this since the outage years, and a slide reading 'last full restore tested 12 September 2026, took 3 hours 40 minutes against a 4-hour target' does more work than any architecture diagram.
4. Carbon per workload. Sustainability reporting is no longer a nice-to-have for UK boards. Streamlined Energy and Carbon Reporting obligations, investor questionnaires and customer procurement forms all want a number. Give them intensity rather than totals — emissions per transaction, or per active user — plus the region choices that drive it. A rising total with a falling intensity is a story you can tell. A rising total with no denominator is not.
5. Concentration risk. The share of your estate sitting with one provider, in one region, behind one account team, or dependent on one person who knows how the billing works. This one is uncomfortable and therefore valuable. It is also the metric most likely to prompt an actual decision.
Three numbers that get politely ignored
Uptime percentage. Once you are above the threshold your service promises customers, 99.95 and 99.99 are indistinguishable to the person buying your product and increasingly indistinguishable to a board. It is a number that flatters suppliers and reassures nobody. Time to recover, and time to detect, say far more about resilience than a monthly average that hides every bad afternoon.
Raw cloud spend. A single total invites a single response: cut it. That is how teams end up with fragile savings, deferred maintenance and a bigger bill two quarters later. Spend without a denominator is a mood, not a metric.
Counts. Number of services in use, number of certifications held, deployments per day, dashboards published. These measure activity, and boards are not paying for activity. If a count cannot be linked to cost, risk or recovery, it belongs in the engineering review, not the board pack.
The 2026 question nobody had on the template
Artificial intelligence has added a line to every board agenda this year, and the naive answer is total AI spend. Resist it. Boards want to know what the spend is buying: cost per inference, cost per active user of the feature, or the gross margin of the AI-enabled product. A team that arrives with 'here is what our AI workload costs per resolved customer query, and here is where it is heading' will get a very different conversation from one that arrives with a token count.
How to present it in one page
Each of the five numbers needs four things: the trend over at least four quarters, the threshold you consider healthy, one sentence on why it moved, and the name of the person who owns it. No more than that. If a number cannot survive that compression, it probably is not a board metric — it is an operational one wearing a suit.
And keep a slide in reserve for the question that always comes: what would you do with ten per cent more, or ten per cent less? A team that can answer in units rather than totals has moved from reporting on the cloud to running it. That is the shift boards are actually waiting for.