Somebody Spends Three Days a Month Building the Board Pack
Four exports, a master spreadsheet, a set of formulas nobody else can follow, and a version emailed round the night before. It works, it costs a senior salary a fortnight a quarter, and page four still gets queried every single meeting.
What the Manual Pack Really Costs
Not the software licences. The time, the delay, and the slow erosion of confidence in the figures themselves.
Why it is nobody’s fault, and why it persists
The pack exists because the board needed numbers and the systems would not produce them together. The person who built it did something useful under time pressure and has been maintaining it ever since. Replacing it is therefore not a criticism of them, and treating it as one is the fastest way to lose the only person who knows how the figures are actually derived. They should be the first person in the room, and usually the one who ends up owning the dashboard.
How a Figure Travels From Four Systems to One Screen
A data pipeline is a scheduled job that copies, tidies and combines. Nothing more mysterious than that, and every stage is a place where things break in a way you want to find out about automatically.
The accounts package, the CRM, the job system, the till. Read-only, on a schedule, never writing back.
An untouched copy, kept. When a figure looks wrong in March you can go back and see what the source actually said.
Dates into one format, the same client under three spellings joined up, cancelled rows excluded, currencies converted.
Active client, completed job, recognised revenue: written down, signed off, implemented in exactly one place.
Dashboard, scheduled email and spreadsheet download all reading the same prepared figures, so they cannot disagree.
A silent broken pipeline is worse than no dashboard, because a stale number looks exactly like a current one.
Where these projects actually get stuck
Almost never at step one. Reading data out of a modern system is a solved problem. The work concentrates at step three, where a client appears as three slightly different names across two systems and somebody has to decide which is right, and at step four, where the CRM counts a deal when it is marked won and the accounts package counts it when it is invoiced.
Both are business decisions wearing technical clothing. We will run the workshop and write the definitions down, but somebody with authority has to choose, and the project cannot finish until they do. Saying that in week one is considerably cheaper than discovering it in month four.
The pipeline, the reporting store and the scheduled jobs are ordinary backend engineering, and they are the bulk of the build.
The backend and integration workThree Tests Before a Number Goes on the Screen
Most dashboards fail by being generous. Everything measurable gets added, the screen becomes wallpaper, and within a quarter people are back to asking somebody for a spreadsheet.
A named person owns it
Not a department. A person, who would be expected to explain it if it moved sharply and who has some ability to change it. A metric nobody owns is decoration.
A decision hangs on it
If it doubled tomorrow, what would you do differently this week? If the honest answer is nothing, it belongs in an annual review rather than on a screen people open every Monday.
It has one written definition
In a sentence a new starter could apply without asking. If two people in the business would count it differently, that is the work to do before any code is written.
What we will argue against putting on it
Totals since records began, which only ever go up and tell you nothing. Percentages with no denominator on screen. Averages hiding two completely different populations, such as a mean job value across retail and contract work. Anything nobody could act on this quarter. And a seventh chart added because there was a gap in the grid. We would rather ship five numbers that change behaviour than thirty that get glanced at.
What Replaces Page Four
A small screen that answers the standing questions, with the detail one click behind it rather than printed for everyone.
Revenue, month to date
£412k
Gross margin
31%
Debtor days
47
Jobs awaiting sign-off
18
Illustrative layout only. Every figure and label above is invented to show how a board summary is structured; none of it comes from a real client or a real month.
Four figures a director can absorb in ten seconds, one comparison across sites, and a date stamp saying when the data was last refreshed. Everything else sits behind a link for the person who needs to go digging.
How Fresh the Numbers Are Is a Budget Decision
Real-time is a genuine engineering commitment rather than a setting. Pick the slowest cadence that still lets you act, then spend the difference on something that changes a decision.
| Cadence | What it costs | When it is the right choice |
|---|---|---|
| Overnight | Lowest to build and cheapest to run. One scheduled job, easy to re-run when a source is late. | Board packs, monthly management reporting, funding returns, anything reviewed on a weekly or monthly rhythm. This covers most of what UK SMEs actually need. |
| Hourly or every few minutes | Moderate. More runs, more chances to fail, so it needs proper alerting and retry handling. | Same-day operations: dispatch, engineer scheduling, stock across multiple branches, a contact centre watching today’s queue. |
| Streaming, near real-time | Substantially more, to build and every month afterwards. A different architecture, not a faster schedule. | Live service dashboards, fraud and safety alerting, anything where a delay of minutes has a direct operational or financial consequence. |
We ask one question to settle it: what would you do differently if you knew this five minutes sooner rather than tomorrow morning? If nobody can answer, overnight is the right answer and the saving is real.
The Numbers That Should Not Be in Everyone’s Inbox
A dashboard concentrates sensitive figures in one place, which makes access control part of the build rather than an administrative afterthought.
Restricting the view
- Row-level access so a branch manager sees their site and the regional director sees all of them
- Column-level control, because headcount cost and margin per client rarely belong to the same audience as volume
- Aggregate-only views, where a team can see the trend without seeing individual people or clients
- Access tied to role in your existing sign-in, so it changes when somebody moves department rather than when they leave
- Exports inheriting the same restrictions, since a download is how a restricted figure usually escapes
The data protection side
- Only the fields the reports need are copied into the reporting store, rather than a whole table because it was easier
- Aggregation on the way in where individuals are not needed, which narrows the whole conversation
- Retention applied to the reporting copy as well as the source, so history does not quietly accumulate forever
- UK or EU hosting regions where residency is a requirement, stated in writing
- A log of who viewed which report, which matters for staff and patient data more than most people expect until an incident
Quite Often You Should Buy the Tool Instead
Power BI, Looker Studio, Metabase and their competitors are capable and cheap. We would rather point you at one than sell you a build you did not need.
Buy an off-the-shelf BI tool when
- The data is already in one or two places and does not need reconciling first
- Your team wants to explore and slice the figures themselves rather than read a fixed screen
- Only a handful of people will ever look, and a per-seat licence is trivial against a build
- You are already paying for it inside a Microsoft licence your organisation holds
- The reporting need is likely to change shape every few weeks while the business finds its footing
Build it when
- Getting the data into one trustworthy shape is the actual project, and the charts are the easy last mile
- The logic does not fit in a chart builder: contract-specific rules, apportionment, or a calculation with exceptions
- Your customers are the audience, and it has to live inside your product under your branding and permissions
- Everyone in the organisation needs access and per-seat licensing becomes the most expensive line on the page
- It has to run unattended and reliably for years, with alerting when a source system changes underneath it
A common and sensible outcome is both: we build the pipeline and the reporting store, and your team points an off-the-shelf tool at it. See everything we build or how we price work.
What Finance Directors Ask Us
How long before the monthly pack stops being assembled by hand?
For a first version covering two or three source systems, usually six to ten weeks, and the pack is normally produced twice in parallel before anyone stops doing it manually. That parallel month is not wasted time. It is how the figures earn trust, because the first time the dashboard and the spreadsheet disagree you find out which one was wrong, and it is not always the new one. We would rather spend a month proving the numbers than launch something the finance director quietly continues to check by hand for a year.
Two of our systems disagree about revenue. Which one does the dashboard believe?
Whichever one you decide, recorded in writing, before we build anything. This is the single most common reason dashboard projects stall, and it is not a technical problem. The CRM counts a deal when it is marked won, the accounts package counts it when it is invoiced, and both are correct for their own purpose. Someone with authority has to name the definitive source for each figure and accept the consequences. We facilitate that conversation and document the outcome, then the pipeline enforces it consistently so the disagreement never resurfaces as a surprise in a board meeting.
Does the dashboard keep its own copy of our data, and where does that sit?
Usually yes, and deliberately so. Querying your live operational systems directly makes reports slow and puts load on software people need for their day job, so we copy the data into a reporting store on a schedule. That store lives in a UK or EU region where you need it to, holds only the fields the reports actually use, and carries its own retention rules so historical personal data is not accumulated indefinitely by accident. Where a report only needs totals rather than individuals, we aggregate on the way in, which narrows what is held and simplifies the conversation with your data protection lead.
Can we still get the figures into Excel for the board papers?
Yes, and we build it in rather than treating it as a failure of the dashboard. Board papers, funding returns, auditors and accountants all want a file, and refusing to provide one guarantees somebody rebuilds the export by hand. The difference is that the download comes from the same agreed definitions as the screen, is dated, and says which refresh it came from, so two people comparing their copies can tell whether they are looking at the same moment. Scheduled exports by email to a distribution list are straightforward to add.
What happens when somebody wants to change a definition six months later?
It is a change request, and it is priced as one, because a definition is code rather than a setting. The important part is what happens to history. Changing how an active customer is counted will move the prior-year figures too, so we agree in advance whether the new definition applies from a date or is applied retrospectively, and the dashboard records which definition produced which number. That way the first person to notice a chart has changed shape gets an explanation rather than losing confidence in the whole thing.
Send Us Last Month’s Board Pack
Tell us which systems it is assembled from and which page gets queried most. You get back a scope, a refresh recommendation with the cost difference spelled out, and a sterling figure excluding VAT.