When Spreadsheets Became the Monster
I remember the first time I opened the management pack of a listed company — a mammoth spreadsheet of over one hundred and fifty sheets. Each tab carried its own burden: statutory notes, governance disclosures, operational metrics, tax summaries, and endless expense analyses. You could almost hear the file groan as it opened, calculating line after line of cross-referenced formulas.
The group’s structure was a tangle — a main entity, a major subsidiary, and five others, some of which housed smaller companies rolled up beneath them. Yet the entire financial consolidation rested not on a proper database or consolidation engine, but on a single workbook so big it felt like balancing the JSE on a USB stick.
Architecture of Madness
Each sheet represented someone’s idea of control. In Air Travel, for instance, every entry had to specify not only the flight cost, but the name of the airline used — British Airways, Emirates, Mango — as if the mere mention of a carrier would add integrity to the data. The intent was noble; the execution was chaos.
In theory, we were building information; in practice, we were breaking it. The design violated the spirit of Codd’s normalization — one fact in one place, relationships explicit and controlled. Instead, the same cost centre reappeared with slight spelling changes; suppliers had multiple aliases; calculations hid in deep columns that no one dared to touch.
The Control Sheet That Controlled Nothing
To tame the sprawl, the team built a master Control Checklist — one sheet to rule the rest, linking to sixty other spreadsheets across departments. If all reconciliations balanced, you’d get a serene wash of green. If not, it pulsed with errors like a Christmas tree.
It took only a renamed file or a broken path to collapse the illusion. Month-end close became a month-long chase. Every fix unseated another link. The controller’s most-used key wasn’t “Enter” — it was “Undo.”
The Human Error that Broke the Machine
Then came the moment that “screwed the cat.” A junior accountant, eager to please, opened a new nominal ledger expense account — nothing dramatic, just a code for Regional Air Shuttle. That single addition sent ripples through the model.
The pack didn’t know the code. Lookups were rigid; logic was static. Air travel totals drifted between divisions; the control sheet screamed red. Formula trails stretched into oblivion. What had been an elegant veneer of control gave way to manual overrides and finger-pointing.
Whatever Happened to Real Consolidation?
Before the spreadsheet empire, group results were consolidated inside the ledger. Subsidiary trial balances fed the parent; eliminations removed inter-company noise; journals had provenance. Double-entry guarded the gates: every debit had a credit, every adjustment a paper trail. You could audit it line by line, company by company.
In the workbook world, consolidation morphed into interpretation. Numbers no longer lived in a system — they lived in cells. And whoever controlled the cells controlled the narrative.
The Cost of False Efficiency
It must have taken months — and megabucks — to build that pack. I sometimes wonder if the board understood how exposed they were: a single misplaced formula could tilt reported earnings. But the deeper ailment wasn’t Excel; it was the belief that IT spend is a grudge purchase, tolerated because it doesn’t “add profit.”
Now the same organizations are hunting an AI silver bullet to cleanse data, forecast sales, and consolidate results. Yet if your data architecture is sand, no model — human or artificial — will hold the house.
Reflection
I can still see that pack — tabs flickering like a casino dashboard, promising insight while hiding disorder. It was an impressive monument to effort, and a cautionary tale of misplaced trust.
We didn’t just build a reporting tool — we built a monster. And when a junior changed a single code, we finally saw how fragile our empire of spreadsheets had become.