The Number Nobody Checked

The consolidation rule that was wrong for three quarters

A mid-market financial services firm. Standard group consolidation. One entity's intercompany transactions were being double-counted, but only on a specific elimination rule that triggered when two particular subsidiaries were involved. The variance was small enough to sit within tolerance each month. Nobody flagged it. The error had been compounding for three quarters before we found it during a routine audit.

What would have happened if we hadn't? Eventually, the external auditors would have. At that point it's not an "adjustment." It's a conversation about controls, about restatements, about why nobody caught it sooner.

The formula that forgot about time

A manufacturing company running a planning model in Excel. One formula was pulling an FX rate from a reference table. The rate was hardcoded from the previous financial year. Eleven months of forecasting at the wrong exchange rate. The variance looked like operational performance noise, not a formula error. The finance team had been explaining away the discrepancy as market conditions.

When we traced the root cause, the fix took five minutes. The misinformed decisions it had influenced over those eleven months were harder to undo.

The 4% nobody could explain

A PE-backed SaaS company. One cost centre was consistently 4% higher than expected. The FP&A team assumed it was a timing difference. The business unit thought finance was double-counting something. Both sides had been having the same circular conversation for two quarters.

The cause: a circular reference in the allocation model. It only inflated that specific cost centre because of how the allocation logic cascaded through three levels. It wasn't visible in the formula view. You had to trace the data flow end-to-end to find it.

The entity that broke the model

A growing group with a consolidation model in Anaplan. The model worked perfectly for 14 entities. When they added entity 15 (a new acquisition), an allocation rule that referenced a hard-coded list of entities didn't include the new one. The model ran without errors. It just silently excluded the new entity from certain overhead allocations. The numbers looked plausible. Nobody questioned them for four months.

94% is not a scare tactic

A 2024 peer-reviewed study covering 35 years of research found that 94% of business spreadsheets used in decision-making contain errors. That's not a figure we use to frighten people into buying software. It's a measurement of a phenomenon we observe routinely.

The errors we find aren't usually the result of incompetence. They're the result of complexity. Models evolve. People leave. Assumptions get embedded and forgotten. Business changes don't always trigger model changes. Over time, the gap between what the model assumes and what the business actually does gets wider. Quietly.

Finding them on your terms

The question isn't whether there's an error in your model. The question is whether you find it on your terms or someone else's.

The Bolt Blueprint includes a risk register built to answer exactly that question. Not because we're pessimistic. Because the errors nobody's looking for are the ones that matter most.