Fix a Model That Does Not Balance

Almost every model fails its balance check at least once. Professionals are not people who never break a model. They are people who can find the break in minutes instead of hours. This lesson gives you a short method for tracking down errors, and shows how to use AI as a debugging partner without letting it guess.

What You'll Learn

• How to read the size of the error for clues
• The five most common reasons a model does not balance
• How to give AI enough context to find the bug
• Extra checks that catch errors the balance check misses

Step 1: read the error before you hunt

The difference in your check row is a clue. Write it down for each year, then ask four questions.

• Which year breaks first? Start there. Later years often break only because the first one did.
• Is the error a number you recognize? If Year 1 is off by exactly 50, look for depreciation, debt repayment, or anything else equal to 50. If it is off by 20, look at dividends.
• Is the error positive or negative? The check is assets minus liabilities and equity. A positive number means assets are too high, often because cash is too high. A negative number means liabilities or equity are too high, or assets too low.
• Is it off by double a number? An error of 20 when you expected 10 often means a sign is flipped. A value that should subtract 10 is adding 10, a swing of 20.

Step 2: check the usual suspects

Here are the errors that cause most imbalances in a model like Brightcup's.

Error	How it shows up	Where to look
Flipped working capital sign	Double the change, for example +20 for receivables	C44, C45, C46
Dividends missing from equity	Exactly -20 (equity too high)	C37
Dividends missing from the cash flow	Exactly +20 (cash too high)	C51
Depreciation not added back	Exactly -50 (cash too low)	C43
Cash not linked	Cash stays at 100 or shows 0	C30 should be =C55
Total ranges wrong	Off by one balance sheet line	C34, C38 SUM ranges

Practice: find the planted bug

Try this on purpose. In your working model, change C44 from =-(C31-B31) to =C31-B31. That is the most common mistake beginners make.

Your Year 1 check now shows 20. Here is why: receivables rose by 10. The correct formula subtracts 10 from cash. The broken one adds 10. The swing is 20, so cash and total assets are 20 too high.

Fix it back before moving on.

Step 3: give AI the evidence, not just the symptom

"My model doesn't balance, help" gets generic advice. AI can only reason about what you show it. Give it the error, the formulas, and the numbers.

A good answer will point at the receivables formula and explain the double-swing. If AI blames something else, ask it to check its reasoning using the error size. Pushing back with evidence is a skill worth practicing.

Step 4: add checks that the balance check misses

A model can balance and still be wrong. If you typed the wrong revenue, everything still balances. Add a few extra checks below the cash flow statement.

Check	Formula for Year 1	Should be
Cash ties out	=C30-C55	0
Retained profit ties out	=(C37-B37)-(C27-C11)	0
Gross margin is as assumed	=C19/C17	40%
Debt never negative	=C36>=0	TRUE
Cash never negative	=C30>=0	TRUE

The last two are business checks rather than math checks. Brightcup's debt is 50 by Year 3. If you extended the model to Year 5 with the same repayments, debt would go to zero and then negative, which is impossible. A check flags that before anyone else does.

Ask AI to suggest more.

Key Takeaways

• Read the size and first year of the error before hunting. Matching inputs and doubles are strong clues.
• The usual suspects: flipped working capital signs, missing dividends, depreciation not added back, unlinked cash, and wrong SUM ranges.
• Give AI the error amount, the formulas, and the numbers, then ask it to explain the exact size of the error.
• Add extra checks for tie-outs, margins, and impossible values like negative debt or cash.
• A balanced model can still be wrong. Checks catch math errors, not bad assumptions.