The Excel formulas people get wrong most often

What the wrong answers were

121 signed-in learners gave 1,093 wrong answers in the same five weeks. The checker recognised 512 of them as a known mistake; the rest were wrong in ways no exercise anticipated. The 439 named mistakes in spreadsheet exercises break down like this.

1. Right pieces, wrong calculation

   37% · 163

   Dividing the wrong way round, subtracting cost from revenue instead of the other way, working out the size of a discount instead of the price after it, or a function's switch set the wrong way.

2. Ranges not locked before filling down

   21% · 93

   Right in the first row and wrong in every row below it, because the ranges slide when the formula is copied. Dollar signs hold them in place.

3. Only works for one row

   14% · 61

   A typed number or a count done by eye that happens to fit the first row, so the result breaks the moment the formula is filled down or the data changes.

4. Off by one

   13% · 56

   A boundary or a rank one step out: greater-or-equal where the rule says greater than, the third-largest instead of the top three added up, the smallest instead of the third-smallest.

5. Approximate match left on

   3% · 15

   VLOOKUP without FALSE, or MATCH without 0, assumes the table is sorted and can return a nearby row's value instead of an error. Wrong answers that look right.

6. A condition missed

   3% · 15

   Testing one condition when the task has two, or testing against zero when the rule has a threshold.

7. Ranges of different sizes

   3% · 13

   A sum range one row shorter than its criteria range. Excel lines them up from the top and silently drops the last row.

8. Wrong column, or one step short

   3% · 11

   A lookup that finds the right row but returns another column, or a two-step lookup that stops after the first step.

9. Compared against the wrong kind of value

   2% · 10

   95% is stored as 0.95, so a test against 100 never passes. Values stored as text likewise never equal real numbers.

10. Error case not handled

    under 1% · 2

    A lookup for a value that is genuinely missing, left to show #N/A in a report.

How these numbers were counted

• Everything comes from this site, 13 September to 17 October 2026: its analytics for every visitor (2,826 in that time), and its database for the mistakes of signed-in learners.
• Started means the first real edit in an exercise, not opening the page. Finished means a first passing answer, counted once per exercise per visit.
• Wrong answers per finish is every failed check divided by every finish. Opened the solution is the people who opened one as a share of the people who started; it includes people who opened the answer without trying, which is why DATE’s figure is so high.
• Each exercise is counted under the first function it teaches. Functions have between 1 and 7 exercises at different difficulties, so a function whose exercises are harder will look harder.
• Left out for having fewer than 30 people: COUNTIFS, IFS, AND, TEXT, DATEDIF, EOMONTH, AVERAGEIF, IFERROR, NETWORKDAYS, ROUND, TRIM.
• Named mistakes cover spreadsheet exercises only; on quiz questions and shortcut drills every wrong option is filed under one code, which says nothing about formulas. Only counts left the database: no names, no emails, no formulas anyone typed.
• The data: starts, finishes, wrong answers and solutions opened, by function (CSV) and named mistakes, by exercise (CSV).