How to Tell Your Boss You Made a Mistake in the Numbers

Abstract

This article is about how to tell your boss you made a mistake in the numbers, and why the first impulse is so often to say nothing. Four findings from research bear on it. A review of research on moral emotions links shame, a judgement of the whole self, with hiding and denying, and guilt, a judgement of one act, with apologising and repairing. In two experiments, an apology restored more trust than a denial when a job applicant's error was one of competence, though neither response restored it fully. In hospital nursing units, the units with better leadership and relationships detected more medication errors, which the author linked to a greater willingness to report. In four experiments, a short note of self-compassion after a failure went with more motivation to improve. The three steps at the end are practical readings of this research, not tested methods, and they are not advice on accounting, audit or legal matters.

The error you see at 7:10 a.m.

At 7:10 a.m. on a Tuesday you open the board pack again to prepare for the 9:30 call. On page 4 the bridge from last year's EBITDA to this year's does not add up. You trace it to a cost accrual of 340,000 that you posted to January. It belongs in February. January is shown 340,000 too low, and the pack went to the CFO and four directors yesterday evening.

Nobody has said anything yet. Your first thought is not about the number. It is "who has seen it?" Your second is that you could correct the model, say nothing, and let the next version carry the fix. Your third is a hot, tight feeling that you will be remembered as the person who got it wrong.

How to Respond When Your Numbers Are Questioned is about what happens when someone else asks. This article is about the other case: you find it first, and how the story starts is still partly up to you.

Why the first impulse is to say nothing

Tangney, Stuewig and Mashek (2007) reviewed research on moral emotions. In the review, shame is a negative judgement of the whole self. It is linked to hiding, denying and withdrawing, and to defensiveness and distance from other people. Guilt focuses on one behaviour. It is linked to confessing, apologising and trying to undo the consequences. The authors also argue that shame leaves little route to repair, because a flawed self is not mended by an apology, while guilt offers several routes.

Read the Tuesday morning through this. The urge to correct the model and say nothing is the direction the review describes for shame. The thought behind it is not "this figure is wrong." It is "I am someone who gets this wrong." The first is a fact with a fix. The second has nowhere to go but out of sight. This is an inference. The review covers many kinds of wrongdoing, not finance errors, and it cannot tell you what you will feel at 7:10 a.m.

What telling does to trust

Kim, Ferrin, Cooper and Dirks (2004) asked 200 business students to play a hiring manager. They watched an interview with an applicant for a senior tax accountant job. The applicant was said to have filed an incorrect return at a previous employer, understating a client's capital gains. In one version the error came from inadequate knowledge of the tax code, which is a competence matter. In the other it was intentional, which is an integrity matter. The applicant then either apologised, taking responsibility and saying it would not happen again, or denied responsibility.

After the competence version, the apology restored more trust than the denial. After the intentional version the pattern reversed, and the denial worked better. In a second study, with 444 people, a ruling either cleared or found against the applicant. Responses that clashed with the ruling, such as a denial followed by a guilty ruling, lowered trust more than responses that matched it. Neither response restored trust fully.

Read the Tuesday morning through this. A wrong accrual is an error of competence, and in the study, owning an error of that kind did better than denying it. The reverse result for intentional acts is not a reason to deny anything: it concerned an alleged deliberate act, and the second study showed that a response that does not match the facts costs trust. The participants rated a stranger on video, not a colleague, and the setting was a laboratory. The study compared an apology with a denial, not with silence, so it cannot tell you what silence costs.

What your team does with errors

Edmondson (1996) studied eight hospital nursing units. She expected the better-led units to make fewer medication errors. The result went the other way. Units with more coaching and direction from the nurse manager, better working relationships and better perceived performance had higher detected error rates.

Detected errors depend on how many errors are reported as well as on how many happen, and the data could not separate the two. So the study does not show that open units make more or fewer errors. The author's reading was that willingness to report differed between units, with the openness of the leader as one reason.

For the Tuesday, this suggests that how many mistakes a boss hears about says something about how safe it is to bring them. Your CFO's reaction will depend partly on that climate, and you may not control it. These were nursing units, not finance teams, and the study cannot tell you what your CFO will do.

What you say to yourself afterwards

Breines and Chen (2012) ran four experiments on self-compassion and the wish to improve. In the third, 86 undergraduates took a hard antonym test that most of them failed. Some then read a short note saying that failing is common and that they should not be too hard on themselves. Others had a self-esteem condition or nothing. All could then study for a second test for as long as they wished. The self-compassion group studied for about 306 seconds on average, against about 230 for the self-esteem group and 203 for the group with no note. The gap with no note was significant, and the gap with the self-esteem group was marginal. Test scores did not differ.

For the Tuesday night, the reading is that a kind sentence after a mistake did not lower the wish to put it right. In this study it went with more of it. The note did not say that the failure did not matter. It said that failing is common. The samples were mostly undergraduates, and the studies measured motivation and short-term behaviour, not lasting change.

How to tell your boss you made a mistake: three steps

None of these guarantees a good reaction. They aim at three things the research points to: the judgement of the whole self that sends you into hiding, the response that matches the facts, and the night after.

1. Write four facts before you write to anyone. Take ten minutes. Write four lines: what is wrong; where it has gone and who may have used it; what it changes and what it does not; and what you can fix and by when. Use no adjectives and no "I should have". For instance: "Accrual of 340,000 posted to January, belongs in February. In the board pack, page 4, sent yesterday evening. January EBITDA shown 340,000 too low, full year unchanged. Corrected model by 9, revised page by 10." This moves the problem from "I am someone who gets this wrong" to "this figure is wrong." It is an inference from the review, and no study cited here tested it.

2. Tell once, early and plainly. Send the message to the person who is relying on the number, as soon as the four lines exist. Use four parts: what you found, what it affects, what you are doing, and when they will hear from you. For instance: "I found an error in the revenue bridge on page 4 of yesterday's board pack. An accrual of 340,000 was posted to January instead of February, so January EBITDA is shown 340,000 too low. The full year is not affected. That was my error. I have corrected the model and will send a revised page by 10, with a note on how it got through." One sentence of ownership is enough. In the study, the apology that worked took responsibility and said it would not happen again. No study cited here tested how early to tell. If the figure has gone to an auditor, a lender or a regulator, follow your company's rules on escalation. If the conversation you have been dreading is bigger than one figure, Difficult Conversations: A Framework for Honest Communication goes further.

3. Make the night after count. Before you sleep, write one line on what let the error through, as a change in the process and not in your character. For instance: "Second reviewer on the bridge before any pack goes out." Then say to yourself what you would say to a colleague who made the same slip. In the experiment, a note of that kind went with more motivation to improve, not less. It is an inference that this holds at 11 p.m. after a board pack.

Questions People Ask

Should I tell my boss straight away or fix it first?

Do both in one message if you can: say what you found and when the fix will be ready. A fix without a message leaves your boss using the wrong number in the meantime. No study cited here tested timing, so this is a judgement. If someone is about to act on the figure, such as a board, an auditor or a lender, tell sooner.

What if nobody has noticed the mistake?

Then you are the only person who can choose how the story starts. Whether to correct a number depends on who relies on it and what it changes. The study compared an apology with a denial, not with silence, so it cannot say what silence costs. If the figure has reached an auditor, a lender or a regulator, follow your company's escalation rules. This article is not advice on reporting obligations.

How do I apologise without over-apologising?

Keep it to one sentence about the work, not about yourself: "That was my error." Then give the fix and the time. A long apology turns a message about a figure into a message about you, and the review links judgements of the whole self to shame. This is an inference.

What if my boss reacts badly?

Some do. In the study, neither response brought trust back to where it started. The aim of the message is accurate numbers and a relationship that can recover, not a guarantee. If every error is met with blame, that is a conversation about the climate for you and your manager, or for HR if your company has a process. Coaching can help you prepare for it. It cannot change how a team treats errors.

When it is more than one mistake

One mistake in a reporting season is part of the work. If errors keep arriving in weeks of rushed deadlines, ask what sits behind them. How to Push Back on an Unrealistic Deadline in Finance is about the request that comes before the work. If you lie awake replaying one error for days, or you check every file three times because you dread the next slip, this is a pattern, not a figure. Burnout in Finance: Why High Performers Hit the Wall describes what that can look like when it builds. If sleep, mood or health are suffering, talk to your doctor. Coaching is educational support for work patterns. It is not treatment and not a substitute for it.

Sources

• Breines, J. G., & Chen, S. (2012). Self-compassion increases self-improvement motivation. Personality and Social Psychology Bulletin, 38(9), 1133–1143. DOI: 10.1177/0146167212445599

• Edmondson, A. C. (1996). Learning from mistakes is easier said than done: Group and organizational influences on the detection and correction of human error. The Journal of Applied Behavioral Science, 32(1), 5–28. DOI: 10.1177/0021886396321001

• Kim, P. H., Ferrin, D. L., Cooper, C. D., & Dirks, K. T. (2004). Removing the shadow of suspicion: The effects of apology versus denial for repairing competence- versus integrity-based trust violations. Journal of Applied Psychology, 89(1), 104–118. DOI: 10.1037/0021-9010.89.1.104

• Tangney, J. P., Stuewig, J., & Mashek, D. J. (2007). Moral emotions and moral behavior. Annual Review of Psychology, 58, 345–372. DOI: 10.1146/annurev.psych.56.091103.070145