First Impression Error

What first impression error is

First impression error, also called first impression bias, is an appraiser forming an early judgement about a person at the first meeting or in the first weeks of work, after which the judgement settles and the evidence that follows is interpreted through it.

Its mechanism has two steps, and the second is what keeps it in place:

• Anchoring. A judgement forms from a small amount of early information and becomes the starting point.
• Filtering. Later evidence arrives, and whatever agrees with the judgement is recorded while whatever contradicts it is explained away. The contrary evidence does not disappear; it is given an interpretation that leaves the judgement standing.

That is why an appraiser can grow more convinced of the judgement as time passes. The conviction grows not because evidence for it accumulates, but because more of the evidence has been interpreted in its favour.

How first impression error differs from neighbouring appraisal errors

First impression error shares its setting with other appraisal errors and parts from them in its mechanism:

• The recency effect in appraisal. It is the opposite of first impression error in time, not a variant of it. The recency effect drops the beginning of the period and rates on its end; first impression error rates on the beginning and then interprets the end through it. One is forgetting, the other is filtering.
• Contrast error in appraisal. Its reference point is another person who was rated before. The reference point of first impression error is the same person at an earlier time.
• Central tendency error. It is a fault in how much of the scale gets used. First impression error can sit at the top of the scale or at the bottom equally.

First impression error beside the halo effect and implicit bias

First impression error, the halo effect and implicit bias all generalise from a little to a lot. What separates them is what was generalised, and along which axis:

• The halo effect. It is one positive impression of a person carrying over into the rating of all their other qualities. The generalisation runs from one quality to another.
• First impression error. It generalises from one time to another: from the start of the observation to everything after it.
• Implicit bias. It is a tendency the person holding it is not aware of, which shapes their judgement according to the group a person belongs to rather than what that person has done. Its judgement needs no first meeting at all, because it comes before the observation instead of starting from it.

The practical difference between the last two is that first impression error starts from something the appraiser actually saw and then stretched too far, while implicit bias does not start from anything the person did. The same line divides the halo effect from implicit bias: the first is an overextended generalisation from a real observation, the second a judgement that comes before observation because of membership of a group.

Where first impression error occurs

• The interview. A judgement forms in the first minutes, from the greeting, the candidate’s appearance and the fluency of the first answer, and the rest of the interview is then spent confirming it rather than testing it.
• New employee onboarding. The first weeks are when the new hire knows least about the context, so a slow start can be taken as a fixed trait. Those are the weeks that employee onboarding covers.
• A move to a new team. The judgement starts again with the new manager, and it may be built on a previous manager’s impression passed on by word of mouth before the two have even met.
• Promotion. Someone who stood out early in front of management can stay in mind as the default candidate after the facts have changed.

What narrows first impression error

• Criteria written before the first meeting. They cannot be cut to fit an impression that has not yet formed, whereas a criterion written afterwards can be.
• Fixed questions and independent scoring. Our definition of a structured interview states that independent scoring prevents opinions from converging on whichever view is voiced first.
• Recording the score at each question. A score recorded at the end of the interview can be written from the overall judgement rather than from the answer.
• Asking about events that happened. It shifts the question from impressions to facts, which is what behavioral interview questions do.
• Several observers. One observer’s first impression is not automatically shared by others observing from different positions.
• Asking the appraiser outright for the contrary evidence. What have you seen that does not fit your judgement? Filtering works when the contrary evidence is not asked for, and it weakens when it is.

A worked example: how much evidence a first impression error filters

What makes the error costly is not that the judgement is early. It is the share of what comes after the judgement compared with what the judgement was built on. The figures below are assumed.

In an interview. Take a 45 minute interview in which the judgement formed in the first 4 minutes, from the greeting and the fluency of the first answer. The judgement rests on 4 divided by 45 of the meeting, which is about 9 percent, and the 41 minutes after it, about 91 percent, pass through the filter. The interview did not get shorter and none of its questions were cut. What changed is its function: it was a source of evidence and became a place for confirmation.

The same effect shows up in the scores as well as in the minutes. Take a structured interview form with six questions, each scored out of five, for a total of 30. If the judgement closes after the first question, then the 5 points that question carries are the only part of the form completed before the effect began. The remaining 25 of 30, about 83 percent of the form’s weight, are filled in after it. That is why recording the score at each question before moving to the next has a visible arithmetic effect: every question closed with its score takes its weight out of the filter’s reach.

In new employee onboarding. Take a judgement that forms in the first three weeks of a twelve month appraisal period. Counting the period as 52 weeks, that is 3 divided by 52, or about 6 percent of the period, and those are the weeks when the employee knows least about the context and its tools. The share is small, and it falls where the employee is least placed to show their work; the two together leave the early judgement in onboarding on a thinner footing than in the interview, not a firmer one.

In the sources we reviewed, we found no reference setting a length of time before which a judgement counts as early and after which it counts as settled, a number of follow up questions that counts as even, or a weight to give the first impression in a decision, and we set none ourselves. The percentages above show the method; they are not thresholds against which any organisation should be measured. What the example does show is that the share of evidence interpreted through the judgement, compared with what the judgement rests on, can be calculated from the length of the meeting and from the weight of the form, and that whatever is built on that figure is an administrative decision taken inside the organisation.

Signs of first impression error on an interview sheet

Filtering cannot be seen in the final score, but it leaves marks on the sheet itself:

• Scores entered all at once. They are written at the end of the meeting in the same hand and the same ink, not at each question, which leaves every score open to having been written from the overall judgement.
• Shorter notes on the later questions. They can shrink to a single word, because the question was no longer being asked to find out its answer.
• Evidence pointing one way only. A sheet without a single line that runs against its conclusion is a sheet on which nobody asked for the contrary evidence.
• Uneven follow up. An average answer from a candidate judged favourably gets one follow up question, and the same answer from a candidate judged unfavourably gets three. The difference in the number of follow up questions can be counted from the sheet.
• A general reason for rejection. Wording such as “not at the required level”, with no event tied to a particular question, is the form an early judgement takes when it is asked to justify itself.

These are signs about the procedure, not about the person, so they can be reviewed without accusing anyone: the sheets are examined and the questions asked are about the differences, not the intentions.

A borrowed judgement is not a first impression error

An impression can reach the appraiser before they have seen the person at all: a word from a previous manager, a remark from a fellow panel member, or a line in an internal transfer file. Strictly, this is not first impression error, because the appraiser did not form a judgement from something they saw. They received a finished judgement and then began filtering the evidence through it.

The distinction is useful because the remedy differs. First impression error is narrowed by whatever delays the closing of the judgement inside the meeting: recording at each question, and asking for the contrary evidence. A received judgement is narrowed by what happens before the meeting: views are not exchanged before independent scoring, and whatever is passed on from the earlier stage takes the form of recorded events rather than descriptions. One measure does not address both, because one arises in the room and the other walks into the room with the appraiser.

Where first impression error leaves no visible trace

• The rejected candidate. Someone turned down in the first minutes never returns to be compared with anything, so the error leaves no trace inside the organisation that could be measured. That makes the interview the setting where its cost is least visible, not the one where it matters less.
• The decision at the end of probation. A judgement formed in the first weeks, with filtering at work for the rest of the period, ends in a decision after which no new observation is available to correct it.
• The internal transfer. An employee arrives with an impression that went ahead of them, so their appraisal in the new team starts where it ended in the old one, with no independent period of observation.
• The unannounced nomination. Someone who stays in management’s mind as the default candidate does not appear in any minutes, so there is no written decision that can be reviewed.

When an early judgement is not a first impression error

Not every early judgement is an error. A first impression can be correct, and the following months can genuinely confirm it; insisting on discarding it simply because it was early throws away sound evidence. The error is not in the forming of the judgement but in its closing. A judgement that stays open to revision at the first piece of contrary evidence is not first impression error at all. It is an observation made at its proper time.

This is an explanation of the concept, not legal advice.

Related terms

• Recency Effect in Appraisal
• Contrast Error in Appraisal
• Structured Interview
• Behavioral Interview Questions
• Performance Calibration