Article
A Faster Way to Remove Bias from Performance Management
April 1, 2020


By Marc Effron, Talent Strategy Group

It’s currently fashionable to sell companies individual bias training to help their leaders improve performance management. While the intentions of these efforts are admirable, the research is clear that anti-bias training rarely has a lasting effect. Science shows that it doesn’t provoke any long-term change in racial bias and in other cases it creates a backlash against the very people it was designed to support (1,2).

The cause of that failure is obvious. As professors Frank Dobbin and Alexandra Kalev wrote in their landmark Harvard Business Review article, “Why Diversity Programs Fail”: “It turns out that while people are easily taught to respond correctly to a questionnaire about bias, they soon forget the right answers (3).”

It’s incredibly difficult to change the deeply-held beliefs and practiced behaviors of adults. There is a strong, natural resistance to admitting bias and it requires prolonged, sustained effort to change biases even with individuals who are highly committed to change them. It’s unlikely that most people want to be biased but also unlikely that they’ll both recognize and stop showing biases because of a training course.

Behavioral economics helps explain why a process can steer us in the right direction when our good intentions aren’t getting us there. For example, most Americans know that it’s important to save for their retirement, but only 57% of eligible employees contribute to a taxfree 401(k) savings plan in the United States. Similarly, nearly every smoker is aware that smoking tobacco is dangerous to oneself and others, yet quit rates remain low despite easily available smoking cessation programs and drugs. The wrong choice in each of those areas will leave you poor and sick, but that’s not enough motivation to convince many individuals to change their behaviors.

In those same situations, people’s behaviors changed when a process steered them toward the right choice. Where there was a system to automatically enroll employees to contribute to a 401(k) plan, 92% remained as contributors (4). When a process forced people into a smoking cessation program, 80% quit smoking or radically reduced their consumption. People still had free choice in both of those situations but a structured process helped strongly guide participants to the right choice.

Use Process to Achieve the Good Intentions

A great process is superior to good intentions and we should use that fact as we try to reduce bias in performance management.

Process controls help reduce bias by using: 1) the power of groups, 2) sound analytics to guide the actions of individual managers, and 3) simple, consistent processes to reduce variation.

These process-based techniques work because:

  • Diverse groups make superior decisions to homogeneous groups or single individuals. The more we subject individual decision to group evaluation, the more we reduce bias.
  • Analytics can give us nearly real-time data to search for and identify bias across the performance management process so we can correct it.
  • Consistent processes ensure a clear standard for managers and employees so it is easy for us to measure quality and completion.

Here’s how you can reduce bias at each stage of the performance management cycle:

Reduce Bias In Goal Setting

Poorly set goals sow the seeds for bias throughout the performance management cycle. Goals of uneven quality or with vague metrics make it more challenging to coach for performance and to fairly evaluate performance. A few different ways to reduce bias in the goal-setting process include:

Goal calibration: A goal calibration process checks to see whether goals within a function or group are set at a relatively consistent level of challenge. This reduces bias by ensuring that no one on a team has goals that are meaningfully more or less challenging than other team members.

In a goal calibration session, each team member reads their goals (no PowerPoint deck, each team member simply reads their 3 to 5 goals). The other team members listen to those goals to identify any collaboration that’s possible or overlap that needs to be corrected. Everyone, especially the group’s manager, listens to hear if the goals are all at a relatively similar level of challenge and if they can all be clearly measured. Too many qualitative metrics allows space for rating bias to creep in later.

The team member’s goals are adjusted by the manager if the challenge is too much or too little or if the metric is too vague.

2-Level Up review: A fast, simple and powerful way to reduce bias in goal setting is a “blind” goal review by the manager two levels above the person who sets the goal. In this process, the 2-Level Up manager gets a list of every goal set by their direct reports’ team members. The employee’s name isn’t listed with their goals. If your managers are setting goals properly (3-4 per person), this list will contain 100 – 150 goals. That may seem like an arduous task for any manager to review but the process is amazingly fast.

Since the 2-Level Up manager understands their goals for their group they can quickly read each goal and flag any individual item that either isn’t aligned to their goals, doesn’t seem sufficiently stretchy or has a vague metric. They only mark goals that need correction. After their review, they send their comments back to their direct reports and ask for the flagged goals to be improved.

The blind review reduces potential bias or favoritism that might arise if the manager directly associated a goal with a particular individual.

Reduce Bias in Coaching

It may seem challenging to believe that bias can occur in the performance coaching process, but research suggests that women receive less specific feedback than men do during performance coaching (5). There are two process controls that can help to identify bias in the coaching process and reduce it.

A simple, structured, consistent coaching process across the organization: When there’s variance in how managers execute a process, it allows bias to creep in. Since most organizations have no structured process for on-going coaching and feedback conversations, there’s significant variance in the quality and content of those conversations.

You can help your mangers reduce bias in feedback and coaching by implementing the 2+2 Coaching approach. The 2+2 Coaching process radically simplifies and standardizes feedback by having every manager structure every feedback conversation in an identical way. Managers are asked to make 2 observations against the employee’s current goals (not two comments for each goal) and two “feedforward (6)” comments for how that person can improve their performance in the future.

That standard 2+2 process will help reduce the bias that can creep in when managers follow their own coaching script, but the bias-reducing power comes from measuring the quality of that conversation using the One Question Survey.

The One-Question Survey with bias review: Each employee will receive a One Question Survey (you can use your HRIS system, web- based survey tools, etc.) that asks: Have you had a quality coaching conversation with specific feedback in the last 90 days? The answer choices are Yes or No. The one-question survey will be sent once a year and results confidentially tracked by employee.

You can help to identify and correct bias by evaluating the responses by gender and race (and other categories if valuable) to see if patterns emerge either by individual manager, group or department or companywide. HR is responsible to work individually with managers to reduce any bias they showed in the coaching process.

Reduce Bias in Ratings

The most effective way to reduce bias in the rating and review process is to execute well on the previous two steps. If clear goals and metrics have been set and calibrated, and 2+2 coaching has occurred and its effectiveness assessed, more employees can be reviewed against clear, consistent standards.

Those steps are wonderful pre-cursors to less biased reviews, but there are still ways to control bias in the review process itself.

Eliminate self-reviews that influence ratings: Many companies correctly start the self-review process long after managers have submitted employee ratings. Why? Because there’s no chance for self-reviews to influence a manager’s review decision or rating.

If self-reviews are considered by managers when they rate employees, there are tremendous opportunities for multiple types of bias to influence the assessment. Employees who write more eloquently may create a more compelling case for a higher rating. Employees whose first-language is English may use diction or word choices that appeal more to a native English-speaking manager. Less honest employees may try to shade the truth about metrics, contributions or accomplishments. Employees who write more or less than a manager wants to read may be considered to be trying too hard or not trying hard enough.

The performance review is a manager’s opportunity to assess how an employee performed against their goals and behaviors. Don’t let an employee’s persuasive abilities inject bias into that decision.

Use rating calibration session with diversity statistics: Many organizations already apply this technique but there’s an opportunity to enhance its effectiveness. In a ratings calibration session, managers in a functional group or geography meet to review the performance ratings and/or performance bonuses they each plan to give to their team members.

In a round-table discussion, anyone receiving the highest rating or lowest rating has to be discussed and approved by the entire group. Next, the group reviews statistics showing the distribution of ratings and bonus recommendations by key groups. Those groups would include the typical categories of race and gender, but could also be presented by native-English (or home-country language) speaker, short vs. long-tenure, potential rating and other cuts that could tease out bias that might not otherwise be evident. A common bias that occurs is what we call “hierarchy bias” where the average performance rating increases as one moves up levels at a company.

After reviewing the data, the group would decide what changes to make to their rating and bonus recommendations.

The recommendations above assume that you have performance ratings. We believe that ratings are essential to managing performance and even more so to manage bias.

Rely on Process, Not Hope

We’re confident that 99.9% of leaders want their people decisions to be bias free. The current trend to train managers on bias does no harm but science suggests that the good intentions it creates rarely translate into changed behaviors. If your company is serious about reducing bias in the performance management cycle, you need to rely on the powerful processes at your disposal to stamp out bias faster.


1. Lai, Calvin K., Maddalena Marini, Steven A. Lehr, Carlo Cerruti, Jiyun-Elizabeth L. Shin, Jennifer A. Joy-Gaba, Arnold K. Ho et al. “Reducing implicit racial preferences: I. A comparative investigation of 17 interventions.” Journal of Experimental Psychology: General 143, no. 4 (2014): 1765.

2. Kulik, Carol & B. Pepper, Molly & Roberson, Loriann & Parker, Sharon. (2007). The rich get richer: Predicting participation in voluntary diversity training. Journal of Organizational Behavior. 28. 753 – 769. 10.1002/job.444.

3. Dobbin, Frank, and Alexandra Kalev. “Why diversity programs fail.” Harvard Business Review 94, no. 7 (2016): 14.

4. Vanguard Group, How America Saves 2018,\\Download 2/20/19 from https://institutional.vanguard.com/VGApp/iip/site/institutional/
researchcommentary/article/HowAmericaSaves2018.

5. Correll, Shelley, and Caroline Simard. “Vague feedback is holding women back.” Harvard Business Review (2016).

6. Goldsmith, Marshall. “Try feedforward instead of feedback.” Journal for Quality and Participation (2003): 38-40.


Stay Up-to-Date

TSG stays at the forefront of a fast-paced industry by constantly creating and curating important Talent Strategy resources. Keep up with the latest by subscribing to our newsletter!