Employee Peer NPS: The Golden Metric For Employee Performance Review

Two decade ago, as an experiment, I started to include the Net Promoter Score (NPS) question in the survey for the annual performance review of my organization. What surprised me was that I was getting the same ranking of the employees whether I was using this simple metric or the standard but complex process including lengthy surveys, hours of verbatim reviews, and days of calibration meetings.

Over the following years, I have continued to use this metric, and I have been investigating why this metric was measuring so accurately the performance and potential of employees. This article presents you with the key findings as well as a step-by-step guide on how to get started with eNPS. If you want, you can directly jump to the end of this post to see the checklist on how to get started immediately.

The employee Net Promoter Score (eNPS) question is “How likely would you recommend working with [employee name] to a friend or a colleague”. The eNPS measurement is based on a survey sent to the employee’s internal network. This network can be, for example, the top 20 most active collaborators (peers, partners, managers, team members, etc.) based on emails and meetings. This collection of metadata can enable much more insights on collaboration and workplace optimization, but this is not the topic of this post.

eNPS is an extension of the Net Promoter Score (NPS) question “How likely would you recommend [brand name] to a friend or a colleague” which is used to measure customers’ loyalty to a brand. In my point of view, this eNPS question measures trust when it is used in a paid environment. When people are paid to work together, the likelihood to recommend working with someone is linked more to a sentiment of “trust” rather than a sentiment of “loyalty” to that person. I firmly believe that this is the primary reason the eNPS score can rank employees so well. Indeed, trust is a composition of the key skills in an organization: Credibility, Reliability, Intimacy, and Unselfishness (see the work by Charles H. Green). You will find more details on trust in my next post which is coming soon.

Net Promoter Score (NPS)

The eNPS is an extension of the Net Promoter Score (NPS) where a company surveys its customers with the question “How likely would you recommend [brand name] to a friend or a colleague” with a scale from 0 = “Not Likely at All” to 10 = “Very Likely”.

Zoom image will be displayed

NPS is believed to measure customer loyalty (see Fred Reichheld and Rob Markey). This loyalty score is calculated as the net proportion of Promoters minus the proportion of Detractors. This framework defines Promoters as people who answered 9 or 10, Passives as people who answered 7 or 8, and Detractors as people who answered 0 to 6.

This is more complicated than using a simple average from all the answers, but as we will see it provides two significant benefits.

First, the segmentation of the customer is not centered around 5, the median of the range. I like this approach as it tends to compensate for the fact that people rate too nicely when asked this question. It also makes this metric tougher and hence it forces us to improve even more to make any significant changes.

Second, the net aspect of the equation (% promoters — % detractors) forces the company to concentrate on two different segments of the customers. By doing so, not only do you need to improve the basics, but you also need to deliver a delightful experience. For example, if the company is a restaurant, not only do you need to keep your kitchen clean (the basics), but you also need to surprise and delight the customers for your restaurant (delightful experience).

Zoom image will be displayed

The NPS can range from -100% where all customers are Detractors (answers 0 to 6), to +100% where all customers are Promoters (answers 9 or 10). A positive score is okay. A score greater than +60% is already excellent.

NPS is a great metric to give to the company’s business units, as it forces them to reduce detractors (answers 0–6) and improve promoters (answers 9–10) which requires two different approaches.

To reduce detractors, a company needs to fix the basics and assure that the minimum expected product/service is offered. For example, a company should fix any broken web pages, or ensure that the customer receives his order. When a company has such severe dysfunction, customers immediately move to the competition (if available).

To increase promoters, a company needs to go above and beyond customer expectations. It needs to delight and surprise. For example, when you purchase a product on eBay or Amazon, and the product does not correspond to your expectations, the process of returning the product is extremely simple and fast. The first time I had to return a product (a large color laser printer) because the color cartridge exploded during transport, I was anxious that it would cost a lot of money and time. Yet, I found out that all I had to do was to fill a quick form on the web, print a label, and drop the shipment at the closest UPS store. I was delighted with the simplicity. What started as a bad experience, turned out to be a great learning and improved my loyalty to that company.

Implementing The Employee Net Promoter Score (eNPS)

Eight years ago I started applying this eNPS question in my team. I sent out a survey with this question followed by the full list of the team members to all our internal partners/clients. Next to each team member was a rating from 0 = “Not Likely at All” to 10 = “Very Likely”. I also added a column for “Not Applicable” in case the respondents did not know enough about the team member, as well as for improving the user experience of the survey. The same survey was also sent to all the team members so they could rate each other. The time to complete such a survey was quite short (2–3 min on average) as the primary element of the survey was this eNPS question in the form of a matrix (one row per employee) and filling this matrix was quite rapid.

What I learned from this experience was stunning, Here are some key findings.

1. Analogous Ranking

The most intriguing finding was that the ranking I got from a simple and quick survey was very analogous to the final ranking from a lengthy process of survey, 1:1 discussions, calibration meetings, and top-down curve adjustments. After much research and discussion, I now firmly believe that this is due to eNPS measuring trust, and that trust is a composition of the four critical skills for any employee’s success and promotability (Credibility, Reliability, Intimacy, and Unselfishness).

2. Distance Matter

A key trend I noticed was that the less a partner collaborates with one of my employees, the more likely he will rate this employee as Passive (7–8), i.e. the partner won’t have a strong opinion (promoter or detractor) about the employee.

By the mathematical aspect of the NPS score, this effect pushes the score range towards the zero score. Indeed, the larger is the share of Passives, the lower will be the share of Promoters and Detractors. Then the subtraction of two small numbers results in a small number. For example, with 10 people where 8 are Passives (who answered 7 or 8), the NPS score can only range from -20% to 20%. Hence the less likely that employee would be on the top (or the bottom) of the pack.

In the annual review meeting, I usually included other statistics. For example, it is interesting to understand the share of people who answered the eNPS question for a particular employee. A low score would mean that very few people know enough the employee to give a rating. This would be bad as this is a sign that the work of that employee got unnoticed and most likely was not impactful.

3. Improve eNPS by Decreasing Detractors

To improve eNPS, an employee needs to change two different aspects of his work. This is very similar to how a company needs to change to improve NPS.

First, the employee needs to decrease Detractors. This can be done by removing the terrible experience that people had working with him. The four pillars of trust (Credibility, Reliability, Intimacy, and Unselfishness) is an excellent framework to help an employee improve.

For example, when an employee does not deliver on his promise (low credibility) or he delivers bad product/service (low reliability), this creates a dysfunctional environment and, over time, the people working with that employee will become detractors. In the same way, when an employee has a low E.Q. (low intimacy) and does not perceive the subtility of social behavior, he will not see the personal issue connected to the surface issue. This creates another dysfunctional setting where root causes are not addressed, and surface issues keep popping up. Last but not least, employees who are self-centered (low unselfishness), will work badly with other teams and create a very dysfunctional organization.

4. Improve eNPS by Increasing Promoters

The fan base can only be built by providing a great collaboration, service, and results. In a very similar way as a company can build loyal customers.

This is done by delivering superior work and services. The motto here is “underpromise and overdeliver” — yet, the “underpromise” should still be credible and challenging.

To deliver on this tactic, the employee needs to find ways to improve significantly his work’s efficiency. He needs to understand which tasks are the big rocks that will move the company, and which tasks are the “hamster works” which will be time-consuming and yield negligible impact.

Beyond this prioritization process, the employee needs also to find ways to create a sense of urgency to deliver the work in a fraction of the initial time: automate, find 3rd party solution, browse the company knowledge base for existing solutions, or disrupt the current process by providing an outside the box solution. For example, in one of my previous companies, the daily sales report required the inputs from half a dozen teams using conflicting data sources. We sat down with all the teams involved, agreed on a common definition and define the source of truth for the data. We used Google App Script to automate the process, and we were able to provide a very robust report through email to all key stakeholders at a fraction of the time needed in the past. All other official reports were using the same underlying data cube and hence all numbers were identical. This project has been praised across the company and has significantly changed the way other teams and functions perceived our organization.

To increase his fan base, the employee will also need to excel in all four pillars of trust:

• be convincing in your words/proposals (credibility)
• consistently deliver good quality & performance (reliability)
• make others feel comfortable talking about sensitive matters (intimacy)
• show that you care about the greater good (unselfishness).

5. Identifying Collaboration Issues Within The Team

As eNPS is measured within the team, you can have a clear matrix view of the internal collaboration. You can quickly identify a pair of employees who mutually dislike each other by giving each other a low eNPS rating. Those types of rivalries can create harm down the roads when the two employees need to work with each other, or when one gets a promotion and not the other one. The manager needs to work on team building to see if the differences can be worked out — or if a more drastic solution needs to apply with one of the two individuals leaving the company.

Working Around Biases

As in all measurements, there are biases. At the minimum, we need to be aware of them. At best, we can compensate for the biases by advanced methods. That is when scientific measurements become an art. Here is a nonexhaustive list of potential issues that might bias your measurement.

1. Subjective List Of Respondents

Depending on how you define the list of people who will answer the eNPS survey for your team/company, you need to make sure the list is fair. If the list is based on the employee, he might miss some people with whom he had issues or with whom he worked at the beginning of the year. Indeed, for an annual review, we need to make sure the full year is being considered. If the list is based on the manager, he might miss some important junior-level employees, and assign indirect contact who did not work enough with the individual.

One way that we found to overcome this issue is to use an application to scan through the last 12 months emails and meeting invites of the employee.

Here is an example of the outcome of scanning company emails & meetings. This list represents the top 16 people I work the most closely while I was at Motorola Mobility. Next to each name are the network weight in points, number of emails, number of meetings and the total duration of all meetings

Zoom image will be displayed

This needs to be executed through a transparent process that respects the privacy laws and the company’s culture. Indeed, while trying to measure the trust between employees, you don’t want to break that same trust between the employees and the leaders of the company. Once this set of metadata is collected, you can use a weighting scheme where the software measures the network weight between each individual.

Here is an example of Google API calls to scan Gmail and Google Calendar to score user connection and establish an objective network.

For example, we attributed a weight of 1 point for each 30 minutes meeting, and that 1 point was then divided by the number of participants. So, if the employee was meeting 1:1 with his manager for 30 min, a weight of 1 was assigned to this connection (employee-manager). For a one hour team meeting of 11 people, each pair is assigned a weight of 0.2 (i.e. 2 points / 10 other people). For email, we assigned the weight of 0.1 points per email, which was then divided by the number of other participants. We decided to use a 1 to 10 factor between emails and meetings, as, in my experience, you would need about 10 emails to get the same work done as a 30 min face-to-face meeting.

Of course, the weighting scheme can be improved significantly, but the results were very impressive for all the people who used this pilot. One of the primary challenges is to capture all the Instant Messaging and offline discussions.

2. Language And Cultural Differences

When you compare the eNPS and NPS results from people living in different countries, you need to be very aware of two main issues.

First, no translation is perfect. Words have different meanings and connotations in various languages. So, the distribution of the answers will be different when you ask the eNPS question in different languages.

Second, there will be cultural differences when scoring someone or a brand. For example, the Europeans will tend to be harsher, and the average rating will be lower than the one from the United States. If the effect is severe in your organization, you can try to correct for this using a calibration factor per market.

For example, let’s take a market A where all NPS answers from the company average to 7.0 and a market B where the average is 5.0. To correct for this, you can increase by 40% (7/5 = 1.4) all answers from market B’s respondents, and then define the grouping into Promoters, Passives, and Detractors. You could also make the adjustment from the other side, and redefine the threshold 0–6 (Detractors), 7–8 (Passives), 9–10 (Promoters). For example, if you see that in your domestic market, the Promoters represent 20% of the respondents' population, but they only represent 5% in another market, you can extend the definition of Promoters for that market, so they represent also roughly 20%.

3. Playing the System

Another potential issue is when some individual wants to play the system. A small group of people could agree to rate the maximum score between them. Depending on the culture of your company, you might have a serious issue or not. In my experience, the individuals who are willing to cheat are noticed by other activities and don’t last long in a company.

4. “Everybody Is Great” Syndrome

You will find on occasion some respondents who will rate everybody at 10. Although this is the ultimate goal to have a perfect functioning company, it is highly unlikely that a group of people rates 10. In my experience, the people who do that are usually underperformers who are worried about their performance. Hence, they tend to play the system unilaterally by rating everybody with the maximum rating, in the hope that they will receive in return a good score. This usually does not happen, as those underperformers usually have a strongly negative eNPS score (below zero, i.e. more detractors than promoters).

5. State Of Mind Of The Respondents

The measurement of eNPS is very sensitive to the state of mind of the respondents. Depending on your survey introduction, or the questions before the eNPS question, you will impact the score. I had seen the NPS of a company significantly improved when we changed the order of the questions.

For example, if you force the respondents to remember a good experience or a bad experience before asking your NPS question, you will impact their state of mind. They will then bias their rating based on what they just remembered. Usually, we suggest placing the NPS question early in the survey to prevent any large biases from the other questions.

6. Statistical Errors

As in any measurement, there is volatility. Hence, when comparing people (which we do in a ranking), understanding the level of volatility is important. eNPS is a mathematically interesting system where you have three states: Promoters, Passives, and Detractors. It is similar to other binomial systems (Male vs. Female, Yes vs. No), but with three states. It is called a trinomial distribution function. The standard deviation, i.e. the error bar for the 68% confidence interval is given by

where N is the total number of answers, p the proportion of Promoters and d the proportion of Detractors.

If you have a small group of respondents or the respondents vary a lot amongst your team/organization, you might want to include this error in the ranking value. To achieve this, you can inspire yourself from the Value At Risk (VaR) method which takes the mean measurement and subtracts two standard deviations. However, I would recommend adding instead of subtracting in the spirit of giving the benefit of the doubt to the employee with respect to this statistical volatility. I would also recommend adding only one standard deviation, to prevent having employees with very low response rate to outrank everybody. Here is a numerical example. If employee A has an eNPS of +60% and the standard deviation is 10%, we know for sure that there is two chances out of three that the real score is between +50% and +70%. Now, if employee B has an eNPS of +50%, but the standard deviation is 30%, the two chances out of three range is +20% to +80%. So, employee A ranking value is +70% whereas employee B ranking value is +80%, and employee B will rank better than employee A.

Let’s Get Started

Now that we have covered the key concepts, here is a checklist to start this process within your company.

1. Start with a small pilot (a team, a function, a business unit, …)
2. Define the list of respondents for the eNPS survey. If you start with a small pilot, the list of respondents will ideally be the same for all employees on which you want to measure eNPS.
3. Create an eNPS survey. The most basic form is a single question “How likely would you recommend working with the following people to a friend or a colleague?”, followed by a matrix where each row is an employee to be scored, and each column is a rating from 0 = “Not Likely at All” to 10 = “Very Likely”. You could add a 12th column with “Not Applicable” for the case when the respondent doesn’t know well enough the employee. You could have a more advanced survey with open questions to collect verbatims or follow-up questions. For the technology, I have been using Survey Monkey, Google Spreadsheet and an internal website that I build which was scanning the emails & calendars from the participants to define their objective key partners, as well as providing the eNPS survey itself.
4. Communicate the project to the key participants and make sure you show a trustworthy process with clear transparency on what this process is about, why do we need it, and how it will be executed.
5. Run the survey and make sure to send reminders to maximize the response rate.
6. Calculate the eNPS score for each employees, including a summary of any verbatims and augment the score with extra statistics: the tier where they rank within the rest of the team/organization (e.g. 1st quartile), the response rate for that employee, the number of Promoters and Detractors, a benchmark across the team to help understand the relative ranking within the team/organization, and key guidelines on how to improve eNPS and the employee perceived trustworthiness.
7. Collect feedback from the respondents and from the employees on the process
8. Improve the process, and iterate with a larger audience

If you are using an automated scanning system (similar to the one I build based on Google App Engine and Google API to scan the emails and meeting invites to determine the objective list of key partners), you could run this process at a much faster frequency. We were looking at a pilot with a two weeks frequency. In this pilot, every two weeks, all the participants would have been rating the top 20 people who they worked the most closely over the last 2 weeks. The idea behind such frequent measurement is to enable the employee to see more frequently his eNPS score over time, instead of once per year. This would have enabled the employee to identify issues and try various tactics to see which ones would have lead to a significant improvement over time.

What do you think? Did you try this in the past? How did it work or not for you? Please let us know!