Case Study: How Google does

Performance Reviews
Everything you need to know about
Google’s performance management
practices

Francisco S. homem de Mello

© 2016 Francisco S. homem de Mello
Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Where did we take all this stuff from? . . . . . . . . . . . 3

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Performance Reviews . . . . . . . . . . . . . . . . . . . . . 5
Self-Evaluation . . . . . . . . . . . . . . . . . . . . . . 6
360-degree Feedback . . . . . . . . . . . . . . . . . . . 6
Calibration . . . . . . . . . . . . . . . . . . . . . . . . . 7
Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . 8

OKRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
TL;DR . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Introduction . . . . . . . . . . . . . . . . . . . . . . . . 13
A Brief History of OKRs . . . . . . . . . . . . . . . . . 14
A bit of goal-setting science . . . . . . . . . . . . . . . 17
Introduction
”We need people to know how they’re doing, and we’ve evolved what
might at first seem like a zanily comples system that shows them
where they stand. Along the way, we learned some startling stuff.
We’re still working on it, as you’ll see, but I feel pretty confident
we’re headed in the right direction. And with any luck I can save
you some of the headaches and missteps we had along the way.”

• Laszlo Bock, SVP, People Operations, Google

Google has probably Silicon Valley’s, and maybe the world’s, most
advanced human resources (or, as they call it, People Operations)
practice. As it becomes clear by books like Work Rules, written by its
SVP of People Operations, and How Google Works, by Eric Schmidt,
its former CEO, and Jonathan Rosenberg, current SVP of Product,
the company continuously iterates on people practices, based on
uniquely huge amounts of data, gathered among its more than 50
thousand “smart creatives”, employees in the fields of engineering,
design, and sales, all handpicked at the world’s top universities.
Google’s people operations cornerstones are:

• Hiring only the best: sourcing, and selecting, only the best fit
candidates amongst the best pool of candidates worldwide,
and, if it can’t reasonably achieve 100% perfection in hiring
only amazing fits, skewing errors towards false negatives
(eventually passing on a great candidate) instead of false
positives (eventually hiring a bad fit);
• Creating a meritocratic environment, where the best perfor-
mances are correctly identified and rewarded; and,

1
Introduction 2

• Developing employees to their full potential, through great

people management and on-the-job coaching (see our Project
Oxygen paper here), peer-to-peer and outside training, and
through a comprehensive 360-degree feedback collection pro-
cess;

The purpose of this case is to explain a bit better, and in detail,

how Google does the second bullet point, meritocracy, and bits
of the third, development, through its performance management
procedures.
But why, you may ask, look at Google for a benchmark? Apart
from the obvious reasons (it seems to be working for them, eh?),
it our view that smaller, less resourceful companies (and here
we’re talking more about sheer cash and headcount, as opposed
to attitude), can greatly benefit from using Google as a starting
point for their own practices, and then iterating on that (what is
commonly known as standing on the shoulders of giants). Why
should you, HR manager, or C-level executive, reinvent the wheel
when this giant company has not only spent millions and millions of
dollars finding its best self, but talked at length about it, so that you
can benchmark yourself and use many of these practices? There’re
no good reasons not to[1].
We’ll have achieved our goals if you find some inspiration and best
practices on this paper. Remember we, at Qulture.Rocks, can help
you get your performance management practices (be them inspired
by Google or not) running in a matter of hours.
Where did we take all this
stuff from?
Great question! Google, as many of the world’s top companies (like
GE, AB InBev, Walmart, etc), talks frequently and openly about its
own culture. So we basically read everything that’s out there – and
actually written by Google and its current and former employees
-. We’ve also scanned online platforms like Quora, Medium, and
Twitter, read many pieces by the press, and, finally, interviewed as
many as 10 former executives, both from People Operations and
from areas like Search and Google Ventures, in order to form a
holistic understanding of Google’s practices.
We’re confident that you have a very faithful description in your
hands. Some of the processes may have already been iterated out
activity, but for the most part, we strongly believe you have an
accurate picture of the company’s current practices.

3
Overview
For the purposes of this case study, we’re calling performance
management the collection of the following human resources tools
and processes used at Google:

• Annual performance review (including mid-year checkpoint)

• Monthly performance check-ins (part of regular 1:1 meetings
that also comprise other themes such as career development,
coaching, personal issues, etc.
• Googlegeist engagement survey (that spans much more than
just the regular engagement axes, but measures basically
everything that’s to be measured)
• Annual Upward Feedback Survey, a feedback review (similar
to 360-degree review) where only supervisors are reviewed
by their direct reports, and that is based on Google’s Project
Oxygen
• OKRs, or objectives and key-results, a mildly different form
of Management-by-Objectives, that we explain in this post,
and
• Meritocracy, or compensating people unequally, based on
their perceived performance, through bonuses, equity stock-
option grants, and prizes Performance reviews

4
Performance Reviews
Google’s annual performance review cycle is comprised of two
parts: a “preview”, in the end of the first semester, and a complete
review, that happens between October and November, and which
happens concurrently with the company’s 360-degree feedback
collection process.
Managers take two main things into account when attributing
their employees’ performance ratings: results attained, or what
the employee accomplished, and behaviors, or how the employee
attained these results. The employee starts with a self-assessment,
which is followed by peer-reviews, whose authors are only visible
to managers (reviewees may have access to the anonymized content
of peer reviews).
On the review side, Google employees are asked to review each
other, and their direct reports, according to the following criteria:

• Googleyness: The employee’s adherence to Google’s values.

This is the main component of the “how” axis.
• Problem solving: Analytical skills applied to work situations
(problem solving).
• Execution (high quality work with little guidance): Deliver-
ing great work without the need for a lot of hand-holding
from managers and peers (autonomy).
• Thought leadership: How much an employee is seen as a
reference for a given niche of expertise. As Google grows in
size, these niches may tend to become smaller and smaller,
but still, Google wants employees that are go-to resources
for specific themes, training colleagues on tech-talks, traning
customers, and producing high-quality content.

5
Performance Reviews 6

• Leadership (or emerging leadership): Albeit many young

Googlers have little or no exposure to managing complex
teams, everybody is required, nonetheless, to show emerging
leadership skills, such as taking the lead of problems and
projects, being pro-active, and owning results personally.
• Presence: Presence is the employee’s ability to make himself
heard in an increasingly large organization, and intimately
related to emerging leadership.

Self-Evaluation

The self-evaluation is the first step in the performance review, and

where the employee evaluates himself in the five criteria described
above (on five grades ranging from “never demonstrates” all the
way to “always demonstrates” and invited to share examples of
his actions that support these grades), and highlights his main
accomplishments for the last cycle (in a text field limited to 512
characters). These accomplishments will appear in the next step
(360-degree reviews) to reviewing peers, who’ll be then asked
to assess their proximity with these projects, and the reviewee’s
impact on their results.

360-degree Feedback

Google’s 360-degree review process serves the purpose of giving

managers a holistic picture of their direct reports, since they may
carry a biased and restricted impression of reports’ impact and
behavior (some employees may be great at “managing up” a rosy
picture of their contributions, for example).
The process starts with a back-and-forth between employee and
manager, so as to pick a representative, fair sample of peers to
participate. The employee suggests a shortlist, that is discussed and
Performance Reviews 7

validated with the manager, taking into account how close the peer
was to the employee’s contributions, and how well she can assess
the employee’s performance.
Peers are expected to give assessments in three different media:
strengths, or things that the person should keep on doing, and
weaknesses, or things that the person should consider working
on/developing; rating each other on the five criteria discussed
above; and finally, commenting on the reviewee’s contribution
to specific projects. These two open-ended fields (positives and
negatives) have evolved from a larger form a few years ago. Laszlo
Bock, Google’s SVP, People Operations, observes in his Work Rules
that the simplification reduced aggregate time spent on this step
by more than 25%, while improving the share of participants who
perceived it as useful from 49% to 75%.

Calibration

After all data has been collected, in the form of self-reviews and
peer-reviews (or what’s known as 360-degree feedback), and results
achieved are understood, managers draft a rating for their employ-
ees, based on the following scale[2]:

• Needs improvement
• Consistently meets expectations
• Exceeds expectations
• Strongly exceeds expectations
• Superb

As you may have noticed, I said they draft their ratings. That’s
because no ratings are final before the calibration process, again,
described by Laszlo Bock:
“The soul of performance assessment is calibration… A manager
assigns a draft rating to an employee – say, ‘exceeds expectations’-
Performance Reviews 8

based on mainly OKRs but tempered by other activities, like the

volume of interviews completed, or extenuating circumstances such
as a shift in the economy that might have affected ad revenues.
Before his draft rating becomes final, groups of managers sit down
together and review all of their employees’ draft ratings together in
a process we call calibration… A group of five to ten managers meet
and project on a wall their fifty to a Thousand employees, discuss
individuals, and agree on a fair rating. This allows us to remove
the pressure managers may feel from employees to inflate ratings.
It also ensures that the end results reflect a shared expectation
of performance, since managers often have different expectations
for their people and interpret performance standards in their own
idiosyncratic manner… Calibration diminishes bias by forcing man-
agers to justify their decisions too ne another. It also increases
perceptions of fairness among employees.”
Calibration, a process also adopted at other leading companies such
as AB InBev, GE, Kraft Heinz, and Goldman Sachs, is therefore of
crucial importance in ensuring the fairness of performance ratings.
It’s where heavy-handed raters are identified and discounted for
(and the opposite is also true).

Outputs

The calibration meetings output each and every employee’s perfor-

mance rating for the period. After the rating is closed, managers go
on to hold two meetings: one where feedback is given, taking into
account peer reviews and managers’ impressions of their employ-
ees, and another where compensation and promotion decisions are
communicated.
The two conversations are held in different meetings and at least a
month apart from each other in order to ensure their quality. Google
understands that a compensation-focused employee is no good a
listener of feedback, whether compensation expectations were not
Performance Reviews 9

met, met, or exceeded:

“a [negative] dynamics exists when managers sit down to give
employees their anual review and salary increase. The employees
focus on the extrinsic reward – a raise, higher rating – and learning
shuts down…. We have an embarassingly simple solution. Never
have the [pay and feedback] conversations at the same time. Annual
reviews happen in November, and pay discussions happen a month
later.”
The theme is also discussed by Prasad Setty, member of Google’s
People & Innovation Lab[3]:
“Traditional performance management systems make a big mistake.
They combine two things that should be completely separate: perfor-
mance evaluation and people development. Evaluation is necessary
to distribute finite resources, like salary increases or bônus dollars.
Development is just as necessary for so people grow and improve.”
Performance Reviews 10

Diagram 1: Google’s simplified performance management schedule

[1] Actually there is one good reason not to: You’re in a business
where the majority of your employees are not “smart creatives,”
but maybe less educated, operational, hourly workers, maybe not as
capable of self-management, and maybe not as high on Maslow’s
pyramid. Valid argument, but we won’t discuss it in detail here.
Enough to say that you’ll have much more to gain from learning
with Google than ignoring it, for now.
[2] Before a five-point scale, Google rated its employees on a scale
from 1 to 5 in 0,1 increments, having, in fact, 40+ possible ratings.
The scale, according to Laszlo Bock, beared many inneficiencies, as
was ditched after more than 10 years in use for a simpler scale.
[3] Google’s People & Innovation Lab, or PiLab, is worth a book
itself. In short, it’s a team of quants whose only attribution is
to study people data (performance, engagement, happiness, etc),
Performance Reviews 11

iterate on people practices (testing them), and to continuously

support Google’s people practices with heavyweight data analytics.
OKRs
*This chapter is an excerpt from our ebook, The Ultimate Guide to
OKRs, which you can find and download at http://qulture.rocks/the-
ultimate-guide-to-okrs

TL;DR

OKRs are an acronym for Objectives and Key Results. Objectives

are high-level, qualitative goals. Key-Results are specific, SMART
goals that support the Objective. When we say support, we mean
Key-Results should include metrics that trully translate Objective
accomplishment.
Some pundits use a very simple statement: We will achieve _____-
___ as measured by ____, ____, and ________. The first space is
filled by your Objective, and the second to fourth are filled by Key-
Results. Let’s use an example to illustrate our definition:
Objective

• Increase the profitability of the company

(Since OKRs belong to cycles, if they don’t have a “date” stamp

to them, you should automatically assume the goals should be
completed before the end of the cycle.)
Key Results

• Increase revenues by 10%

• Reduce costs by 3%

12
OKRs 13

• Maintain general, and administrative expenses nominally

constant

As you can see, the Objective is a bold goal, specific, time bound,
but still achievable (as opposed to doubling my profit in a month).
Key results are the actual targets, as measured by KPIs, that will
really acid test Objective achievement. They are what, in MBOs,
we know by “goals”. As you can see, I’ve used Andy Grove’s
method of Key-Results breakdown: using KRs as milestones for goal
achievement. There are other possible approaches. Some people
defend that Objectives have to be qualitative, whereas Key-Results
have to be quantitative.

Introduction

If we come to think about it, there’s a blurred line that separates a

job description, a goal, and standard operating procedures.
Job descriptions would be fine by themselves if we substituted
humans with robots, and robots were always pre-programmed with
job descriptions and a manual of standard operating procedures:
they would perform their jobs and responsibilities perfectly, capped
only by the machines mechanical/processing limitations. But it
turns out there are many tasks we haven’t figured how to program
robots to do better than us, yet. We humans are very good at
having common sense, and transferring knowledge from one set
of applications to another; at figuring out weird patterns; and at
coming up with novel solutions to problems.
With this set of amazing skills comes also some limitations at
performing and sustaining performance at our fullest potential, two
problems that are closely related to our - unique? - faculty: volition.
That could happen because a lack of challenge (related to our frontal
cortex) or a lack of purpose - the “why” of what we’re doing it. If
OKRs 14

we’re supplied only with job descriptions, we may forget them, or

lose our motivation to perform it at our best. We may even lose
track of what “best” means.
Goals come in to fill this need. They help us find a sense of
purpose when we properly participate in their setting, but also by
linking our responsibilities at tasks’ outcomes to greater outcomes,
like the success of the group we work in/with; they provide us
with a challenge to constantly perform at our best; they provide
gratification when we reach them, which reinforces the cycle.
OKRs are basically goals: an old staple of business management,
rebranded, repurposed, and tweaked to 21st century necessities of
companies and professionals.

A Brief History of OKRs

OKRs are an old staple of business management, rebranded, repur-

posed, and tweaked to 21st century necessities of companies and
professionals.
It all started with the fathers of management, Taylor, Ford, and
the sorts, who began facing business like a science. How so?
They figured they could measure outcomes, and then formulate
hypotheses as to how they could improve these outcomes. The main
outcomes back then were productivity, as measured by output per
employee. These guys figured out optimum work schedules, break
times, and lightning arrangements for factories. They also started
streamlining production, adding specialization to the factory floor.
These practices all brought incredible, tangible improvements.
In the 50s, a fellow named Peter Drucker, who’s believed to be the
greatest management guru that ever lived, figured out that adding
goals to managers could be a great thing. Not only they had to
improve their outcomes, as Taylor and Ford had, but they also had
to aim at specific target outcomes from time to time. Drucker called
OKRs 15

this framework Management by Objectives, or MBO, a concept

introduced in his seminal book The Practice of Management.
Since the introduction of MBOs, practically every modern Fortune
500 company practices some sort of goal setting. It’s proven to bring
better results than not having goals . Some companies set them once
a year; some companies set them twice a year. A number of them tie
variable compensation to reaching your goals, and another number
of them perform some sort of performance review based on these
goals and results. The term OKRs was introduced by Andy Grove,
a de facto cofounder of Intel (he joined the company on the day of
its incorporation, but is not listed as a cofounder,) and its former
CEO, in his great management book High Output Management .
Grove didn’t bring any transformational insight to MBOs, but spoke
about appending key-results to goals, and calling goals objectives.
But making key-results an integral part of the MBO process is very
important: it brings clarity to how goals can and should be attained,
and makes this “how” evident and transparent to everybody.
In Grove’s view, key-results had to be chronological milestones that
took professionals in the direction of reaching their goals: a one-
year goal could be broken down into 12 monthly key-results, or 4
quarterly key-results. He treated them specifically as “milestones.”
That use was attuned to Intel’s 80s reality: a large company, already
on the top of its game, looking to translate its strategic planning into
actionable goals and milestones for the whole organization.
Another Grove tweak to MBOs was his belief that goals (objectives)
and key-results had to be set in a bottom-up process, from the em-
ployee up, so as to bring buy-in and empowerment to the process.
Before that, companies would shove goals down the organization,
from the Board to the CEO, down to VPs, and so on. Grove enabled
employees to set their goals according to a broad guidance from the
company, to be then calibrated with direct managers.
Last, but not least, Grove insisted that OKRs be aggressive, meaning
– very – hard to achieve, what he called “stretched.” He went further
OKRs 16

along, and instituted 70% as the new 100%, meaning that achieving
70% of your goals was as good as hitting them, since they were
purposely baked very hard .
In the late 90s, OKRs spread out to other Silicon Valley companies,
through the inspiration of Jon Doerr, a partner of Kleiner Perkins
Caufield Byers, one of the world’s foremost venture capital firms.
Doerr had worked for Intel under Grove’s leadership, and got
acquainted with its use of OKRs, later thinking it could be adapted
to other companies of KP’s portfolio. That’s how Google, and later
on Zynga, became fierce advocates of OKRs, tweaking the tool to
their specific needs.
But what made Google’s version of OKRs different from Intel’s?
Not much. Google shortened – a lot – the OKR cycle, making it a
quarterly process. That means the company, its senior executives,
and basically every employee, sets his or hers objectives and corre-
sponding key-results quarterly, a practice more attuned to the in-
credibly fast-paced reality of web 2.0 technology companies. Google
enforced Grove’s position that goals should not be cascaded down
the organization in a top-down manner, and greatly expanded upon
it, according to Laszlo Bock, its SVP of People Operations:
“Having goals improves performance. Spending hours cascading
goals up and down the organization, however, does not. It takes
way too much time and it’s too hard to make sure all the goals line
up. We have a market-based approach, where over time our goals
converge, because the top OKRs are known and everyone else’s OKRs
are visible. Teams that are grossly out of alignment stand out, and
the few major initiatives that touch everyone are easy enough to
manage directly.”
That means at Google, everyone’s OKRs are set by themselves, and
made public via its intranet. Google ensures that individual OKRs
are aligned with its own through a mix of supervisor oversight, peer
pressure, and psychology.
OKRs 17

A bit of goal-setting science

(Or, the goal of having goals.)

Goal-setting has historically been used in the corporate world for
two main purposes:

• To motivate employees (efficiency)

• To assess their performance

Let me explain this: HR common sense has always said that goals
motivate employees towards achieving better results. Goal achieve-
ment, on the other hand, has historically been used as a proxy for
performance: if I’ve hit 100% of my goals, it must mean I’m a good
performer. But we thought it made sense to briefly review goal-
setting theory, or GST. We think HR professionals deserve to have
this widespread practice correctly understood with a theoretical
basis, because there is more to it than just these two axes of purpose.
According to GST, ˆfoo3¹ goals serve three main purposes:

Focus

Presuming that goals have been come up with according to the

company’s long, medium, and short-term strategies, according to
myriad methodologies like BSC, Hoshin Kanri, etc, goals help the
company focus effort, attention, and energy on what’s relevant,
relative to what’s not relevant. According to Johnson, Chang, and
Lord (2006), “goals direct individuals’ attention to goal-relevant
activities and away from goal-irrelevant activities.” It is proven that
“individuals cognitively and behaviorally pay more attention to a
task that is associated with a goal than to a task that is not.”
¹MarionEberly,DongLiu,TerenceMitchell,ThomasLee;
AttributionsandEmotionsasMediatorsand/orModeratorsintheGoal-StrivingProcess
OKRs 18

Effort

Another very important purpose of goals is to increase the level of

effort that people exert at work. It is also proven that “goals energise
and generate effort toward goal accomplishment. The higher the
goal, the more the effort exerted.” This is a tricky equation: too
hard a goal, and, as you’ll see in a bit, employees get demotivated;
too easy a goal, and employees will also get demotivated. In sum,
there’s a right amount of hard, which pushes people to challenge
themselves, but within a reasonable chance of achievement, that
optimizes performance, which links us to

Persistence

Persistence is probably the trickiest thing to get right when setting

goals: The right ones produce high effort input for longer periods
of time, but the wrong ones can really wreak havoc: “large negative
discrepancies may lead to a withdrawal of effort when individuals
are discouraged and perceive low likelihood of future goal attain-
ment.” (Carver & Scheier, 1998). As we’ll see, there are derivative
factors that influence persistence towards goals.