Key Takeaways
Measure what customers do, not merely what your team ships
Work shipped is not value. An outcome is a measurable change in a customer's behavior that creates value. A product, policy, or campaign is an output: it may work as designed and still fail to help anyone. Google Glass was technically impressive, yet customers largely did not buy or wear it. Companies introduced unlimited vacation policies hoping employees would rest more, but employees often took less time off. In both cases, producing the thing did not produce the intended result.
Start with the human response. Ask who you serve, what you want them to do differently, and how much change would matter. A target such as active customers using a service measures value more directly than units manufactured or a launch completed.
The distinction is powerful because organizations naturally count what they control: releases, campaigns, and completed work. Customer behavior is harder to control, but it is closer to the reason the work exists. This reframing also exposes a subtle measurement problem: a behavior can change for reasons unrelated to the team's work. Outcomes are better evidence than output, not automatic proof of causation. Pairing behavior metrics with customer feedback and careful comparisons makes the conclusion more credible.
Treat the colleagues who use your work as customers
Customer distance is not customer absence. At a shoe company, designer Stuart initially thought his customer was the person buying the finished shoe, not his drawings. But brand managers, manufacturers, and marketers used his designs to do their own work. If clearer design materials helped them respond faster, their behavior could show whether his work had improved.
Map the chain of value. An internal team can ask which colleagues consume its work and what those colleagues need to accomplish. A data team may serve marketers who need better evidence; a human resources team may serve employees and managers. The goal is not to call everyone a customer casually. It is to identify who depends on the team's work and whether that work makes them more effective.
This idea makes customer focus relevant to roles far from a sale, and reveals that value often travels through a chain of people. It also gives internal teams a way to test whether a service is useful rather than assuming that delivery equals usefulness. The concept needs a boundary, however: serving an internal requester is not always the same as serving the eventual user. Teams should understand both the immediate recipient and the wider human consequences of the work.
Give each team one focused objective and a few behavioral measures
Separate aspiration from evidence. An Objective describes a specific, time-limited goal and gives the work a reason. It should be inspiring and qualitative, not a disguised task list. Key Results show measurable progress toward that goal. Each should identify the person whose behavior matters, the behavior sought, and the size of change that would count as progress.
Keep the set small. The book recommends one main team objective per quarter, usually with two or three key results. For instance, a support team might aim to provide excellent help, then track whether customers find answers online, reach trained agents, and get problems resolved. The key results should collectively make the objective credible, while leaving the team free to choose solutions.
This structure separates purpose from proof, two things organizations often blur. The objective helps people understand why a goal matters; the measures make progress discussable. Limiting the number of goals is also a practical defense against attention fragmentation. Still, a small set can create tunnel vision if routine responsibilities disappear from view. Keeping essential operating measures visible, while distinguishing them from strategic priorities, preserves focus without pretending the rest of the business has stopped.
Choose the strategic problem before writing the goal
Goals need a reason. Strategy is an opinionated, coherent approach to an important challenge. It requires choosing what problem matters, which people or market to focus on, and how the organization believes it can succeed. Without those choices, OKRs can be neatly written and still direct effort toward the wrong work.
Make the logic visible. A technology startup struggled to convert lengthy customer pilots into paying business. Its leaders settled on faster time to customer value as the central challenge. That choice gave teams a shared direction: product, marketing, sales, and research could each identify a relevant contribution. A useful strategy also says what not to prioritize, and it should be explainable as a clear story rather than a list of disconnected targets.
Strategy acts as a constraint on possibility. A team can generate many plausible goals, but a coherent choice makes it possible to coordinate scarce time and resources. The approach also usefully treats strategy as a working opinion rather than an eternal truth. One limitation is that leaders may mistake conviction for evidence and frame a favored solution as the challenge. A strong strategy names the problem and rationale clearly enough that teams can question its assumptions without losing the shared direction.
Set goals together, combining leadership direction with frontline knowledge
Neither level sees everything. Executives understand broader business pressures, while teams closest to customers often know what is feasible and where friction lies. In one acquisition story, new owners gave a small company a revenue target of $19 million after it had just earned about $3 million. The team had no chance to explain why the demand was unrealistic or what strategic disagreement sat beneath it.
Use two-way goal setting. Leaders first explain strategy and high-level goals. Teams then draft goals within their influence, connect them to those priorities, and discuss them with leaders and peer teams. Revise quickly, publish the result, and clarify dependencies. This creates shared understanding without requiring leaders to prescribe every task or teams to operate in isolation.
This process addresses an information problem as much as a motivation problem. Frontline teams can surface operational reality; leaders can explain constraints invisible to them. The result is more than buy-in, it is a chance to discover that the parties may be pursuing different strategies under the same target. Collaboration is not meaningful if teams can speak but cannot affect the goal. Real autonomy requires leaders to treat disagreement as useful information, not disloyalty.
Treat target numbers as informed estimates, not promises of certainty
A useful target is both plausible and valuable. Teams rarely have enough evidence to calculate the one correct number, especially when trying something new. They can make an informed estimate, discuss what level of change would matter, and refine the target as they learn. A concrete proposal, such as raising average order value by 5%, gives colleagues something specific to evaluate and debate.
Measure the customer, not the system. A supermarket team might target the share of shelf space occupied by store brands. That describes the store, not shoppers. If the real goal is more repeat visits or larger baskets, those customer actions are better measures. The book advises keeping scores secondary and avoiding incentives that reward easy targets or encourage teams to sandbag.
Numerical targets make trade-offs visible, but precision can masquerade as knowledge. A measured target is still a judgment about what matters, what is feasible, and what the team's work can influence. The distinction between system performance and customer action helps prevent proxy metrics from becoming the goal. Even customer behaviors can be poor proxies if they reward short-term activity at the expense of lasting benefit. Reviewing the metric's meaning, not just its value, is essential.
Test small ideas before committing months to a solution
An OKR names the result, not the answer. Teams work backward from a key result, generate several possible solutions, and express uncertain ideas as hypotheses: if the team does something, it expects a particular behavior to change. Then it decides whether to execute or learn first, based on the likely value and the uncertainty involved.
Short tests limit wasted effort. A donut shop seeking more foot traffic could commission a full advertising campaign, or try a person directing passersby to the store for a week. Neither approach is guaranteed, but the smaller test can produce evidence sooner. Teams can learn by talking with customers, observing them, prototyping, and measuring both numerical behavior and the reasons behind it.
This is a practical way to make uncertainty discussable rather than hiding it inside a confident plan. A small test is not merely a cheaper version of a large project; it is designed to answer a question before more resources are committed. The analogy to experimentation has limits: customers and markets are not controlled laboratory conditions, and tests can affect behavior themselves. Teams should therefore treat findings as decision-relevant evidence, not universal proof, and choose the test size according to the risks.
Use regular check-ins to change tactics when evidence disagrees
A goal needs a working rhythm. The OKR Cycle connects strategy and team goals to execution, learning, and progress reviews. Weekly check-ins can surface immediate issues, monthly reviews can examine results and next steps, and quarterly reviews can reconsider priorities. These conversations should focus on evidence, learning, obstacles, and decisions, rather than turning into recitations of completed tasks.
Follow the outcome chain. At FastTax, marketing improved email opens and nearly doubled click-throughs, but customer sign-ups did not rise. A check-in makes that mismatch visible and prompts investigation into whether the work is moving the broader goal. Plans and roadmaps should remain adjustable: keep useful work, stop or defer misaligned work, and add experiments where the current plan leaves a gap.
The FastTax example shows why a metric can improve locally while the intended result remains unchanged. Regular reviews make such divergence visible before teams invest too long in a tactic. They also create a governance question: who can revise work, and how quickly? A calendar of meetings alone does not create agility. Check-ins matter when participants have access to trustworthy data, can raise unwelcome findings, and possess clear authority to change the plan.
Leaders must make experimentation safe and support it with resources
A framework cannot outrun the culture. Teams need access to customer information, time to learn, training, and permission to change direction. They also need psychological safety, the belief that raising concerns, admitting mistakes, and testing ideas will not lead to humiliation or punishment. Otherwise, teams may avoid research or conceal results that challenge a leader's preferred plan.
Managers become enablers. Their work shifts from assigning every task to clarifying purpose, protecting focus, removing obstacles, connecting the team's goals to strategy, and making decisions when progress stalls. Performance reviews should assess contribution, learning, collaboration, and customer understanding, rather than punish a team for missing an uncertain target. Individual OKRs tied to compensation can encourage local competition and defensive goal setting.
The claim that psychological safety matters is especially compelling because experimentation creates visible failure as well as discovery. Leaders shape the incentives around that exposure through their reactions to bad news. A useful test is whether teams can report that an idea failed and still retain the resources and respect needed to learn from it. Safety should not mean low standards: it means separating honest, disciplined experimentation from negligence, while expecting clear communication and responsible follow-through.
Scale OKRs through visible pilots, shared goals, and patient iteration
Do not impose a template everywhere at once. Large organizations can take years to build an effective OKR practice. Start with a supported pilot, make its work and lessons visible, and use the experience to improve training, tools, and processes before expanding. Infobip's attempts illustrate that simply renaming existing production targets or appointing a few champions may not give employees enough understanding to work differently.
Connect goals without dictating every detail. Each team goal should support a higher-level goal, while retaining local ownership. Interdependent teams can share an OKR when their results depend on one another, such as teams responsible for different stages of online checkout. Shared goals can reduce siloed optimization. Scaling is not the same as cascading identical instructions downward; it is maintaining a traceable connection between strategy and work.
The pilot approach treats organizational change as something to learn, not merely announce. Visibility matters because colleagues need to see how teams reached decisions, not just a polished result presented at the end. Shared goals also expose dependencies that a simple hierarchy can conceal. Yet pilots can mislead if they use unusually capable teams or unusually favorable conditions. Their lessons should be tested across varied roles and constraints before being treated as a universal operating model.
Analysis
The book's central argument is that organizations often confuse activity with value. Output is attractive to manage because it is visible, schedulable, and largely within a team's control. Outcomes are more consequential but less convenient: they depend on other people's behavior, which is uncertain and affected by forces outside the team's influence. The proposed OKR practice addresses this gap by joining strategic intent to observable customer response, then using short cycles of work and review to revise assumptions.
Its strongest contribution is not a new goal-writing formula alone. It is the attempt to connect measurement, team autonomy, customer knowledge, and organizational design. A designer's internal colleagues can be customers; a key result should describe what those people do, not merely what a system does; and leadership should create conditions in which teams can discover how to improve that behavior. This makes OKRs a management philosophy as much as a planning instrument.
That breadth also creates tension. Behavioral measures can look objective while still being weak proxies for customer benefit. More clicks, purchases, or faster processing may not always mean a better experience. A team can also be accountable for a behavior it influences only indirectly. The book recognizes scope and uncertainty, but organizations must continually test whether each measure represents meaningful value and whether a team has enough agency to affect it.
The framework is therefore most credible when treated as a learning system, not a promise that a metric will reveal truth by itself. Quantitative data can show what changed; interviews and observation can help explain why. Transparent check-ins can expose mismatches, but only if leaders tolerate evidence that challenges their assumptions. The practical discipline is to make a clear bet, state what evidence would change the bet, and preserve the ability to change course. Done well, OKRs replace the illusion of certainty with coordinated, testable judgment.
Review Summary
Who Does What By How Much? is a highly practical guide to implementing OKRs (Objectives and Key Results) with a customer-centric approach. Readers praise its clarity, accessibility, and actionable advice, making it valuable for both OKR newcomers and experienced practitioners. The book emphasizes aligning teams around customer outcomes rather than just internal metrics, helping organizations break down silos and improve collaboration. Reviewers appreciate the real-world case studies and step-by-step implementation guidance. While one reviewer found it average for experienced users, most consider it essential reading for driving organizational transformation and creating meaningful business results.
People Also Read
Glossary
OKRs
Goals and success measuresObjectives and Key Results form a goal-setting approach. An Objective states a time-limited, qualitative aim, while Key Results measure progress toward it. In this book, key results are framed around changes in customer behavior, rather than completed tasks or delivered products. OKRs also refer to the recurring process and organizational practices used to pursue and review those goals.
Who does what by how much?
A behavior measureA compact way to define an outcome. Who identifies the person or customer whose behavior matters. Does what names the valuable action or change being sought. By how much specifies the measurable target. The phrase helps teams write key results about observable human behavior instead of describing only a product, system, or activity.
Outcome
Valuable behavior changeA measurable change in a customer's behavior that creates value. An outcome describes what a person does differently after encountering a product, service, or other work. In this approach, outcomes serve as key results because they indicate whether the work is helping people, rather than merely confirming that a team completed its planned output.
Output
Work produced or deliveredA product, service, policy, campaign, feature, or other thing that people make or deliver. Outputs matter as possible means to create value, but their completion does not establish that customers benefit. A product can function as designed and still fail to change behavior in the intended way.
Impact
Long-term organizational effectA high-level, long-term measure of success, such as revenue, market share, customer satisfaction, or progress toward a nonprofit mission. Impact can guide organizational strategy, but it is often too broad for a team to influence directly or use as a day-to-day measure. Teams therefore track nearer customer behaviors that may contribute to impact.
Logic Model
A results pathwayA framework for understanding how work is expected to create longer-term effects. It distinguishes outputs, the things produced, from outcomes, changes in people's behavior, and impact, the broader result those changes may support. The sequence helps teams ask whether an output is actually producing the customer response and organizational benefit they intended.
OKR Cycle
A recurring work rhythmThe repeating process through which an organization sets strategy and goals, teams define their own OKRs, work toward results, learn from evidence, and check progress. The cycle includes execution and learning rather than treating goal setting as a one-time event. Regular reviews help teams adjust tactics and, when warranted, revise goals.
Scope of control
Work a team can influenceThe portion of a product, service, customer journey, or organizational process that a team can meaningfully affect. OKRs should reflect this boundary. When success depends on other teams, the groups need to coordinate, use complementary goals, or share an OKR rather than accept responsibility for results they cannot influence.
Business as Usual (BAU)
Ongoing operational responsibilitiesThe necessary day-to-day work that keeps an organization functioning but is not necessarily its current strategic priority. Examples include maintaining website reliability or running a call center. BAU measures should remain visible, but placing every operational responsibility inside an OKR can dilute the focus that OKRs are meant to create.
Parent OKR
A supporting higher-level goalA higher-level OKR that gives a team's goal a clear strategic connection. A child OKR should support its parent, and teams should be able to explain how the relationship works. This creates traceability and shared direction while allowing teams to shape goals suited to their own responsibilities.
Psychological safety
Safety to speak candidlyA shared belief that people can raise ideas, questions, concerns, and mistakes without being humiliated or punished. In an OKR environment, this matters because teams need to report failed experiments and evidence that challenges existing plans. Without such safety, people may hide uncertainty, avoid research, or continue work they believe is not helping.
HIPPO
Highest-paid opinionAn acronym for the highest-paid person's opinion, used to describe decisions dominated by the most senior or highly compensated person in the room. The book contrasts this habit with decisions informed by customer behavior and evidence. OKRs can make an opinion testable by connecting it to an expected result and checking what actually happens.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.