Applying OKRs

Over the last year or two I have been exploring OKRs—Objectives and Key Results—with several organisations, from a few hundred people in size to a couple of thousand. Some are well over a year in, some are just starting out. There doesn’t seem to be much out there in terms of experience reports or hands-on advice so I have tried to capture the advice I wish I’d had when I started out.

As with any tool, OKRs are going to be misunderstood, misapplied, and just plain abused, but when they are framed properly and applied sensibly I’ve seen them bring order and direction to organisations that lacked both.

A brief introduction to OKRs

The story starts with Management by Objectives, or MBO. Management by Objectives sucks, or more accurately MBO doesn’t work. You and your manager come up with a bunch of objectives at the beginning of a review period, typically six months or a year. At the end of the period you are measured against those objectives, which form the backdrop to your journey through the organisation.

This leads to one of two scenarios: you are ambitious and set stretch objectives, fail to meet them, and are punished, or you are conservative and set modest objectives that you easily achieve. So the organisation sets its good people up for failure, and rewards mediocrity. In any case the objectives you set are often out of date within a month because things move on, and your interests and opportunities, and the organisation’s goals, will move with them.

Andy Grove, founder and CEO of Intel, recognised these problems and devised a deceptively simple model he called OKRs, or Objectives and Key Results. His colleague John Doerr then introduced them at Google, and since then lots of companies, including some high profile tech firms, have been adopting them.

A typical enterprise defines its “true north” on several levels:

  • Its Mission is (mostly) unchanging, and is the reason it exists: to index the world’s information; to land a man on the moon and bring him safely back to the Earth; to inspire and nurture the human spirit – one person, one cup and one neighborhood at a time. That last one is from Starbucks!
  • Its Vision is a statement of how it plans to realise that mission, and is usually stable for a number of years, or until a major regime change in the organisation.
  • The Strategy is most likely an annual conversation: “What are we going after this year? Geographical expansion? Product diversity? Product consolidation? Reducing our operating footprint?”
  • Then Objectives are the quarterly milestones by which you navigate the strategy, and Key Results are how you measure progress towards those objectives. So OKRs are about applying and measuring the strategy throughout the year.

Defining OKRs

Every three months the exec team defines Objectives for the organisation. These are the three or four things—not more than that because they won’t fit in your head—that we are going after this quarter. They need to be broad enough that they capture the whole organisation, but specific enough that you can see how you could contribute to them.

For each Objective we identify three or four Key Results, which are measurable indicators of success. Objectives are qualitative; Key Results are quantitative. I like to use the language of “outrageous success,” so we are genuinely unlikely to achieve a key result. This is important. They are aspirations rather than realistic goals.

Finally each department, team and individual asks themselves: what can we do in the next quarter to contribute to those key results? This process is supposed to cascade neatly down the organisation, but as you’ll see below, in practice this isn’t as obvious as it sounds.

Getting started with OKRs

I started by watching the Google Ventures talk all the way through, which is 1h20 long and pretty dry but it has two advantages. It describes the OKR process in detail along with a worked example, and it references John Doerr, whose own talk is 2h30, which I figure leaves me net one hour.

I then described OKRs as I understood them to a couple of clients, with the full disclosure that I had never tried them but I had a good feeling about them. They decided to give them a try so I facilitated a couple of workshops where we cooked up some OKRs and tested them by imagining it was 3 months hence and we’d had a great quarter. We also asked ourselves what things could fall through the cracks without negatively impacting the OKRs, and whether that mattered.

We are already entering the phase of people selling OKR tooling, but I’m clinging to the fact Google just uses their internal directory and a Google doc to hang the OKRs off, and that I’ve so far managed to get teams writing half-decent OKRs just using regular facilitation tools like Post-Its and index cards.

In terms of pitfalls and gotchas, I always err on the side of more disclosure is better, so I tell my clients they will mess them up and to roll with it and see what happens. If they freak out about that I suggest that maybe we’re not ready yet for experiments like this, and to do some other things in the meantime.

The rules

This is what I’ve learned coaching and applying OKRs in a number of different contexts. The sample size is still modest, but the organisations vary in size, domain, age, and business model, and there are definitely lessons to learn.

There are only a few rules with OKRs, but if you scrimp on any of them the whole thing will most likely fail at best, and backfire badly at worst.

1. They are not a target, they are an aspiration.

Management By Objectives rewards mediocrity as people set easy, modest objectives and score big by landing them. The Key Results should be outrageous in their ambition. I coach them as “Imagine it’s three months time and you’ve just had the best quarter you could possibly have had. What did you achieve that amazed you?” One quote I like is “If you aim for the moon and miss, at least you had a great view!”

A good quarter will land 70–80% across most of your KRs. If you are getting more than this they aren’t ambitious enough.

2. You never measure people on their OKRs. Never.

As soon as you do this they become targets (see 1) and people will start to game them. They are an alignment tool not a measurement tool. In organisations that do annual or semi-annual reviews, you can use your individual and team OKRs since your last review to remind yourself of the journey, and to pull out what you and your team(s) achieved as part of the narrative. But rather than saying “We aimed for 4% increase in market share but only achieved 2.5%,” it is more like “We achieved 2.5% increase in market share over a quarter, which is fantastic! And I had forgotten about it because of all the stuff that has happened since!”

3. OKRs cascade down from as high as you can start them.

In one organisation I have the CEO and his board setting them quarterly; in another the IT leadership; in another the Head of Engineering and his team across about half the IT organisation. In each case, each Outcome (I prefer this language to Objective) is something the organisation can rally around, and the KRs are big ambitions for each outcome. It is not unusual for the Os to roll on for several quarters, identifying different KRs in each quarter.

4. “Cascading” is not as obvious as you think.

In any decent-sized organisation you have lots of managers and management groups—think practice or guild leads, department managers, middle and senior line management functions—and you have control functions like PMO or Change groups, compliance, security and legal teams, and so on. It turns out cascading isn’t so obvious for them.

To help them identify meaningful OKRs I start with my working definition that the goal of an organisational system is to optimise for rapid and sustainable flow of value (a blog post for another time, work with me here), and then to think of what that means for this organisation for this quarter.

Some OKRs will be about value in the form of measurable, customer impact. Others will be about improving flow by reducing lead time, reducing batch size, eliminating obstacles or changing the org structure to make it easier to get work done. Others will be about the sustainable elements such as happiness of people, managing costs, reducing fragmentation across different autonomous groups, or whatever else we feel might make sustainability harder if we don’t go after it.

In the organisations I work in, a person could be a member of various groups with different responsibilities, so a manager might be part of a working group looking at improving flow in the Path to Production, they might be the executive sponsor of a product delivering customer value, and may also have a formal compliance role. Their individual OKRs would express each of these, and describe how, if they have an amazing quarter, they would contribute to the success of the various groups they interact with, which roll up to drive the organisation’s goals for the quarter.

5. So if they aren’t a target, what is the point?

The reason for having individual and team OKRs is to create a sense of alignment at various levels of the organisation, and importantly to provide people with a vehicle to say No, or at least Not Yet. This is surprisingly difficult in most organisations, especially if the ask is yet another of the thousand tiny cuts that sound fine on their own but which together will derail everything you are trying to achieve. “I would love to but that doesn’t align with our OKRs. Maybe next quarter.” becomes a legitimate pushback.

You are only ever pushing things out by at most 12 weeks, which is a blink in corporate time, but seems to be just enough for the request to (usually) dissipate naturally. If it is still there it becomes a candidate for the next quarter.

I’m seeing some real behavioural change from senior managers and execs due to OKRs, as well as an increase in focus in the organisations I’m coaching, but it is taking a lot of “No, you still can’t use them to measure people” conversations.

6. You will mess it up at least twice

OKRs are hard. They are proper strategy rather than the usual hand-wavy stuff. They are a deceptively simple model but there are a number of failure modes I’ve been seeing:

6a. Conflating the department head with their department.

Department heads will declare the things they personally want to go after as department-level OKRs, which can make them too tactical, e.g. “Fix Project X” rather than “Ensure all major projects are on track”. This can leave other teams that aren’t working on Project X thinking “How does that apply to me?”

Conversely their individual OKRs might say something like “Decommission 3 legacy systems.” I counter this by asking “Are you personally going to do that in the next 12 weeks?”

6b. Thinking about them as MBO-style objectives.

People start treating them as personal goals they will be measured against, and hedge accordingly. Cue the usual hand-wringing and gaming of traditional MBOs. Or they will move heaven and earth to achieve their personal KRs at the expense of broader organisational goals.

Fortunately both of these behaviours are reasonably easy to spot, although they can be challenging to manage. This can be a presenting symptom of the culture of the organisation so you may need to just wait it out until the behaviour self-corrects.

6c. Getting stuck in the weeds.

Honestly, it’s only for 12 weeks. Even if we mess this up badly (which we will), what is the worst that could happen in a quarter? People can definitely overthink OKRs and it is important to coach them past that.

7. They only happen quarterly!

So even if you pick things up quickly, and you work in an organisation that picks things up quickly, it will still take at least three or four quarters before you start to reap the real rewards. One of my clients started a little over a year ago, and they recently told me OKRs have “transformed how we do strategy”, “created alignment between Engineering, Marketing and other departments that we didn’t think was possible”, and “caused peace to break out in the management team.”

Concluding thoughts

I have become a fan of OKRs over the last year or so. As with any planning activity, the value is in the conversations around producing the OKRs as much as in the measures themselves.

I’ve watched teams up and down the organisation from middle management to execs, wrestling with the hard choices of what not to do this quarter, and OKRs seem to provide just enough structure to cause the team to come together and agree a shared direction.

I recommend giving them a go. And remember, you will mess them up at least twice, so go easy on yourselves.

16 comments

  1. Hi Dan. Interesting write up on OKRs, thank you. Do you have any examples of OKRs used specifically for software engineers?

    1. dalbuschat · ·

      I was going to ask the same, would love to see some real (fake) examples.

      1. As you can imagine, even fake real OKRs can be revealing so I’m not really able to share even doctored OKRs. However someone recommended Radical Focus, which starts with a fictitious worked example of OKRs and then explains the process. I’ve bought it but haven’t read it yet, but the reviews are good.

      2. You can find a bunch of OKR examples here: http://okrexamples.co/

      3. Thanks for posting these. I looked at your engineering examples and for me they fail the “outrageous” test (and the full-screen ad is annoying too). They say things like “Improve our testing procedures” or “Review and improve our security procedures.”

        One of my criteria for a good Objective is that it make me want to jump out of bed and rush in to work. I like an Objective like:

        Slash the time needed to release our products!

        which might have KRs like:

        • Halve the time to release three applications. (measurable and hard to do!)
        • Eliminate 1/3 of the manual effort involved in releasing our core product (may be impossible but we will learn a lot finding out why)
        • Eliminate rework (number of issues submitted down to zero) in preparing our staging environment. (May need to ensure this doesn’t add lead time through extra checks.)

        These kinds of KRs drive the right behaviour and imply things like value-stream mapping the release process, without having to spell it out.

  2. Jan Helwich · ·

    Hi. Good read. My personal experience is that especially traditional management tries to cascade literally. Meaning each KR is used as O in the next lower level. Is that what you had in mind in the “cascading” paragraph?

    1. Hi Jan, that’s a good question. How the OKRs cascade is a function of your organisational structure. Each sub-organisation, whether it be a department, a programme, a delivery team, a management team, a community of practice, or anything else, looks at the OKRs of the “next level up” to see how they can contribute. Individuals are usually members of multiple organisational groupings so their personal OKRs should reflect this.

      OKRs do cascade in the sense that a project team within a programme can’t define their OKRs until the programme has defined its own, but the relationship between the OKRs themselves isn’t strictly hierarchical.

      You start with the organisation agreeing its overall OKRs, typically at the exec level. Then the Engineering leadership, for instance, thinks about how their department can contribute to one or more of those key results for the quarter. Each KR doesn’t necessarily drive a single objective below, and each objective and key result further down can drive more than one key result further up.

      The only rule as such is that a lower-level key result must drive at least one key result further up the organisation. You can’t introduce new things further down otherwise you don’t get the alignment.

      As an example, say the Engineering department had one objective about reducing its operating cost and another about being more confident about programme deadlines. A programme within the department could set an objective to identify and release a cost-saving subset of its scope early. This would be a good thing for both those higher level objectives, and the programme team would most likely choose key results that are easy to relate to the KRs above.

      At the same time, cross-functional groups like compliance or finance, or structures like communities of practice, would be thinking about how they can best support the various value streams in the organisation, and setting their own OKRs accordingly.

  3. Hi Dan. Nice post. I recognise the ‘mess it up’ part. People find it difficult to set good objectives an measurable key results. But it starts the discussion, which is the main benifit of the OKR tool in my opinion. But what is your opinion/experience on scoring the key results after each quarter has ended?

    1. Hi Bart. It is important to track progress throughout the quarter as well as reviewing at the end. KRs are stretch goals, so if you are comfortably landing them all the chances are you are “sandbagging”—setting your sights too low. Conversely setting them too aggressively is demotivating because you know you don’t stand a chance.

      If you find some of the KRs end up at only 20%, say, it either means it was much harder than you thought, or that you weren’t that interested in progressing it. Both of these are useful feedback. As long as you stick to the OKRs for the whole quarter and critically review them at the end of the period you should be ok. My clients tend to reviews progress against their OKRs in a regular scheduled meeting every one or two weeks, so they can track progress and identify obstacles.

      The goal of OKRs is to set direction and keep you focused. Measuring people against Key Results defeats the purpose, because then you end up back in a world of Manaagememt By Objectives.

    2. What we’ve found by working with thousands of customers on OKRs at Weekdone, it’s that if the key results are set well and there is automatic 0-100% objective fulfillment calculation in the software, there is no need for additional grading. OKRs should be objective not subjective in saying how far you are and how much the fulfillment is, so there is no need to turn measurable numbers into subjective grades later on. That would create more trouble than solve them.

  4. Thank you for an insightful article Dan. One of the better ones I’ve read. Disclaimer: I am a founder of a Swedish OKR tool.

    We actually found a useful variation to the standard approach of Company-, Team- and Individual OKRs. It solves the issue with boss’s KR becoming my Os automatically and it also creates clarity between what is actually the individual’s focus and department’s focus (sum of individual’s focuses). In short – it keeps OKRs as an aspiration method rather than to be one of targets and a copy of the Balanced Scorecard (which I dread just as you dread MBO). So as you understand, your points in this article are very interesting to me and the way we have designed an updated OKRs method, and I would be happy to hear your thoughts on it.

    So how do we do it? – We throw out Company- and Team OKRs and replace this with something we call ”Focus Areas”. They are as well set quarterly, just as the OKRs.

    The main difference with focus areas are that they are not measurable, instead they serve as an area of focus that the company has set up for the individual to align their individual aspiration towards (OKR). The focus areas are also nested so it can have several levels, like ”Product”/”Product Efficiency”/”Agile Development”, where a developer or designer could link their objective to any of those three levels for example. The reason why the focus areas shouldn’t be measurable is that if you let measurements take place on both team/company level AND individual level, the individual measurement will always be second hand, at least in the case where the two are not perfectly matching. And in reality – they never are perfectly matching. Making the individual’s measurement second hand will kill engagement for the individual, and will, in turn, kill the overall engagement around OKRs for the whole company.

    So keep OKRs on the individual level, and let Focus Areas be the connecting source all the way up to the Company’s Mission, Strategy and Long Term Plan.

  5. Hi Dan,
    Great article. As the key results are quantitative how do you get around individuals OKRs becoming a measurement of that individual? I understand the definition around communicating the positive progress made rather than the part that wasn’t achieved but I still think it will feel like a measurement to the individual.
    Have any organisations you’ve worked with steered away from individual OKRs and decided to keep it at the sub-organisation level? This would maintain alignment as long as you have functional and committed sub-organisations. Alternately does this need to flow down to the individual in order to build that commitment?

    1. Important to look at the measurement as a measurement of the Key Result, and not of the individual. The individual is the captain and has chosen a vehicle – the Key Result. The measurement is an objective measurement of how far the vehicle has been driven, not how the driver has driven it – if that makes sense.

      Let me try to present an opposing view – and hopefully a bit more hopeful view on the individual perspective – OKRs is foremost an engagement tool and measurement of innovation (aspiration). Innovation doesn’t happen anywhere else than with the individual and OKRs are one of the best methods to stimulate this engagement. But a lot of OKRs initiatives fail when they include the Company- and Team OKRs into the mix with individual OKRs. In fact – I believe that team-OKRs are directly harmful to the individual engagement, and that you should throw them out and replace them with Focus Areas or something similar. You will never achieve maximum individual engagement with team-OKRs in play as they will indirectly describe my individual OKR for me, which, over time, disconnects me from my individual engagement towards the company.

      So, let the individual OKRs be the champion and align them as a roll up towards strategy.

  6. Hi Dan, great high-level view of OKRs! I wrote a guide to help teams get started implementing these, and would love your feedback. http://discover.usenotion.com/team-goal-setting-and-team-metrics

  7. brunns · ·

    I do have concerns about the 1/4ly horizon of OKRs. Activities which will take a long time to pay off can be discouraged. Also, when we learn something, and have the opportunity to respond to it quickly, benefitting the organisation, we may find that also doesn’t affect an existing key result, and be discouraged from doing that too.

    1. Those are both valid concerns. For the longer-term goals I’ve seen the same Objective sustained over several quarters, with the KRs underneath it changing each quarter to reflect what we’ve done and where we are going.

      If you learn something within a quarter that means you should legitimately question the current OKRs, then that is what you should do. The quarterly cadence is there to challenge this, and to reduce the likelihood of thrashing or changing direction on a whim. It gives you a structure to say Not Yet: let’s factor that assumption or learning into our next OKRs, which are only a few weeks away, and see how it affects them.

      You may be spotting good ideas all the time, but the opportunity cost of acting on them may be silently hurting your strategic goals because no-one has time to do the important stuff—they are too busy delivering the tactical good intentions. Perhaps being discouraged is a valid counter; at the very least it may cause you to think critically about the amount of time and effort your “quick” opportunity will cost.

%d bloggers like this: