Data Science OKR Examples

Data Scientists are great candidates for successful OKR projects because of their extensive knowledge about data. Browse the list of OKR examples and learn more about the OKR framework.

Improve our payment fraud detection algorithm

Key results:

  • Increase the AUC of our fraud detection algorithm by 20%

Improve readability of database metadata

Key results:

  • Increase number of database columns with descriptions by 50%

Improve data science reporting setup

Key results:

  • Decrease number of reports done ad-hoc from 10 to 2

Improve data cleansing processes

Key results:

  • Decrease % of rows unsuable for data model x by 20%

Achieve lightning-fast data model calculation

Key results:

  • Decrease dashboard data calculation time delay from 1 day to 3 hours

Improve data model accuracy

Key results:

  • Increase data model accuracy by 20%

Optimize new user activation

Key results:

  • Increase percentage of new users who are considered "activated" from 30% to 45%

What Do Data Scientists Do?

The job description of a Data Scientist is very rarely similar across different job descriptions or job posts, like many other titles in tech. However, what all can agree on, is that a Data Scientist spends most of the day working with data.

The work is often related to creating advanced statistical analyses, pulling and gathering relevant data, as well as setting up dashboards visualizing business-critical metrics.

The job title has been one of the fastest-growing not so many years ago, and it remains very critical, especially for technology-powered companies that create a lot of data and wish to gain value from it.

A few examples of large companies that employ Data Scientists:

  • Uber

  • Airbnb

  • Tesla

  • Facebook

  • Google

Like with many other roles, the bigger the company, the more specific tasks a Data Scientist is assigned. In smaller companies, the tasks may vary more, but the amount of data is typically smaller.

Why Data Scientists Should Work With OKRs

Data scientists can benefit from using OKRs because:

  1. OKRs should always be based on data (which Data Science people love)

  2. The framework focuses on the outcome instead of the output (which helps set the direction for Data Scientists)

  3. It helps keep value-adding work the top priority

What’s Important For Data Scientist OKRs

To start working with OKRs, it’s important to sit down and discuss the following two key aspects:

  1. What’s the primary goal of a Data Scientist?

  2. When is data science work considered a success?

Data Scientist: The Primary Goal To Achieve

One of the benefits of the OKR framework is that it helps teams focus. Focus on what’s important and zoom in on, optimally, a single goal for each OKR cycle. Before defining the OKRs, it’s a great idea to discuss the scenario where the team could only achieve one thing for the next OKR cycle. What would be the primary goal?

For Data Scientists, this could often be related to providing the necessary insights for other teams in the company. It could also be related to delivering growth in the business based on experimentation, which is why Data Scientists are often part of a smaller growth team. If you’re curious about growth-related OKRs, see examples here.

When Is A Data Scientist A Success?

Related to the primary goal of a Data Scientist, it’s important to define when Data Scientist efforts are considered a success. This will help clear any doubt about whether some project added value or not, as it will be obvious, due to the nature of OKRs if it was achieved or not.

As just discussed, this role is often part of a team, so it’s relevant to discuss when that team is considered a success. Let's say that a Data Scientist was part of a growth team in a SaaS company. In that context, success should be defined at the team level, and not at the individual level.

What Makes A Great Data Scientist OKR

As you can see from the examples above, there are plenty of ways to create and formulate OKRs. But there are key requirements that OKRs should pass.

Why Data Scientists’ Key Results Should Not Be Binary

If you google “key result examples”, a lot of times you’ll stumble across many sites suggesting that you could create key results like “Run experiment X”. Don’t! Why? What if that experiment is run and no improvement or insight is found?

The purpose of Key Results is to enable you and your team to track progress on them, objectively. You should be able to pull in a person from the street, give them access to your data and have them answer how you’re doing.

But if your Key Result is binary, you’ll only be able to answer the progress questions once you’ve finished the OKR cycle. That’s a problem and it’s one that lots of teams who work with OKR face.

A great exercise that will enable you to get rid of binary key results is to ask yourself “Why do we want to do X?”. Maybe your team has observed a problem and plans on doing something to fix it. If you do, what state will that put you in? What will the desired outcome be? That should be your key result.

How Often Should I Track Key Result Progress?

You should track progress on your key results as often as you expect there to be changes in progress. Imagine that I have a key result saying “Decrease the query time-related KPI dashboards by 50%”. My guess is that I would be able to optimize the query at least a little bit each week. So it makes sense for me to track it every week.

On the other hand, if I only expect there to be noticeable progress on a monthly basis, reporting on it every week makes little sense. I should instead use the time to make sure that the initiatives I plan for are adding true value to my progress.

How Ambitious Should Data Science Key Results Be?

In short, key results should have a probability of being achieved of 50%. But why 50%? One of the requirements for key results is that they should be very ambitious. 

The reasoning here is that people very often overestimate what they can do in a day, but underestimate what they can achieve in a month or quarter. And so, setting ambitious targets often lead to extraordinary results. This is one of the reasons why the OKR framework is so popular.

In addition, when scoring your key results, it’s usually told that achieving 70% of your target goal is enough to call the goal “achieved”. This is again one of the reasons why key results shouldn’t be binary as you’ll then only succeed or fail. Nothin in-between.

How Many Key Results Per Data Scientist Objective?

You should create as few key results as possible to achieve your objective. The overarching benefit of OKRs is laser focus. Creating too many key results will spread focus and you’ll likely have too many initiatives not being related to each other. Also, improving on one key result should ideally not affect progress on another. If so, then they’re likely too dependent on the same things and should instead be merged into one.

We usually say that unless your objective is covering a larger team, sticking to 1 or 2 key results should be enough. Remember, the key here is to be very focused on moving in the same direction.

Large corporations often have more key results, but that’s usually also because each department or team is responsible for one each. This makes excellent sense, but if you’re creating them for your own small team or yourself, stick to as few as possible.

How Many Key Results Per Data Scientist?

If you’re setting key results on an employee level, you should aim to stick to as few as possible. The reason is that employees are likely also working on team or company OKRs and their focus should be as tight as possible.

Prioritize Outcomes Over Outputs

Working with Data Science, it’s easy to get caught up in focusing your day-to-day work on the outputs of what you do. Doing X gets us Y. You shouldn’t do that.

OKRs are meant to force you to think about your desired outcome. The happy place where you’ll be once you’ve done the necessary work to get there. Think: “If we succeed with this, which things will have changed and to what?”. Another way to frame it is to finish this sentence: “We are now at a state where we are/have …”

Then activate the people in your team and democratize how you’ll get there. The best way to turn output-focused goals into outcome-focused goals is to ask “why?”. “Why do we want the output you’re describing?”.

The Ingredient That Will Get You Over The Finish Line

OKR initiatives are day-to-day tasks that you do in order to achieve your OKR objectives. The results are measured by your Key Results and will ultimately determine whether your initiatives were the right choice or not.

Initiatives are super important because they make sure that everyone involved is working towards the same goal; achieving the objective. Whenever you start planning your OKR cycle, make sure that you immediately start brainstorming on initiatives as well. This often gives people a great idea about whether you're stretching too far or if you're being ambitious, but also realistic.

How Data Scientist OKRs Should Be Structured

When you create your OKRs, they should include these elements at a minimum:

  1. Objective

  2. Key results

  3. Owner

Before the cycle, it’s important to have these three defined, as they’ll help set direction and keep people accountable.

Ideally, as mentioned earlier, you should also start brainstorming early on initiatives that you think could potentially drive the OKR forward.

Criteria For Data Scientist OKRs

Overall, a great Data Scientist OKR should follow the SMART guidelines of being:

  • Specific

  • Measurable

  • Achievable

  • Results-oriented

  • Time-bound

1st Data Scientist OKR Criteria: Specific

Your OKR should be specific enough so that when other people within the organization, that aren’t necessarily on your team, know what you’re working on. No one ever got punished for having an OKR that was a bit too long. Instead, writing fuzzy or unclear OKRs is a definite no-go.

2nd Data Scientist OKR Criteria: Measurable

Because OKRs are usually working with specific metrics, it’s a lot easier to check the box that the OKR is measurable. We measure things that are relevant to progress within SaaS.

3rd Data Scientist OKR Criteria: Achievable

It’s important to have achievable OKRs. They should always be ambitious, but nothing is more demotivating than unachievable goals. A rule of thumb is that the chance of achieving a Key Result should be around 50% from the start. If you reach 70% and above, it’s considered a success.

4th Data Scientist OKR Criteria: Results-oriented

Being focused on results is a very important aspect of OKR. It’s so important, that we’ve dedicated an entire section to describing why you should focus on outcomes over outputs below.

5th Data Scientist OKR Criteria: Time-bound

A key part of defining a goal is also defining when you’re expecting it to be met. For OKR, your goals are usually divided into different cycles and your goals should of course be reached within the end of the cycle.

Created by OKR Framework © 2024