Atlassian's big experiment with performance reviews

No readers like this yet.
Carrot on a stick

Do you ever wonder if and how you could call a halt to yourperformance review process? Do you think traditional processes aremarred by the distribution curve (and forced rankings), huge timeinvestments and low impact on performance improvements? Maybe you agreethat your processes have their faults, but you think that it's notsensible to abolish performance appraisals altogether or replace them with coaching sessions.

Keep reading... Let us be your guinea pig! Our organisation, Atlassian, is conducting an open review of our new performance review model for everyone to see.

From speaking to HR executives, we've learned thatmany people face similar issues with their current process. Often theythink their organisation's review and reward culture is too entrenchedin their organisational processes, that it's hard just to change theperformance review model without any evidence and alternatives. We hopethat other growing, or even established companies can learn fromour experimentation, challenges and results.


For years, Atlassian's performance review model was in line with 'HRBest Practice'. Twice a year, people would review themselves and theirpeers via 360-degree reviews. Managers would review their team membersand determine their final performance rating on a simple 5-point scale that determined their bonus. I believe it's a similar model to that ofmany other tech companies like Google and Salesforce.

So, what was the problem? In short, twice a year the model didexactly the opposite to what we wanted to accomplish. Instead of aninspiring discussion about how to enhance people's performance, thereviews caused disruptions, anxiety and de-motivated team members andmanagers. Also, even though our model was extremely lean and simple, thetime investment was significant.


If you ask your managers and staff to invest time and money in aprocess, it'd better be worth it. And in our opinion it wasn't - well,not in its traditional format.

First, we analysed the traditional performance review model in detail.We asked what made people perform better and what parts of the reviewsworked well. We talked at length with other tech companies about theirexperiences and asked if the negative aspects of reviews, mainly related to de-motivation and the high levels of anxiety, tend todisappear after a while (they don't).

The triggers for this experiment was that we found nothing out there to copy (that would have been much better and easier). Sure, there are books and blogs dwelling on the negatives of performance reviews, but their solutions are too impractical to implement often ignoring the need for solid performance feedback. Also, there are systems that have a new approach to performance reviews, but no HR system supported our approach. We needed to start from scratch.

Key innovations & timeline

1. Rip apart the traditional performance review

We've replaced the traditional performance review structure with a more lightweight, continuous model.

    • We incorporated the constructive aspects of reviews in the existingone-on-one meetings. Atlassian managers already have weekly one-on-oneswith their team members. Now, every month, one of these meetings isdedicated to a discussion on how the person can enhance their ownperformance and play to their strengths.
    • Removed the unconstructive focus on ratings and get rid of the distributed curve.

2. Stop paying individual performance bonuses

Instead, we gave everyone asalary bump. Similar to Netflix's approach, we prefer to pay top market salaries rather than bonuses. However, we continued to pay an organisational bonus like before, so people will share in the company's success. Also, we've recentlygiven every staffmember stock options so people will benefit from thecompany's growth/increased value over time

3. Create bite sized chunks

The thing with our traditional review was that, despite goodintentions, it focused mainly on two sections: the manager rating andthe employee's weaknesses. This makes total sense as the first thing aperson will be curious about is their rating (which also affects theirbonus). Even if a person receives a "good" rating, most of the time willbe consumed by justifying why the person didn't get an "outstanding"rating. We wanted to introduce a lightweight and continuous model of conversations designed to remind people to - every now and then - talk about topics other than daily operational stuff. We changed the following:

  • All sections should receive equal attention. We think that the 360review feedback can better be discussed in a separate conversation. Samegoes for performance ratings, strengths, weaknesses and careerdevelopment, etc.
  • We split the sections into separate conversations with their owncoaching topics. Every month we allocate one of our weekly 1:1s to acoaching topic.
  • We created conversation guides to help everyone stay on topic, andto supply tips and tricks on how to facilitate the conversations.

4. Performance still evaluated - with some crucial changes

We think it's important not to shy away from giving honest performance feedback. Some 'anti-performance review' books suggested skipping that altogether, but we think it's a invaluable element in every manager/employee relation. In twoof our monthly conversations, managers will still check in with theirreports to evaluate performance. However, we make a few crucial changes:

- No 'exact rating'
We prefer people not to concentrate on theexact rating definitions,but rather on having a good and honestconversation on how they havegone in the past 6 months. However, there will be no rating on anumerical scale. Instead, during in the check-in the manager indicatesroughly how often their report hasdemonstrated exceptional performance.

- Two Axis
In addition to an evaluation of performance/achievements, we've added the scale on 'how often you havestretched yourself'. Put simply, thereason forusing both scales is to acknowledgebotheffort andresults. Two peoplecan both deliver similarresults, yet one ofthemhas solely focussed on their normal work responsibilities whilst the otherperson has really injected extra energy in improving things outsidetheir normal job (e.g. themselves, their team, etc)which has made a real difference toAtlassian. Feedback from Atlassian managers was that they wanted to acknowledge this.

- Focus on frequency of behaviours
The scale from 'never to always' helped people focus the conversation onhow to improve the frequency of certain behavious, rather thenfocussing on a number. Recently, I shared this idea with the cool team of Sonar6 who translated this in a blogpost in their usual colorful way. Both staffmember and manager will decide on the person's position on two scales. When the a certain position on the axis is chosen, the logical subsequent question is: "why didn't you position this person more to the left or right, higher or lower?" This approach also encourages a better coaching style review conversation.

Challenges & solutions

We've learned a lot and but it will probably take several iterations togetthis right. Along the way, we bumped into several challenges:

  1. Lack of best practice: there was not literature or best practice out there (or we couldn't find it) Lots of books focus on the negatives of performance reviews but they offer solutions that are too much focussed on just coaching, in not enough on the alternatives to the actual performance conversation. They also didn't offer much guidance on how to structure the coaching conversations themselves. We ended up merging different ideas including suggestions from the HR industry, strengths based coaching, motivational theories, and different rating evaluation scales.
  2. Systems: To my knowledge, there is no HR system that offers anintegrated approach including regular 1:1s, coaching 1:1s andperformance review check-ins. There are the heavy and traditionalperformance review tools like Successfactors and SumTotal. Conversely,there are some cool continuous feedback tools like Rypple - who Iunderstand are designing a more integrated solution for Facebook. We finally found an awesome start-up who has created a really smart solution - called Small-Improvements. Small Improvements features all sorts of goodies in a nicelightweight app, can facilitate our performance check-in process andcapture continuous feedback/coaching.
  3. Coaching conversations: The managers (some very new to the role)had little to no experience in how to handle a coaching conversation. Weended up engaging an executive coach who trained our managers in how to have a good coaching 1:1s.
  4. Team sizes: One of the first problems we highlighted was when a small group of managers said they wouldn't have enough time to discuss these topics once a month. And that highlighted a larger problem. Some teams were plainly too large for one manager to handle. This meant that, whilst the team was productive and well oiled, some people rarely met with their direct manager to discuss non-operational topics. That was one of the reasons we decided to form smaller, more agile teams and introduce a group of new team leads.
  5. People love numbers. They are objective and easy to analyse. So, when we decided to take out performance ratings many people felt uneasy. Truth is that people generally understand that, if their performance is indicated to be closer to the bottom-left corner of our performance/stretch axis, they have to step it up. You don't need a rating or a number for that! Also, we noticed that salary reviews and talent management didn't need an exact performance rating - as a matter of fact, we've never had a direct relation between salary and people's ratings. Other factors such as market salary movements, internal and externalsalarybenchmark data, years in the job and skill-set all helpdetermine salary growth. Few organisations would calculate their salary reviews in a simple formula.

Benefits & metrics

Honestly, this experiment may be better categorised somewhere between a 'hack' and a 'solution'. Most other entries in this category are hacks - probably because changing 'traditional best practice' is relatively new territory. And as added complexity, when you just changed your (performance) measures, it's hard to say by how much people's performance has increased :-) Despite all this, I thought it would be worthwhile to share our story, if only to help other organisations who are currently considering changing their process. So, although it's hard to quantify at this stage, some outcomes below:

  • More focus on people's strenghts: instead of only operational 1:1s, every staffmember (100%) has now had discussions about their likes and dislikes of their current work. Coaching conversations uptake has been awesome. Within weeks of launching this new approach, all staffmembers have had converations about their strengths and how to dedicate more of their time on things they love.
  • Motivation & recognision model changed. Instead of performance incentives that don't work, our model has shifted to encourage desired behaviours through better coaching. Also, a Kudos model where every staff member can recognise great performance of co-workers without their manager's approval has been implemented to supplement this new performance model. This peer feedback model has been extremely popular. since launching this, already 150 peer-kudos gifts have been handed out. 75% of staff who displayed outstanding performance were recognised by peers outside any formal process
  • Engagement: Atlassian has been awkowledged as an Employer of choice internally and externally. Independent and internal staff engagement surveys showed extraordinary high engagement scores of 87% and 83% respectively. Last year, we've won several Best Employer awards, including Best US medium sized companies to work for, and a Highly commended award in Australia for HR Leader's Employer of the Decade.


  • Challenge HR best practice - if the whole organisation hates something, yet you invest huge amounts of time and money (through bonuses) in performance reviews, something is wrong, best practice or not. It is HR's job to fix these things!
  • Don't be stopped by the lack of HR systems - It's very difficult to change your model when HR systems can't support your ideas (and we can't go back to pen & paper). Big HR system providers will encourage you to keep with the current model and present traditional reviews as best practice in magazines and at conferences. Keep charging ahead - there are solutions out there developed by new start-ups.
  • Make bold decisions - Performance bonuses are ingrained in our compensation model and staff ratings are the starting point of our talent management programs. It's hard to change these things. be courageous.
User profile image.
I manage the Global HR/Talent team of Aussie Software Company Atlassian. Offices in Sydney, Amsterdam + San Fransisco.


This is a great read and very timely as we are just taking on our first permanent hires rather than contractors so giving a lot of thought to compensation, bonuses etc.

My real horror stories stem from my time working for investment banks, where the annual bonus cycle meant zero delivery for 9 months of the year, with everyone holding back for Q4 (then a few months of positioning /backstabbing to claim the biggest slice of the pot!)

The use of distribution curves gave me a particularly warm feeling come appraisal time, when I knew I had to pick 2 of my motivated and performing team and reward them with a D grade (no bonus, no payrise and a threat to improve of face firing), simply because 10% had to be in that curve.

I understand now why every other team had a proportion of "dead wood" - simply to be held back for appraisal time rather than my (maybe naive) approach of weeding them out through the year and hiring motivated replacements.

So now I have my own company, I am very keen to avoid the negative aspects of traditional compensation and appraisal schemes and find something better suited to to today's working styles and people.

Some good food for though here - I especially liked the performance / stretch curve being plotted together.

I'd be happy if my employer just kept my base salary up with inflation. I just make sure that there is an equivalent drop in my productivity (they're losing about 7 weeks from me now and I find this very relaxing) and maintain the same quality and we're all even.

I applaud Atlassian for questioning the status quo and forging ahead to find their own solutions to this problem. I particularly commend them for the openness of the process. We all can learn from each others' mistakes, but only if we're not only bold enough to make them but also (harder) to make them public.

The honest fact is that the 'HR Best Practice' does not apply to most teams in the software industry, yet we keep trying to follow this lead simple because the Practice is labeled Best.

I've blogged about something along these lines ( The post received no online comments but several offline and all parroting the "but that's not the way it's done" party line.

To that I say: So what? Take the time to challenge the way it's done. If it ends up that "the way it's done" equals "the way it should be" then you've just validated your time and effort. If it doesn't then you know it's time to change course rather than spin your wheels on a process which doesn't work.

Creative Commons LicenseThis work is licensed under a Creative Commons Attribution-Share Alike 3.0 Unported License.