Squad Health Check is a team health assessment model invented by Spotify, and today we’ll tell you how we measure our health in Yandex.Money and what came of it.
How to measure team health?
Typically, managers in large companies want to get an idea of what is happening in the development teams. This is necessary in order to eliminate systemic problems in time, focus on something specific, and direct the team to improvement. Therefore, such companies experiment a lot with various methods of measuring “internal temperature”. And when it comes to assessing how things are, the answers most often come down to numbers – KPI, metrics, or setting goals for any period – for a month, quarter, six months. For example, how many tasks have been completed, how many have gone into the backlog, how many have saved and so on.
The problem with quantitative assessments is that they do not show what lies in the sphere of “psychological comfort” of the team. It seems that all the tasks are completed on time, but by hell refactoring and a significant deterioration in the quality of the code. Or there was not a single fakap, but PM drove the team into terrible depression.
When the Scrum methodology was developed, a good “tool” appeared, which was called a retrospective. In a retrospective approach, the team meets and discusses what went well with them and what was bad: in fact, this is the team’s self-esteem.
What are the disadvantages of retrospective?
- She has a small time span: the team answers the question “What bad / good happened during the sprint?” – not all things are visible from such a close distance.
- In retrospect, many prefer to remain silent and not talk about negative factors. For example, it’s quite difficult to tell the manager in the face that, to put it mildly, he behaved incorrectly during the sprint.
- And finally, the retrospective often comes down to the format of “came, chatted, parted, repeated at the next meeting”. There is no analytics or problem solving.
The Spotify guys moved further: in 2014, they came up with an idea like Squad Health Check that answers questions that retrospective cannot answer. At some point, we stumbled upon this article and decided that it would be nice to try to host the same. We have been doing this process for exactly a year, so it’s time to share our experience. Next, a little materiel and practical tips that will help to avoid our mistakes.
What is Squad Health Check?
Squad Health Check is a team assessment model that allows you to understand the status of a team or, in Spotify terminology, a team’s health, in the form of a general meeting with voting on a specific set of indicators. With this approach, the team independently evaluates itself.
When we hold such meetings, we explain to the teams as follows: Squad Health Check = Retrospective + Planning Poker
Like any other model, SHC has its own vocabulary.
Halschek – Actually, Squad Health Check itself. We have reduced the big name to such neologism and are already used to it.
Indicator – a question that the team must answer. Spotify offers 11 indicators, we limited ourselves to 10 indicators. I think more than 15 will already be too much.
Semaphore – This is an assessment that an individual participant must give. In our case, it can be red, yellow or green.
Dynamics – how the state of the indicator has changed in comparison with the previous health check. If this is the first time you are doing this assessment, then suggest recalling the situation 3-4 months ago. We have three types of dynamics: “better”, “worse”, “unchanged”. Sometimes options are offered: “incomprehensible”, “all in a circle”, “in a sinusoid”. This is a matter of agreement.
final grade – a combination of dynamics and semaphore of the whole team, or in a different way a general feeling from the state of affairs. The team independently chooses this rating during the discussion.
By the way, in order to understand whether your company has “logged in”, it is worth listening to what words people use in conversations. If the team begins to say something in the spirit: “Guys, we are in the red zone by this indicator,” then at least SHC did not cause rejection in your company. And if a lively discussion ensues and the discussion of problems is in full swing, then the helcheck works.
Step number 1: determine the indicators
The Spotify website already has the proposed set, we thought it over, and at first our indicators looked like this:
- technical quality
- suitable processes
- the help of others
- convenient features and releases,
- product value
However, in the process of work the teams always had questions about such names. For example, whose mission? My mission as a person, developer or team, company? Other names also included misunderstandings. Therefore, after some time, we clarified some of them so that the team clearly understood what it was. We generally excluded one indicator (training), because people always evaluated their own training and it was often impossible to come to a general assessment of the team.
Here is the list:
- technical quality
- team processes
- pawns or players
- the help of others
- team atmosphere
- development processes
- product value
- team mission.
Is it worth it to constantly change indicators or leave them unchanged – a debatable question. On the one hand, ideal options will not work right the first time, and questions may also arise like: “What about pawns?” On the other hand, if the question is constantly changing, then how to evaluate the dynamics of health?
Step # 2: set the value to semaphores
Green color means that the situation suits, nothing much needs to be changed.
Red is a disaster, the situation definitely needs to be changed.
Yellow is somewhere in between, neither good nor bad (Spotify uses the expression “meh”).
We think that the players in Magic: The Gathering have already recognized the familiar mana symbols. Yes, that’s right – for the semaphores we took the basic lands that were familiar to them, which can be easily obtained.
You can use any other option (for example, ordinary colored cards), but in our case a small fan element is added that sets the team to the desired mood.
By the way, if you go to the nearest MTG club to buy land, then do not forget to buy protectors for them, otherwise you will have to throw out cards after a couple of health checks.
Step # 3: describe each semaphore and indicator
The slide structure of our presentation for helchecks is as follows:
- indicator name
- brief description / indicator question,
- Description of green assessment
- Description of the red rating.
Each indicator should come up with a set of questions to make sure that each team member understands the indicator in the same way as everyone else.
Below on the slide we show two extreme situations: definitely green and definitely red. We decided not to describe the yellow semaphore: this is such an average position on the scales when it is difficult to lean into the red or green zone.
For example, at first the questions for the Teamwork indicator looked like this: “Are we a team or not? Do we hear each other and help? Can each of us openly express our opinions? ”
Then we realized that we made a mistake – the indicator had too many questions, so people often threw yellow cards. Most argue like this: “Yes, like we are a team. Yes, we kind of hear and help each other, but here it’s not accepted to express our opinion openly, I’ll put a yellow mark. ” As a result, we greatly simplified this approach and are now trying to ask one specific question.
As a result, the question about the team turned into: “Are we working as one team?” Semaphores were described accordingly.
The most incomprehensible for the teams was the indicator “Responsibility”. The question was: “Are we responsible for our decisions, for our products and for our services?” Then the stream of clarifications began: “For which decisions should be held responsible? I am always responsible for the code that I write. What is the question? ”It turned out that in the green semaphore, people reflected responsibility for their own decisions, and in red they saw a completely different situation: the solution of quarterly plans.
They began to think: what do we want to know in general? We saw that Spotify had a similar indicator called “Pawns or players”. Initially, we threw off such a formulation, deciding that serious things needed to be solved, and then some pawns. However, it turned out that we had this in mind, but could not articulate it explicitly: “Does the team feel influencing something or is it decided by someone else.”
You can see the full list of indicators here – This is a minimalistic presentation that we show on our helchecks. You can take it and use it as it is, but our experience has shown that even each team has its own specifics and it is better to tailor indicators and semaphores to fit your needs.
Step number 4: find a disinterested person
In addition to the team itself, a meeting should be attended by a person who will act as the leader and moderator of the discussion. It is also advisable to call a secretary for a meeting – this is the person who writes down all the main points (not on the recorder, but immediately on the laptop).
To achieve the most honest answers, the facilitator and secretary should be relatively uninterested. Imagine that you come to a meeting as a team, and the IT or HR director asks you, “Are we a team or not?” And the team will immediately want to answer as it will be right, and not as they really think.
When all the necessary terms were found, we decided to try. Further more about how meetings in Yandex.Money are held directly.
Step number 5: hold a meeting and collect the results
Meet worth it live. Yes, this statement is in the spirit of the captain of evidence, but at the stage of preparation for the helschek there were proposals to make a bot for voting or use profiles. Tip: just do not do this.
For the meeting, we choose convenient meeting rooms where there is always a TV or projector (for slides), as well as a marker board or a flip chart (for recording results).
Also, we do not advise participants to take laptops. We use one laptop only for displaying a presentation with indicators and semaphores.
How is the assessment going? An indicator is displayed and the semaphores described, the host asks: “Is everything clear?” – then the vote goes on. Blindly. This is necessary to level the effect of the crowd or the halo effect.
Then the cards are “revealed”, and the team begins to discuss. At this time, the secretary should write down key comments from the participants of the check. Without commentary, it is sometimes difficult to understand why the team gave six green ratings, but the final one suddenly turned yellow.
As a result, on the marker board:
- the number of cards of each color is recorded,
- the final mark is circled – by a common vote the team decides which color to choose as the final,
- and the dynamics results are written in a separate column.
It turns out approximately the following:
If you carefully look at the picture above, you can see that even though the team is in the yellow zone (most circles are circled), the dynamics goes up – this sets up for an optimistic mood.
Step 6: analyze the data
After the first series of meetings with teams, a lot of these photos can accumulate:
Then we digitize all this into our confluence:
As a result, we enter the results for all teams in the summary table:
After this step, there is an important point. In general, the entire Squad Health Check is about evaluating the team inside. Needless to say, some team has something better and something worse. We look at how we are doing, not blaming for the results. If the team gives a red mark – excellent, it means that the team is aware that it has problems.
What to do next?
According to our indicators, all problems can be divided into 4 types:
If the guys say that they have problems understanding what product they are making, what is the purpose of the team’s existence, then we have a problem in the product area. In this case, you should communicate with product managers, show them that the team does not understand its purpose.
If the problem is the indicators “teamwork” or “atmosphere”, then perhaps you should try to change the composition of the team. Go to HR and say: “We have problems, people don’t want to go to work” … It happens that people don’t get along, and that’s normal.
If the team signals that they have a lot of legacy, some old processes, this is also a signal to the business that the team should suspend a little in grocery features and give them time to do refactoring.
It’s best to invite the team to meet and conduct a retrospective on problematic issues after the helcheck. Give the children the opportunity to try to change something themselves.
Some practical tips
Some things are not always obvious, and we reached them only with time. We share with you our work on the bugs:
- Do not make too many indicators. Ten is the optimal amount. If you can shorten to seven, do seven. More than 12 is already too much.
- Call the secretary. Call the person who will record what is happening. For example, we write in a chat and ask: “Guys, who wants to record?” A sudden discovery that many people like to see how the team is doing, to be a detached observer.
- Optimal time: 1.5-2 hours. Less – just don’t get it. And if you get more than two hours, be sure to take a short break. I am making a coffee break after the fifth indicator.
- Meet at least once every six months. Otherwise, a lot of water will leak, and it will be very difficult to compare the readings. Once a month – often. Once a year – the team may have time to forget what happened to them 12 months ago.
- One indicator – one question. Or two, but very much connected with each other. Do not try to put a lot of questions into one indicator, leave the most important.
- Red marks are normal. No need to put KPI on the fact that there should not be red marks. I repeat that this is the team’s self-esteem. This is what people themselves feel. And if you then punish them, then the next time they will put all the green. And what to do with it?
If you have a lot of teams and you are going to conduct HELSEC checks, then over time you will have to delegate the assessment to several people. What to do?
- First you need to choose people who match the disinterest criterion.
- Teach them all the points above 🙂 On several health checks, they can be secretaries and write comments.
- And then let them spend their first health checks, and you will be the secretary. So you can see that everything is going well.
Instead of output
Squad Health Check is actually a very simple thing. This model does not require anything special, except that you need to get the base lands of MTG 🙂 Therefore, we fix the algorithm again:
- Define indicators.
- Have a meeting.
- Gather Results /
- Analyze the data.