How to tmlidu survive in a scalable scram and keep control over the quality of the code

    The technical director of ivi, Evgeny Rossinsky ( eross ), is well known to the participants of our conferences on reports on the technical side of streaming. But today you have before you the decoding of Eugene's report on  TeamLead Conf about how Agile-transformation is reflected on team leaders.

    Not so long ago, ivi underwent an agile transformation and got a good profit in terms of it:

    • business,
    • development speed
    • time to market;
    • other interesting metrics.

    But the consequences of this creative decision quite seriously hit the team leaders. The report is about how to deal with it, and what tools to use.

    Before transformation

    In order to understand what I'm going to talk about, you need to get a little familiarity with our product. Then we analyze why we have such a format of teams and why we chose this path.

    Development in ivi comes under five major platforms :

    • iOS,
    • Android,
    • Web,
    • Smart TV,
    • Windows + Xbox.

    And, of course, the backend, without which nowhere. Before we decided to carry out the transformation, our teams were formed according to competencies.

    Guys who know, for example, Swift or Objective C:

    • sit in the same room;
    • receive tasks from product managers;
    • working on them and producing a result.

    It seems to be all right, but there is one problem - the platforms are very different in terms of product and user consumption of content.

    This means that each team began to appear its own backlog and its priorities. For example, users of a web application do not like to pay, and the consumption of content is carried out through an advertising model. The features that are needed for this platform naturally appear in the backlog. Smart TVs are characterized by the fact that they buy well, and the features of the paid model appear.

    It turns out the following situation: someone came up with a brilliant idea how to improve the product; wrote a lot of tickets that go to the product managers, each responsible for their platform. Managers pass ideas through the lens of their perception, perhaps they are modernizing something and inserting it into their work plan. The problem here is that one feature can be released on all platforms for more than a year, because “I believe that it is not a priority for my platform”. And in a year, half a year or just a few months, anything can happen to a product - for example, a redesign or a concept change will happen, and this feature should already be rolled out, and it is completely different from all other platforms.

    As a result, after some time it turned out that on different platforms we have completely different productswith different communication messages and business logic. The user in such a situation begins to get confused, because he does not have a single space in which he feels confident. In addition, it is very difficult to build hypotheses , since it is not always clear on which platform which functionality is better shot. It all depends on the size of the screen, on how people consume content. And it all looked something like this: The

    ingenious productist brings the idea, and the developers on different platforms perceive it rather unfriendly, with the exception of the backend, which in general does not matter what to write.

    So, since the company had enough competences regarding what flexible methodologies are, we agreed with the business that we need to remove the barrier between:

    • product,
    • business
    • technologies

    interact productively and do one thing in common.

    After transformation

    This resulted in an architecture with dedicated Value Stream , in which each team deals with a specific business area.

    To form a new structure we:

    • conducted several trainings;
    • we determined the main directions for our business;
    • in the voluntary-forced broke people on the value stream;
    • have begun to implement.

    True, before this, we made one such team, by the example of which we held demonstration performances , gashted one feature on all platforms with an excellent time-to-market.

    In addition, the feature was chosen very well and was successful on all platforms. Thus, we had an example that showed that everything can be done faster and better . This allowed the entire development team of 130–140 people to understand that it is possible to work differently.

    When we determined the value stream, the question arose of what to do with the timlids. Previously, they were the basis for everything in their direction, and now appear value stream and greater influence of business. What have we done? We in each value stream gathered people of different competencies, at least one:

    • iOS developer,
    • Android developer
    • Javascript developer
    • Smart TV Specialist,
    • coder,
    • to the tester.

    It turned out like an independent team, but it is  independent in words and on beautiful Agile-coach slides. In fact, everybody forgets about the fact that there is something in common between people who can write in the same programming language - this is their program code, through which they must somehow communicate. For some reason, many are silent about this, but this is really quite a big problem that needs to be remembered.

    Suppose you are seating people on different floors or rooms, but the code they write is the same. There is a release cycle and features that should not interfere with each other. There are many problems.


    We adopted several basic concepts:

    We have the value stream commands and  platform commands . In value stream, the guys implement a specific feature, and in the platforms there is a timlid - a center of competence. Inside the platforms, they can also develop some features, but this is more an architectural look at software development.

    We have introduced the concept of " Guild ". Placing the same iOS developers in different rooms, we needed to give them a formal sign and even informal that they remained the same iOS developers, and not just participants in the value stream advertising product, for example. And then with this guild you need to do something and somehow teach people to communicate inside.

    The next question waswhat to do with release cycles . On those platforms where you can roll out as the feature is ready, there are no questions. And in the case of iOS or Android, when due to the specifics of obtaining approve from Apple, it is impossible to send an application to release twice a day, you have to accumulate some packs.

    We decided that every platform that cannot be released immediately will be released once every two weeks . We have created a special calendar available to everyone, where the team publishes the date feature freeze and the release date. If, being in the value stream, you want your task to be available to the real user, you should have time to feature freeze. I managed - well, I didn't manage - you are waiting for the next train.

    Timlid in the new scheme

    What has become of our tmlidy? They went according to the classical pattern of awareness of problems that cannot be solved.

    At first we ran into  denial . Because for several years, the team leader was assembling a cool team, cultivating each employee, lured someone from competitors. And these specialists are now assigned to jobs that do not always meet their expectations.

    Of course, after denial was anger .

    There were many scandals, after which bargaining began . Everyone came up with the question: “Let everyone have your way, and I’ll have it as before, but I’ll take a marker, write Agile, I will wear a T-shirt with this inscription, but everything will be as before”. Did not work out.

    But in the end something happened that I was very much waiting for - people understood why we are doing this. I hope they understood, though maybe they lied. The very first successful case showed a real striking difference between what it was before and how it can be done differently. Twice the reduced time-to-market allowed us to define a benchmark for all. The next step has taken place, which in the classical model of psychology is called the Stockholm syndrome . People who adopted the new rules learned to exist in them and began to slowly promote this "religion." I can not say for all tmlidov, but about 30% of them are now active "preachers" of this direction.

    Problems and difficulties

    Now I will try to more structurally list the difficulties we faced:

    Due to the fact that the employees were seated in different rooms, the verbal communication was lost. For a month it became very difficult , because earlier there was an opportunity to quickly ask a neighbor, for example, what should be inherited from, and immediately get an answer. And now you are sitting in a separate room, around you are people whose technologies you may not like and have no one to consult with. To ask someone a question, you need to write in the chat, and the person who can answer, left - everything, you are left to yourself.

    We immediately started having problems with the Code Review , which I consider to be not problems, but a great discovery. We found out that a large number of employees are not independent.. Suppose there was an employee who, according to statistics, reviewed the tasks at most from the second time, usually from the first time everything was fine. And when he left to work on value stream, for some reason, it became 5-6 iterations.

    They began to understand, and it turned out that the authoritative opinion of the symbolic figure of the team leader, who controls, centralizes all the information flows in relation to himself, does not very well influence the development of developers. People are lazy , ask for help on all issues and get competent answers. And therefore review passes quickly and cool. Being alone with themselves, many developers had to grow quite significantly above themselves in order to make the same architectural decisions.. No one likes to hear five times in a row that your code is not very good. It is frustrating, makes you consider yourself an insufficiently qualified employee, and can even lead to depression or the belief that maybe the work is bad. But the point is that you need to look at yourself, understand that in the previous mode of operation you did not think about these things, but now you are developing and have to start thinking about them .

    On the other hand, we had some very interesting things that, in my opinion, are redundant with respect to the Code Review. In one of the teams there was a rule: in order for the task to pass the review, all team members must accept its decision. When they were all sitting together, they did it more or less, and now they all sat down - some had a daily stand-up meeting at one time, others had some other business - and the review of tasks became a narrow bottleneck. Over and over again, they learned in retrospect that some tasks hung on a review for two weeks and were forced to change something.

    Then began: separatism, discrimination and envy .

    Previously, a man thought he knew what he wanted to do. And with the division into value stream and platform, a suspicion began to arise that “the neighbor has a better taste”: the tasks are more interesting and growth is faster. The second problem is that when everyone becomes feature-centered , the quality of the code gets very bad . There are people around you who want to quickly get a result, and it’s not their task to think that it will need to be somehow maintained, modernized, and colleagues shouldn’t break anything from a new feature.

    There were situations in which the high speed of development had its opposite side - low - quality codewhich began to multiply in inhuman quantities. When the responsible team leader takes up the correction of these errors and refactoring, it turns out that his life turns into sweeping garbage for careless employees. The developers quickly succeed, everyone is happy, and they do not notice the titanic work of the Timlid, thanks to which everything actually works. After about two months, each team leader expressed dissatisfaction with the situation.

    To solve these problems, we had several meetings with the Timlides first, and then with all the guilds in order to explain to people that it is possible to work differently. If you want to try something else, then with this you need to come to the team leader, and you can easily swap places. The main thing is to agree among themselves.

    The power of flexible methodologies is that it does not matter to everyone how everything happens, the main thing is that people agree and be happy.

    This is on the one hand. On the other hand, so that there is justice regarding who is engaged in refactoring, and who makes new features, the Timlids started working with backlogs, I will return to this a little later.

    And then the difficulties began with Release Management . The value stream has its testers, the platforms have their testers, something is automated, something is not automated, you need to think about how to do a general regression. And then there are deadlines. We voluntarily decided to be released once every two weeks, and what if the partner comes asking him to do everything by tomorrow (and a bag of money, for example) to tell him to wait for the next cycle? Still, you need to look for a compromise . And then a beautiful scheme with releases can change a lot.

    A good example is the law FZ-54, according to which all purchases on the Internet should be accompanied by sending electronic checks to users and to the tax one. One may speak at conferences and talk about flexible methodologies and the fact that there are regulations and so on, but as soon as there is a risk of receiving a fine on 70% of your revenue, you switch to a completely different mode and make sure that your company is not fined. Such cases are infrequent, but there are. In particular, we had to make a second level of communication in case of changes in the scheme. It is not easy, but possible.

    The next problem we faced as a result of the transformation is with  new employees.. In my opinion, beginners should be immersed on Wednesday, as close as possible to the one in which it will work. And there, at the expense of the environment, he will quickly obtain the necessary knowledge. And we took the specialists who made up this environment to different floors and rooms. The life of a newbie in a company becomes a rather serious managerial task, the solution of which must be thought out.


    Here are the tools we use.

    I will begin the story with a  route map for a new employee. Unfortunately, this is something that we have not yet learned how to fully automate and, in fact, think over for each employee individually, depending on what he will do.

    There was a rather interesting example: we needed to grow a universal testerwhich would be able to test absolutely all products for an ad stream. They have a lot in common, but has its own specifics. It’s no secret that the tester, who is famously able to test a web application, will get lost the first time when working with the toolkit, say, for iOS. For this tester, our director of quality control has built a special route map. According to this map, a new employee in each competency gained knowledge for two weeks, participated in regressions and so on. With developers, the implementation situation is about the same. Thus, even at the interview we find out what the candidate is for and what is interesting to him, and we are trying to choose in which direction to send him.

    Of course, at first, a new employee pumps competencies, that is, in our terms, is in the platform team, and then can migrate to the value stream we need. By the way, route maps are good, but their adaptability should not be ruled out either.

    Then we introduced Guild Sync - synchronization of the guild actions .

    Why do you need it? Everyone listened to Scrum lectures, which talked about the need for a daily stand-up meeting, which allows them to find out what's going on around them. But our expert, for example, on the one hand is a developer for Android, and on the other hand is a member of the product team. If he has to walk and communicate twice a day, he will have some kind of Cargo-cult. At this rate, you can spend a lifetime on the meeting. Definitely, this will annoy people.

    If we say that we are based on a feature-centric model, then the daily stand-up meeting is more important in its value stream. But at the same time you can break away from what is happening in the guild. To do this, some guilds hold meetings once a week, some - twice a week. And then timlid methodically thinks about what to talk about. This should not be translated into an empty conversation, it is the development strategy of the technology team, which is the guild. Guild Sync meetings are needed so that everyone understands what technologies the company will use, what strategic decisions are made regarding this platform. It is a partial review of architectural solutions.

    In some teams we have more than one team leader. In general, the definition of team leads is very ambiguous. There are opinion leaders, which can be tmlidy and carry on themselves the decision of some organizational issues. There can be several opinion leaders with management skills in a team. And then they can organize themselves, who is responsible for what. For example, one of the opinion leaders could go to value stream. In this case, you need to synchronize your architectural solutions, which will later determine the development strategy of the entire technological platform.

    Then we began to conduct internal and external certain directions. In general, this is a fairly standard partly HR case, partly a team building case. For internal meetings, sometimes a vote is taken on which topic is the most interesting, speakers are searched either within the company, or we invite someone from the outside, we rent a special room and discuss various important points. On average, we have two internal mitaps per month, but sometimes four. People from other competencies can come to these meetings to listen to how their neighbors live and see if they want to change their specialization. This also happens.

    Now the labor market in matters of developers is not very simple. And to retrain a good, intelligent guy from one programming language to another is sometimes much cheaper than taking such a specialist from the market. Therefore, if a person is bored, we welcome the practice of transferring him to another direction , which is interesting to him. And internal mitaps help a lot to understand what this direction is.

    I'll tell you about another interesting point. I collected a small statistic, on the basis of which it turned out that during the Agile-transformation, smoking people were more knowledgeable, and Guild Sync and other meetings were almost not needed by smokers and drinkers.

    People working together do not often discuss general everyday topics, somehow it all comes down to work. Guild Sync is needed less often for teams in which many people smoke, because they already discuss everything during smoke breaks.

    Of course, there are non-smoking tmlids, then, for example, they practice joint dinners , where everyone gathers together at one big table. True, in a team of more than 10 people, this is no longer the case.

    The next tool to create a comfortable environment is rotation.. Initially, the biggest problem in all transformations is that people are afraid to ask whether it is possible in a different way. For some reason, it is believed that if the management said to do so, then it is necessary to do so. And only the most courageous and brave go and change something, while the rest think about changing the workplace, believing that they are not valued here. Although they did not even say that they do not like something. Therefore, we held a series of meetings and small trainings on the topic that the company does not matter who exactly is working in a particular place; it is important for her that there and someone work there.

    And then we started to apply various motivational attract performers in unpopular tasks. There are tasks that are very difficult to automate. Nobody likes to do them, but they need to be done, because this is, say, reconciliation of data with one of our partners. This partner constantly sends data, a lot of data that we need to process, in a different format. It turns out that once a month you have to greatly upset someone from the developers, because you need not just to run the script, but to understand what is wrong there, fix the script, run it again, etc. Automation such a process gives in badly. In this case, small bonuses , such a “harmfulness” surcharge, help perfectly . And there are more people willing to do an unpopular task, and the price of the issue is low.

    Guild Chatbecame much in demand after people sat down. Some teams have agreed to record the results of some agreements in general chat rooms, so as not to overload tickets, regulations, and so on.

    The most interesting thing is that our team leaders have learned to communicate productively with the product owners in the value stream. They began methodically explaining to them that if they made 10 features, and in a certain piece of code, something would need to be changed, then it would take half a year, because initially there would be a number of features. The positive factor is that the product owners took such treatment normally. And in the backlogs, tasks began to appear to refactor and fix their own crutches. And the Timlides breathed out quite seriously, as they realized that their teams remained their teams. They were convinced that they managed this process in the same way (not in terms of management, but, rather, in terms of communicating to everyone how to do it right).

    In my understanding, the task of a team leader is, above all, the definition of a strategy for the development of your software product.

    Teamlead - glue

    The main conclusion that we have learned for ourselves is that in a good way the failure or success of any brilliant transformation depends on the people who believe in it .

    If you venture something like that, then the focus of your impact should be primarily on the tmlid . Because it is very easy to come up with a beautiful picture from which it is easy to speak at the conference, and it’s not easy for people to run away and understand that this is a working scheme.

    For example, I realized for myself that after transformation many Timlides got rid of a large number of “internal running”. If earlier on all, even insignificant and unrelated to management or code, the Timlid was to blame, now with a similar scaling scheme for a number of issues it simply does not come to him. Timlid no longer needs to answer silly questions, because they are asked to the one who developed this feature, to the one who is responsible for the desired value stream. The team leader has fewer incoming distractions, and that's great. The Timlides became calmer , their eyes stopped twitching, and on the whole their self-confidence became higher.

    The next meeting on the exchange of Timlid wisdom is scheduled for September 24 and 25. We will meet in Saint Petersburg at  Saint TeamLead Conf to discuss all kinds of team management problems.

    The program is already beginning to form, there are more than 40 applications in the list of submitted reports , and as long as it is possible to add a topic there, the Program Committee will accept applications until 10 August .

    Also popular now: