Usually the salary increase looks like this. Method # 1, humanitarian: after a year of work, an employee thinks that something has gone wrong, and it's time to ask for a raise. Waiting for his local maximum effort, and on this wave goes to the manager to ask for more money. From the point of view of game theory, it looks like "well, I asked, it suddenly works." The manager has no reason to raise the salary.
Then the employee can raise the rates. "Promote, otherwise I'll quit." In this situation, both are losers - the manager loses the time of training a new employee and the cost of recruiting. The employee loses on the fact that he may unexpectedly quit.
Developers traditionally use method number 2: first, they go through several interviews somewhere, collect offers and come with them to the manager. "Look, here they offer me 20% more, but I like it here, raise it by 15%, otherwise I will move." This is already a subject of discussion. In a trivial case, it is easier to raise and retain a valuable employee, but this will ensure losses in related games. That is, it will create a precedent. Therefore, the decision is made (in a simplified model) with some degree of randomness.
Many of us have a history of mathematicians. Considering this game further, we can draw a simple conclusion that such a dialogue is always stressful for an employee, and it happens in a moment after a crisis. That is, at first a person is worried, then he makes potentially unprofitable actions (passes interviews in other places), then comes. Parts need to be raised, parts not needed. The next question is: can you find a function that will provide a fair estimate? Will this function relieve these stressful situations?
Regular reindexing every year is a variant of this feature. Conventionally, if the contract stipulates that the salary grows every year at the rate of inflation - probably you can not worry. But Vadim came up with a more interesting trick - to tie this to the assessment of the usefulness of the employee's actions for the company. But as an adequate person, without KPI.
What is the utility of an employee's actions for the company?
The contradiction is that the leader wants the result, and the subordinate wants money for his time. In an ideal world, one is easily converted to the other. In the real world, this is not even close to being true, because there are tons of options that can go wrong. I will assume that each of you got into a situation of filing a feature that was not included in the release later; or urgently-urgently-time to the deadline, which turned out to be not so that line, and not so that it is completely dead.
The contradiction can be resolved with a stake in the company for a specialist, but there are about the same pitfalls. In practice, a large company does not do this for most employees.
In general, hello. The last time I wrote about the business was when Mosigra was not sold. Then I began to plunge into the dark depths of the development of domestic tourism in Russia. I must say, the timing was perfect: on the one hand, where he temporarily went, you know. On the other hand, it turned out that a lot can be changed right here and now. Because those who weren't paranoid died. It's nice to work in a huge team with people who think like physicists. Tutu's size is exactly such that it is not yet a corporation with all the bureaucratic idiocy, but also not a small business with the fact that you know everyone by name. That is, there are already idiocies, but they are still rather amusing and manageable. About 400 people, of which 180 are IT. Leaders of various levels are often mathematicians, former developers, or someone else who began to dive into the business from the wrong side,where the soft skills are. And, just as autistic people I know build crutches in the spirit of “a macro of the beginning of a dialogue: smile, ask how are you, smile for 4 seconds, nod twice,” so the company turned out to have several rather strange things that correct the deficit of social skills.
In short, Vadim was drinking compote and suddenly came up with a robot that goes every month to demand a raise. In general, he was initially completely not for this. Let's step aside just a second.
A couple of aspects of leadership
From the outside, it may seem that the task of the leader is to determine in which direction to dig and set tasks. This is yes. In fact, the first function is to provide a good team, and the second is to synchronize understanding within this team. This is when it's about a big team, not about rushing ahead of everyone and doing the same job. That is, as a company or department develops, the role of a leader may change, but in a more or less stable state, the priority is precisely to say the right words to the right people.
In short, it works like this: without a team, you can only do something at a speed of X. By hiring 5 people, you will already get a speed of about 2X, because some of the resources will go to transaction costs within the resulting architecture. For example, people are extremely ineffective in negotiating, so instead of short mental dialogues, you get meetings every day. I used to think that there are some kind of silver bullets that help to quickly sort things out. This was around the time when 15-minute meetings once a week seemed to be the pinnacle of efficiency.
Different control systems allow you to balance between reducing these costs and the clarity of the task. Turquoise schemes, for example, assume that there is very little cost, but so does unity.
But to hell with abstraction. Here's a small team on the product. The elementary question "what is the best thing to do next" can cause a fierce holivar. The most striking example is to redo the architecture or screw in a feature, which then will have to be written from scratch on the new architecture. In fact, there are dozens of such elections every day, and they are of very different sizes and genesis. In the case of a question with architecture against a new feature, you can arrange a stabbing, you can listen to "do as I said" from the manager and sadly go to write a statement, or you can agree among yourself about the most fair solution. Separately, I note that the level of justice is different for everyone. In construction, for example, I heard about the criterion "so ugly", which everyone immediately understood. We usually introduce some kind of evaluation function,which allows you to calculate the benefits of one and the other solution in parrots or elephants by a certain date. It is not always money, because there are things that, in principle, are quite difficult to measure.
In general, an understanding of what exactly needs to be done, how and in what order comes from somewhere. I have already described this painful procedure once, so let's get closer to the topic - salary reindexing.
What is employee efficiency for a company?
In short, the loop looks like this:
- The employee constantly knows what to do.
- The manager knows that the employee is doing the most useful thing at the moment.
- Corrective feedback is received in response to the completed pieces of work.
We started by making reports where three things should be sent at the end of the month: what was required of the employee, what worked and what did not work. The logic was to simply get the summary from the fields on time. End of month, took and sent the questionnaire.
And here - hoba! - it turned out that people do not always understand what is most valuable in their work. That is, someone can polish some trifle for weeks when the next commercial feature on the list, which will cost millions of rubles. In a day. And I'm literally now. The first days of the transport crisis looked like this: millions of rubles in losses. To simply stop them, you need to do a certain amount of work, a plan of about fifty points. And if you do it inconsistently, out of order, with unnecessary meetings - all this is very expensive.
That is, it is not necessary for a specialist to understand the market, the company's place in the market and other subtle matters. You can just get the TK and work on it. But it is important to do what the leader expects. We have had, are and will have problems with this. To make it clear what kind, I'll just say that there were situations when a subordinate did not see the head for six months. They just gave him a task, provided everything he needed, and he sits there and does. As in a joke about dumplings. Conventionally, this is setting such parameters as “in this release we make tasks slower, but with a clean architecture”, “we sculpt crutches at the speed of light, we need a release on Friday, and then let everything burn with fire”, “let's agree that the designer helps people, but does not send them in a forest if they cannot formulate a task, "" it would be good if lawyers would answer exactly how this can be done, and not why it should not be so "and so on.Well, and also often people do what they understand and what they like, and for all things with development (and usually they are the most valuable), you need to do something new and scary.
That is, the effectiveness of an employee is the compliance of the manager's expectations for his area of work and the fact. Simply put - when at the beginning of the month it is possible to predict approximately what will be the result by the end of the month and in what way it will be achieved. This is called “syncing expectations,” and it needs to be done in personal relationships, on a project, and with clients. True, the project can greatly raise the degree of cynicism of the process.
It seemed logical to have meetings to discuss tasks at the beginning of each month. The link between the report and the meeting allowed, in theory, to receive everything at once. The real world, as always, hit in the face with the fact that it just didn't take off. As elsewhere, if you want to control some parameter, you first need to measure it. Plus, among other things, it turned out that an awesome result is achievable "the manager did not set up a meeting."
That is, the idea of the humanitarian “well, we will somehow agree there” and the binding of the rule “somehow happens there on the first Monday of every month at 11:00” had to be dismissed. The solution was seductively easy, but it didn't work.
Second model
In many companies there is some kind of silence that the salary is revised once a year. Either it is indexed to the inflation rate, or it remains the same ("well, you didn't do much work"), or it changes upward. It suddenly turned out that people grow professionally more than once a year. Obviously, it was necessary to have a more flexible system.
Logically, an increase in the frequency of revaluations suggested itself. We started reviewing the results on a monthly basis.
The first version of automatic salary growth looked like this: the manager rated the employee at 0-1-2-3, where 0 is below expectations, 1 is within the limits, 2 is above, 3 is surprised. The estimate was equal to an increase in salary percentage, that is, it was possible to gain both 0-0-0-0 at the minimum, and 3-3-3-3. The salary increases at the end of the month were consistently based on appraisal; there were no one-time bonuses for particular success. An obvious problem appeared in the fact that often people did not understand, on the basis of which the leader makes the assessment, which affects the result. In fact, it turned into a mini-analogue of a special hell called "360 rating", when managers collected feedback from the environment. That is, he asked colleagues and clients - whom he could catch. As you may have guessed, this scheme has as much to do with performance as the number of pirates is related to global warming.
In this scheme, the model of auto-increasing salaries was good without going to the manager with a couple of offers in hand. Each month the report suggested that such an increase be requested. Automatically. Did a good job - here's the score. I just worked - here's the estimate.
There were problems with the assessment itself. There was almost no feedback.
The next pitfall is called KPIs. I have seen enough bonuses for this scheme at one time, and I can say that it only works when the system is not being hacked. That is, about + 0% of the time. Or, these same KPIs need to be changed faster than employees can figure them out. Plus, in a year we get demotivation, because “today is a feat = tomorrow is a plan”. KPIs are good for two things: for shift competitions and for evaluating efficiency in financial planning. In case of motivation, the plan is not the best.
In the case of smaller companies, everything would be solved by hand, but somewhere deep in the company's DNA - to automate what is usually not automated. And so Vadim proposed an unusual, but quite elegant and harmonious system.
Third model
The next update is this: what if at the beginning of the month we set the rules of the game, and at the end of the month the employee will evaluate himself according to these rules?
Made the form in Confluence. In addition to "what was expected of me this month", they also made an item in the methodology "what is expected next month" - that is, we shifted the cycle forward by 1 cycle. And we filed two estimates: one comes from the employee based on the results of his own report and comparing the tasks of the month set to himself, the second from the manager.
That is, at the beginning of the month you need to find out how to get a good grade. Receive it at the end of the month, based on the expectations agreed upon the month before. This is much simpler than solving the problem "what the market needs".
In this case, reports must be sent every month. It turns out that the leader sees how correctly everyone in the team understands their main focus. And, which is very important, the tasks are formulated by the employees themselves - this is a double verification. If they were formulated in the wrong way or in the wrong direction, then they are discussed again. For this there is a meeting following the results of this self-review.
What is important - the rules are discussed on the shore. No "well, I had to understand that it was not so important" - at the beginning of the month everything was fixed.
I.e:
- The employee gives himself a grade for the month.
- The manager agrees or gives the employee a different grade for the month.
- If the assessments differ, they discuss why this is so within the framework of the established conceptual apparatus. I know of many examples where the manager disagreed and raised the grade. And vice versa too.
The assessments themselves are:
- 0 - worked badly
- 1 - expectedly good
- 2 - greatly exceeded expectations
At that moment, cadres intervened and said that once a month it was not a plan to re-index salaries, because there are many places where you need to do everything physically on paper and give a bunch of reports. The logical compromise was a quarterly revaluation. The idea of a quarterly assessment based on the results of three months' work was shallow at the start: in this situation, only the last month is assessed, because this is how memory and psychology work. Therefore, the frequency of assessment is monthly, and the recalculation of salaries and bonuses is quarterly.
It turned out that for each quarter the employee receives three grades: [0..2] + [0..2] + [0..2]:
- 0, 1, 2 - not well done, the quarter was not successful, there are no bonuses or bonuses.
- 3, 4 - the salary will grow by 2% from the next month, a one-time bonus of 10% and 17%, respectively, of the monthly salary.
- 5 — 4% 30%
- 6 — , 50% .
There is also an award for special merit (conditional hidden grade 3, which can only be given by a leader). At first they called it the "Nobel Prize", but then they renamed it, because there were outstanding achievements during the crisis, but Nobel implies a breakthrough, and it was not always just some concentrated blow.
According to practice, for a quarter 3-4 in divisions have in the region two-thirds of employees. About 10% have 2 or less per quarter, the rest are higher. Any case of assessment outside the 3-4 corridor is a reason to go to a performance review, that is, the possibility of inconsistency with the position. This procedure is done every six months in April and October. An invitation to a performance review does not mean auto-promotion or auto-dismissal.
One of the most recent reviews was on why an employee looks like 2-2-1 and 3-1-1 for the last two quarters. It turned out that she was part-time, did the full job during the crisis, and the manager gave her a high result. In the course of the quarter, they were transferred to full time (for the same job). At the review, we talked about the whole situation, no additional changes were required.
What turned out in about a year
The system is much more flexible than KPIs, since it takes into account the plot twists in the real world and the real contribution to the work on the one hand. Focused on employee honesty. On the other hand, I don’t think that it will go for a ride somewhere in retail, for example, without some modification. Not suitable for streaming repetitive work such as a call center (there are still different types of development and KPIs are better), but very well suited for development and project work.
There were concerns that the manager might abuse the assessment. This was feared by those who have a bad relationship with the leader. A logical question was asked: why then work in this team at all? The general solution is that the manager is generally interested in the employees doing well. If this does not happen, he will also feel bad. That is, I'm not sure if the system is applicable to state-owned companies.
However, an additional procedure has appeared: a manager cannot simply take and give an employee a score of 0 if he has 1 or 2 in his own. To do this, you need to coordinate such an assessment with a higher-level manager. In the case of development - with the technical director. On the other hand, the CTO must track the average output in the divisions and see what his team leaders give grades (2) for - because there is still a certain standard for the company, that there is good work. We found this on the fact that the percentage of marks (2) in one of the teams exceeded one third. They began to understand - and this is April, the beginning of the crisis, and there they really overcame.
There were only three estimates (3) in the last two quarters. We introduced a restriction that there can be no more than 5% of them (in the number of people) per company per month, so as not to dilute their value.
results
First and foremost, the system guarantees a 2% increase per quarter during normal operation. That is, if you just do what you need, as usual, without going into a binge - there is an auto-increase of 8.2% per year. To take a 4% increase per quarter, you need to do something that does not fit into the plans and job responsibilities. Plus there are one-time bonuses. In fact, it is impossible to take 4% all year long, since "I am the fire" means that either the employee is not working in his position and needs to be raised to a suitable level of incompetence, or he will break down now. The comments “don't wait any longer from me, I can't do this a second time” are quite normal.
From what I want to fix:
- We need a reciprocal system for evaluating managers by subordinates. Testing options. While it looks strange.
- , — . , .
- , , « , - ». .
- , — 10- . , . , , . .
- , , 5 - .
Plus, all this still had to be merged with the Labor Code, which does not allow two identical specialists in the same positions with the same job responsibilities to have different salaries according to the staffing table.
In 2-3 months, expectations were aligned throughout the company.
Here are the results of an anonymous survey:
Now, please, dispel this scheme. She seems suddenly healthy to me, but I suspect there is better.