Descriptive (descriptive) analytics. Using available information, it forms an objective and most accurate description of the object/situation under study. Answers the question: "What's happening now?". This is the basis of any analytics in HR. Operates with such data as personnel structure, labor standards, headcount standards, salary reviews, process efficiency metrics, internal and external benchmarks.

Predictive analytics. Based obvious dependencies and confirmed statistical hypotheses allows you to “look” into the near future by forecasting the number and planning the workload of personnel. It makes it possible to create a profile of successful specialists, develop an action plan to increase their involvement, and determine standards for passing tests.

Predictive analytics. Uses non-obvious dependencies, data from descriptive analysis, predictive analytics and Big Data to “influence events in the distant future.” Automated using artificial intelligence (AI). They are used to prevent/predict/identify the reasons for the dismissal of employees even before it happens, to predict their possible success/fiasco in a specific position.

Predictive modeling uses both traditional and innovative analysis methods:

Theoretically, predictive analytics in HR together with AI are designed to:

• achieve long-term company results (earning profit);
• identify problems with staff turnover and propose ways to solve them;
• monitor the costs of attracting new and retaining existing specialists, thus reducing operating costs.

However, in practice, not everything works out so smoothly.

Predictive analytics in HR and the Black Swan

Opponents of predictive models believe that analytics, by definition, is not “predictive.” As justification, they cite the idea that was popularized by Nassim Nicholas Taleb in his work “The Black Swan. Under the sign of unpredictability." According to the author, all incidents with significant consequences are black swans, that is, rare and difficult to predict.

For reference. The first European to see a black swan was the Dutch traveler Willem de Vlaminck. This happened in 1697 in Western Australia. Before that, representatives of the Old World were absolutely sure that swans were exclusively white. And now the final question: how likely is it that one of the inhabitants of Europe could predict the appearance of these birds with a black color?

If Taleb's theory is correct, then in relation to HR, predictive analytics cannot predict events that can significantly influence the resolution of personnel issues.

But, on the other hand, many large companies continue to invest heavily in the development of artificial intelligence, introducing predictive modeling technologies into the work of HR departments. Why do they do this? Let's try to find out.

Predicting the Unpredictable: 5 Examples of Using Predictive Analytics in HR

US Special Forces. You may be surprised, but US Special Forces HR uses predictive analytics tools to predict which candidate will become a successful fighter in an elite unit. The requirements are strict. We need to choose the best. And so to be sure. Guess, predict what is inherent in a tough special forces soldier: 1) a high level of IQ; 2) the ability to do 80 push-ups; 3) courage, endurance, strength of character?... The analysis pointed to point 3.

Google. In his book “Work Rules! Why most people in the world want to work at Google" Laszlo Bock, senior vice president of human resources (HRM), writes that the most important HR tool at Google is statistics. The initial interview with the applicant is fully automated, conducted on a computer, and configured to select the best candidate. The search giant also estimates the likelihood of people leaving the company using predictive analysis. One of the findings: new sales employees who are not promoted within 4 years are more likely to leave the company.

Hewlett-Packard. The HP guys are real fans of predictive analytics in HR. The company began to trust predictive technologies back in 2011. Then, in-house scientists pooled HR data from the previous 2 years and used a predictive model to try to predict the likelihood of each of the 300,000 employees leaving. As a result, the so-called “Flight Risk” indicator was generated, using which managers could respond in a timely manner to warning signals of layoffs. Thanks to this, HP saved up to $300,000,000. What else we found out: unlike Google, for Hewlett-Packard employees, not only promotion, but also an increase in salary is important; otherwise, they are dismissed within 5 years.

Best Buy. This consumer electronics retailer is one of the leaders in predictive HR research, during which company management determined that a 0.1% increase in employee engagement leads to an increase in annual revenue of $100,000. These results prompted a review of the frequency of internal engagement audits. : It is now conducted quarterly instead of once a year.

IBM. The Blue Giant has at its disposal IBM's Watson supercomputer, which evaluates a range of criteria that are believed to influence staff turnover. Among them: level and field of education, hourly rate, passion and satisfaction with work, frequency/absence of business trips, marital status, age and even distance from home to place of work. The column containing an indication of the employee’s “burnout” (yes/no) is highlighted in green (see screenshot below).

By analyzing this information, the HR manager can prevent the departure of a specialist and identify the reason for dismissal.

Note. The table shows structured data, that is, predictive analytics almost always begins and ends with descriptive analytics. Therefore, do not neglect collecting traditional statistical data. They will definitely come in handy if you decide to develop predictive models.

5 steps to develop a predictive model

In general, the progress of work will be as follows (using the example of the forecast of dismissals of qualified specialists at their own request):

1. Preparing Assumptions, why highly qualified workers quit of their own free will. Study:
   • External factors: demand for competencies, dynamics of salary changes, employer’s position in the labor market.
   • Internal factors: working hours (schedule, overtime), presence/absence of business trips; level of comfort in the office; salary level; career growth, development prospects; travel time to work.
   • Individual characteristics: gender, age, marital status, presence of children; psychological profile; knowledge of foreign languages.
2. Determining available data. They decide where to find the information necessary to test the hypotheses formulated in the previous stage. Possible sources:
   1. External data from recruiting sites, salary reviews.
   2. Internal information of the access control and management system (lateness, breaks, early departures, changes in established behavior); data about visited sites; results of engagement surveys and performance assessments; personnel records data (salary and career changes)
   3. Results of psychodiagnostics.
3. Collection, verification and structuring of information. As a rule, the longest stage.
4. Model development: 1) creation, 2) testing, 3) visualization. They create and test a prototype model based on real data, develop a user interface, and train HR managers to interact with the system.
5. Application of the model. Ideally, not just forecasting, but monitoring (including remote) indicators of the risk of employee dismissal.

Let’s summarize and think about how to apply the article’s material in domestic HR realities.

They write and talk about HR analytics much more than they understand what it really is and what problems it is designed to solve.

This article is about what the main difficulties in application are and where you can start analyzing.

So far, in the public space, HR analytics means either metrics and working with numbers, statistics in HR, which are indicators of local processes or departments, or sharing cases on how to involve the IT department in the creation of HR data archives. In fact, HR analytics solves the problems of strategic development of a company and determines the main predictive trends.

So, what is HR analytics?

HR analytics is a process in which data science and business analytics (BA) techniques are applied to the processing of HR data. It is sometimes also called talent analytics. Additionally, data mining in this context refers to the practice of mining databases to create new information.

Why is this so relevant now, in the light of the global digitalization of the economy, business, and people. Data, friends! Big data is everywhere!

And then neuroscience arrived, showing us how human decisions are subjective and filled with emotions instead of a rational approach.

How to live with this now?

Of course, we need a different way of making decisions - so that everything is logical, reasonable, based on data and with a guaranteed result. Who doesn't want that? Everybody wants! Why don't they do it?

HR analytics readiness remains a major challenge. According to a 2017 Deloitte report, after years of discussing the issue, only 8% of respondents reported that they had useful data; only 9% believe they have a good understanding of what employee characteristics lead to success in their organizations; and only 15% overall have deployed HR and talent performance systems for line managers.

“It’s been a mystery for the last decade – why, given the obvious importance of human capital, are organizations not investing in it and requiring leaders to make their decisions about people using evidence-based analytics?” ("")

So what's the deal? Why do they talk and write about the benefits and necessity of using analytics much more than they use in real work?

Let's look at this phenomenon from different angles.

What's the difficulty?

First of all, it is worth considering the deep, fundamental reasons for such “inhibition” on the part of company leaders. I will describe this in terms of a model that Pete Ramstad and John Boudreau presented in Beyond HR (Boudreau and Ramstad, 2007) called the LAMP model. logic, analytics, measures and process).

To simplify what is described in this model, the reasons why the use of analytics is inhibited are the following:

• Logics: we cannot explain why high-performance work systems work. It's still a black box. We understand that there is a certain relationship between the nuances, but we certainly cannot say what directly depends on what and what needs to be done with X to get Y.

• Analytics: there has traditionally been a lack of depth and rigor in analytical models. Leaders at Google and other leading companies are turning to industries like rocketry, where there are models that take into account a huge number of factors. Simply put, this is not methodologically sound.

• Metrics: Most often, data sets relate to current employment status, employee costs and HR programs. At best, this data represents operational or advanced reporting rather than strategic or predictive analytics, which includes analytics, employee segmentation, and which is tightly integrated with strategic planning.

• Process: it is the presentation of analytics to decision makers. Here, the main success factors are the timeliness and degree of visual appeal of the presented data. We are talking about obtaining data in real time in a form that is accessible and understandable for decision-making, and such tools using artificial intelligence are just being developed. For example, most managers have no idea how to interpret the employee turnover rate because they usually know that low turnover is not always beneficial, and conversely, they do not know how to determine what is best in the situation they are faced with. From this point of view, we are at the stage of reviewing HR tools.

I think from what has been described above the complexity and depth of the problem seems a little more clear. So, there are objective reasons why investing in analytics seems to be quite risky. Roughly speaking, we do not have clear, reliable, unambiguous tools for making analytics-based decisions. More precisely, for very simple local areas there are, but they are not worth the cost. Spending this level makes sense if we can get reliable predictive trends that are key to business success. But analytics alone cannot guarantee this.

We don't want to just process data. We want to have reliable tools for making business decisions with more or less guaranteed results. And in this sense, the main thing still remains with the person:

• ability to ask strategically relevant questions and present them in a logical framework that shows the relationship between HR investments and critical organizational outcomes;

• Possessing in-depth knowledge of your business;

• understanding the logic of analytical models in the sense of their applicability to explain vital processes in an organization and much more.

To sum it up in very simple words, the main difficulty of analytics as a way of working with data is that we first need to determine what results we want to get. And to do this, we need to ask very correct questions that require a deep understanding of the business, then determine with the help of which analytical model we can get these results, in accordance with this, determine what data and in what volume we need, and only then come up with, How can we get them exactly in the form we need?

A complex approach

To illustrate the complexity of the approach, look at the picture showing the composition of the HR analytics team:

That's not all. It is very important to remember that in general the phrase “HR analytics” is extremely rare today in the works of researchers and authors. This is such a familiar Russian-language term. In English, the concept of People Analytics is now used - people analytics. This is not a simple synonym. Vice versa. From local areas related exclusively to HR - turnover, recruiting metrics, employment status, etc., the West has moved to global “people analytics” or “human analytics”. All data about people matters - their movements, health status, activity on social networks, etc. Only by using the full amount of data can we talk about an acceptable degree of reliability of forecasts and strategic decisions. To collect such data, companies must implement new tools based on mobile applications and more, and attract specialists who could work with it.

But this is far from the end of the problems, this is only their beginning.

Context. Combining big and dense data

Context is critical. What does it mean? This means that in addition to big data, we need the so-called. dense data: this is all that valuable information from people - stories, emotions, communications - that cannot be quantified, but carries meaning of incredible depth. What makes them profound is the experience of correctly perceiving what people say - this is what helps to recognize gaps and holes in predictive models. Dense data immerses business issues into human issues – providing context. Therefore, combining large and dense data gives a deeper picture. You work with both collected and uncollected data: this gives you the opportunity to ask the right questions “why?” Why is this happening?

To illustrate the importance of context, I will give two examples: negative and positive.

Negative example- this is the story of Nokia, which has already become a sad example of how you can fly out of the market at the peak of its form. The essence of the main strategic miscalculation was that the company's leaders ignored dense data, which could not be compared in array with big data, but absolutely accurately predicted the huge interest in smartphones even among the poorest segments of the population.

A positive example also in plain sight. This is fantastic growth for Netflix. There, on the contrary, they saw holes in the analytical models and invited a technological ethnographer (there is already such a specialization) to work with dense data. And he figured out something that wasn't visible in the big data. The ethnographer noticed that people like to “get stuck” in front of the TV; they don’t feel guilty about it, but simply enjoy it. And by combining big data with dense data, they did something simple but effective: instead of showing different genres of series, they started showing the same ones to make it easier for people to get stuck. But that was not all, they changed the very practice of broadcasting in accordance with these findings. By bringing together big, dense data, they not only improved their business, but also changed the way people consume media information. Their shares are expected to double in the next few years.

Data is nothing. Context is everything!

Resources

We are gradually moving forward in considering our problem, and if you are still with us, the last bastion is ahead.

These are resources. As can be seen from everything described above, serious work with data requires “heavy” and expensive software, highly qualified specialists and a lot of time. All this adds up to costs that are virtually unaffordable for most organizations. If you follow the topic, you may have noticed that most of the published cases are cases of huge companies describing global research. In this case, you need to remember about the so-called. survivor's mistake.

The published cases are mostly those in which it worked. And how many of those in which it didn’t work out at the same cost? There are no low-cost and relatively simple tools and models yet. But the market is a market and most likely, after some time they will appear as a result of accumulated experience. Therefore, large companies are now trying, and everyone else is waiting for something more accessible to appear as a result of the activities of the first ones.

These are, in fact, the main reasons why only 8% of respondents reported that they had useful data; only 9% believe they have a good understanding of what employee characteristics lead to success in their organizations; and only 15% overall have deployed HR and talent performance systems for line managers.

But the need and benefits of working with data are obvious and cannot be discussed. So what to do?

Where can companies start?

People analytics is a large-scale direction due to the global nature of the tasks being solved and is quite new. However, among analytical approaches there are already sections that have been well developed for a long time. They provide powerful yet accessible tools and can provide significant insights to a company. One such approach is organizational network analysis(ONA, Organizational Network Analysis). What it is?

The purpose of ONA is to measure and display relationships and flows between people, groups or organizations. What makes ONA unique is that there is no other way to see the real connections between people in an organization. It is effectively an X-ray of your organization, or your organization's relationship with the external market, or your workforce, or your candidate pool. In short, those relationships that you need can be analyzed.

ONA emerged at the intersection of sociometry and network analysis and appears to be an extremely useful tool.

A huge advantage of this approach is its visuality.

For example: an analysis of managers in the exploration and production division of a large oil company revealed the following difference between the formal and actual organizational structure (Figure from Rob Cross's blog):

From the right picture it can be seen that the company has one of the middle managers, a certain Cole (see left picture), who is almost invisible in the official hierarchy, but in fact it is through him that all the flows of information and the actual distribution of work go. He is the main information hub and he decides who gets what information. The vice president is located on a very distant periphery and, in fact, has no influence on operational management.

I think you have already begun to guess what role such a scheme can play, for example, in change management.

The next big application for ONA is, of course, knowledge management. If at the input you ask questions like: “Who is the coolest expert at work?”, then the picture at the output will show the main bearers of expertise in the organization.

How can we not talk about the task of creating an information field in a company? Any communications manager simply must have this kind of analysis if he does not want to move forward blindly. Such an analysis can show both relationships and information flows between departments, between the company and other stakeholders, and between people. In our training course “HR Unvarnished” we touch on this topic in more detail.

For example, how does the interaction between finance and marketing actually occur in your company? Through whom does all information go? (Fig. from Rob Cross's blog)?

The same goes for any innovation, leadership, talent development, etc.

We looked at the prospects for using ONA within an organization, but with equal success this tool can be used to analyze external relations - with competitors, suppliers and contractors, etc.

Main applications of ONA

ONA is the art of getting useful results: you get maps and metrics that lead you to really good questions. That is, ONA, like any analytical tool, does not answer the question “Why?” This answer can only be given by a person. But cards do two things:

• They provide indicators of where there might be something interesting to explore.

• They provide interesting visual results to support outcome stories.

Of course, in reality this is not as simple as it seems at first glance. Behind all this inspiring beauty and apparent simplicity there is serious mathematics and fundamental research, but it is much simpler than what is in “big analytics” today. ONA will give you extremely useful results right away and save resources.

Victoria Buznik And Liliya Grabovskaya, authors of the resource Talent Management.com.ua and the training course “HR without embellishment”

Please tell me where should I go from here?
-Where do you want to go? - answered the Cat.
“I don’t care...” said Alice.
“Then it doesn’t matter where you go,” said the Cat.
“...just to get somewhere,” Alice explained.
“You’ll definitely end up somewhere,” said the Cat. - You just need to walk long enough.
Lewis Carroll "Alice's Adventures in Wonderland. Alice in the Wonderland"

Have you come across useless HR metrics from the “historical” category? Let's discard everything unnecessary and check what is vital: the site told the portal about the most important indicators of HR analytics Dmitry Supronenko , Head of the HR Department at Metal Profile Company.

Analytics in the field of personnel management as an independent and quite important block - on the one hand, the topic is quite new for domestic companies. For myself, I link its formation to the period of formation of the HR function in its modern form from the HR department, directly subordinate to the CEO of the company, and O&P as part of the financial and economic block.

On the other hand, this period turned out to be more than enough for some companies to take not just a step, but a whole leap forward, while others continued to remain in a state of suspended animation. As a result, even within the same industry we can observe such a significant difference in the approaches and level of development of HR analytics that in the digital age we can only shrug our shoulders.

But since nothing happens for nothing (both action and inaction), let’s consistently understand the reasons for different approaches. I divide them for myself into objective (they exist independently of the HRD company) and subjective.

Let's start with the objective ones. Firstly, from the most important factor, which, in my personal opinion, most directly affects the company’s HR analytics and indirectly influences other factors, this is level of competition in the industry.

Let me explain, for me, as an economist with basic education, a highly competitive market is a market in which a product/service, production technology, logistics, availability of raw materials, etc. are maximally unified, and the buyer makes virtually no difference from whom to purchase this product/service. In these conditions, the level of service comes first, which is why many strategic gurus have already dubbed the economy of the 21st century the economy of impressions. And the key strategic advantage in these conditions is the personnel, or rather, their quality. It is logical that competition for strong employees in such highly competitive markets is also significant.

HR specialists simply have no other choice but to use all possible analytics apparatus in these conditions in order to quickly find/entice the best specialists from the market and retain them with the required level of involvement for the longest possible time.

Therefore, everything starts with analytics on the “recruitment funnel”, ends with analytics on the reasons for staff turnover, and between them there is a whole world of indicators on adaptation, motivation, training and development, involvement, corporate culture, etc. For example, we can take the market for IT specialists. Despite the fact that real wages in many industries have been declining in recent years, here we see steady and stable growth.

The time it takes to find a new job for an IT specialist is minimal; the level of their aspirations in such conditions is constantly growing. And how do companies react? Individual approach. Starting from a dedicated one or several (depending on the scale) recruiters, and ending with the fact that HRD knows highly qualified specialists by sight (even if the number of the company itself is 10,000+), because in terms of the volume and frequency of changes in standard conditions and approaches to personnel, they come out on the same level as the company's top management. But this is for highly competitive industries. If the industry is monopolistic/oligopolistic in nature, then all this “tuning” is not necessary.

I’m not saying that they don’t use HR analytics, no. But its content will be more meager, or (which in my experience is more common) its use in work for decision-making will be formal. I myself worked in a company where Moscow regularly requested a significant list of metrics developed by a well-known foreign consulting company. The communication scheme has always been typical: requested - provided - forgotten.

Secondly, industry affiliation (specifics). Let me explain with an example. Several years ago I was interviewed for the position of HRD of one of the largest agricultural equipment manufacturers in Russia. The general director approached the matter more than responsibly (I borrowed his experience in my work), not limiting myself to the results of assessments and several interviews, but provided free access to the company’s facilities, as well as to all HR information, and requested an action program for 2 years. And although at the end of everything I accepted an invitation from another employer, the experience of this company is still interesting for me to this day.

A key feature in HR’s area of ​​responsibility is pronounced seasonality in relation to the agricultural year. Seasonality will surprise few people, but here you need to take into account the scale (every year you need to recruit, train, and then lay off more than a thousand production workers), the high level of requirements for the qualifications of workers (this is not hiring warehouse workers), location (all production divisions are located in the same territorial area of ​​the city), frequency of repetition of the cycle (at that time it was already the 5th hiring/downsizing cycle) and all the ensuing consequences for the HR brand (the first taxi driver on the way from the airport told me the whole history of the plant since the change of ownership of the company, and that you should come here to work as a last resort).

It is clear that the key HR department at the enterprise was not the occupational health and safety department (as usual), but the personnel selection department, which was supposed to ensure the selection of such a number of employees within a month. I was surprised by 2 facts: all the employees in this department were men (which is absolutely not typical for the HR function in general, and for selection in particular), and the quality of HR analytics.

It seems to me that they monitored everything on an ongoing basis. I’m not even talking about the “selection funnel”, it was compiled for each workshop by position, the effectiveness of attracting candidates through all external and internal channels known to me was compared, a separate place was devoted to the analysis of the candidates’ areas of residence, and the entire adaptation period was divided into blocks, according to internal statistics of staff turnover in the first three months, and throughout the entire adaptation chain the same funnel with staff turnover during the probationary period.

Third, stage of company development. Here it is necessary to recall the famous model of the life cycle of an organization by L. Greiner. I see no point in disclosing it in detail in the article - material on the topic is easily available on the Internet.

I will focus on the fact that very often supporters and opponents of implementing a KPI/BSC system in a company (and analytics in HR is an element of the KPI system, since without planning, control, motivation, in itself it does not bring value to the company) cannot agree among themselves , since the basis for their dispute are organizations that are at different stages of development according to Greiner. And if at the “Creativity” stage the analytics is, rather, a rare exception to the rule; at the “Directive Management” stage, HR analytics is simplified (2-3 general indicators), then at the “Delegation” stage there is a significant quantitative and qualitative development of indicators , and at the “Coordination” stage, analytics begins to become redundant (when the transaction costs of conducting HR analytics exceed its economic effect for the organization). And here the opponents of KPI rise to the podium and begin to criticize.

And they are partially right, but we must immediately clarify that most domestic companies have not “grown up” to stage 4 and are unlikely to ever grow up. And only a lucky few (leaders of the RBC 500 list) seriously think, when faced with a crisis of confidence, about the need to move to the fifth stage: “Collaboration.”

According to L. Greiner, special attention at this stage is paid to the creation of teams and interpersonal cooperation, and formal control systems are gradually replaced by social control and self-discipline. Each stage of an organization’s development has its own approach to the system of HR metrics, and it is unacceptable to simplify the approach by reducing it to uniform requirements, just as it is unprofessional to compare HR analytics of organizations at different stages of development.

Fourthly, level of automation in the company. Here, in my opinion, everything is quite simple: you cannot carry out a large-scale offensive if your convoys with ammunition and uniforms are far behind. HR analytics that are collected “manually” is a sure way to discredit an excellent tool. Therefore, we are friends with the IT department; ideally, you need to have a dedicated specialist for analytics in the ERP system in which your statistics are kept.

Depending on the scale of the company, a block of tasks may be added to integrate data from several HR systems. From my own experience, I can say that setting up integration takes the most effort, as a result of which some companies made the logical (but not cheap) decision to transfer all HR functionality to a single platform.

Now for some subjective factors.

Firstly, presence of a customer: The HR department should not exist on its own. Each task/project should always have a beneficiary: the owner, the Board of Directors, the Management Board, the CEO, the director of a business unit, the head of a functional unit, or someone else, no matter what the name of the position/body is. If the task/project does not have this beneficiary (and this, unfortunately, happens), then the HR manager’s chances of being known as “the cause of happiness” in the company increase sharply. This does not mean that if you have not been assigned a task, then you do not need to do it. It is necessary if you understand why this may be of interest to the beneficiary, but the first thing you need to do is obtain his consent.

Let me explain in practice. Let’s take, for example, a manager who headed the HR department in several large companies, where one of the key indicators in the area of ​​HRD’s attention is staffing, both as a total value for the company and in various analytical sections (by business units, by functional blocks and other).

Firstly, the entire budget (and the maximum share in it is the payroll) in the HRD area of ​​responsibility directly depends on the planned number of personnel in a specific period (usually a quarter, for companies with highly seasonal sales/production - a month). Secondly, other key indicators (for example, staffing levels, the proportion of overdue vacancies) depend significantly on staffing levels.

We are moving this manager to an equally large company. Where will he start? Most likely, one of the first points will be an audit of current HR processes (including collecting analytics in a familiar coordinate system). And here it turns out that the company’s share of vacancies is, say, 30% of the total staff, and applications for personnel selection are no more than 10% of the number of vacancies. What does this mean?

Most likely, the problem is not in the selection, but the staffing table is overloaded with “dead” vacancies. If the duration of these vacancies is significant, then taking into account the constantly carried out measures to optimize the number of personnel, a reflexive decision (postponed at the subcortical level) would be to reduce overdue vacancies or, at a minimum, clear out those for which there is not even a recruitment application, go to the basic worker level in terms of numbers, and then think about what to do with it all. Decided - done. And only a few months later, with the start of the season, business units make a fuss: they need to quickly hire a sufficiently large number of workers, but they cannot do this, since there are not so many vacancies in the staffing table.

And it turns out that staffing is a technical indicator, vacancies are needed only for the registration of workers by the HR department, and resource planning is built solely on the basis of the payroll indicator, which is used by the branch director when deciding whether to hire an employee or not, since he is generally responsible for branch costs. At this stage, it doesn’t matter to me whether the process is built correctly or not, something else is important: HR department analytics cannot exist on its own, separately from the business. It is always secondary, formed for specific tasks. Then it won't be redundant.

Secondly, a significant influence on domestic analytics in the field of personnel management is played by a factor that I call for myself: “historically.” You often come across a situation where your question “why?” (not to be confused with “why?”) you regularly collect this HR analytics, you get an honest answer: “we’re used to it,” “we’ve always done it this way,” “it’s so convenient for us” or something similar.

Once upon a time, someone from the company’s management, in order to solve a specific problem, requested monitoring of a separate indicator/s for control. Since then, the situation could have changed significantly, the task could have become unnecessary altogether, but since no one gave the “hang up” command in a timely manner, the HR service continues, by inertia, to collect information and send it according to the approved list of recipients.

In my practice, I have seen that after several years (years!) of such reports, the recipients find out from each other who the initiator was and why they need this information. And the higher the status of the person who originally set the task and the level of authoritarianism of his leadership, the greater the likelihood of getting a line of HR metrics from the “historically established” category.

Now, after the theory has been completed, and everyone has been informed about the possible consequences and nuances, you can move on to driving, namely the list of HR analytics indicators. I would like to note right away that this list was formed only from those indicators with which I worked, and for the conditions in which I worked at that time; moreover, this list was not even 80% used in any of the companies (who did not understands why, I suggest reading the article again).

I. Finance:

Share of personnel costs in revenue (expenses take into account all items in HRD’s area of ​​responsibility);

The ratio of the growth rate of revenue/margin/net profit per employee to the growth rate of payroll/personnel costs for the same period (fundamentally I do not compare with the growth rate of the average salary, since its dynamics are significantly influenced by the number of personnel, which can lead to a distortion of the economic essence of the indicator);

The level of financial risks for the HR function (can be measured as the number of risk events, which means any failure in the HR process, which resulted in damage to the organization exceeding a certain amount, or the total amount of damage for all risk events);

Company expenses for 1% staff turnover (all direct costs are taken into account: selection, training, adaptation of personnel (distraction of the mentor and manager), salary until the employee reaches the target level of productivity and lost income during the period of absence of the employee and until the employee reaches the target level productivity);

Average wages of employees (by position/function/location) relative to the median salary based on a review of the labor market in similar analytics.

II. Clients (internal):

The level of satisfaction of internal clients with HR services (measured both as a weighted average and separately by function. As a rule, subcontractors are most interested in selection, motivation, headcount management, assessment and development);

The share of vacancies filled within the standard period (standard terms are differentiated depending on the level of the position, functional block, location, etc.);

Staffing as a percentage of the staffing level (the indicator is an alternative to the previous indicator, was used in conditions of mass selection, unlike the previous one, it is momentary and not interval, therefore it is especially inconvenient for selection when the weighted average staffing is calculated by analogy with the SSC);

Average performance of employees at the stage of entering a position (often used for sales managers, the time interval is set differentiated for the position, as a rule, it is linked to the period of training of the newcomer and admission to independent work, can be measured in percentages, conditional products, etc.).

III. Processes:

The cost of recruiting 1 employee (for me it was divided depending on the level of position/profession into 4 categories);

- recruiting “funnels” in various analytical sections (positions, locations, functions, channels for attracting candidates, etc.);

Turnover of employees during the probationary period (an extremely useful indicator for building relationships with an internal client, especially in terms of function and location, for this it is necessary to divide it into two components: on the initiative of the employee and on the initiative of the manager);

Attrition/staff turnover (the division between indicators is conditional; by attrition I mean an indicator when all dismissed employees, regardless of the reason for dismissal, are taken into account; if any reasons are not included in the calculation of the indicator, then for me this is turnover. Fundamentally important in the second case, conduct exit surveys, collect telephone numbers of dismissed employees and make audit calls on dismissed employees;

Traditional analytics for reasons of staff departure/turnover;

Departure/turnover of key personnel (counted for employees who received grades A (high potential) and B (high performance) at the last personnel commission, as well as those included in the personnel reserve for a higher position according to approved career routes with the readiness level RN (ready now));

Analytics on the reasons for the departure/turnover of key personnel (see previous paragraph);

The average percentage of employee absence relative to the FRF according to the staffing table (includes all types of absences);

The share of commercial service employees who have successfully completed sales/product training by location/function;

Level (percentage) of compliance with the company's Sales Standard according to approved checklists (analytics are collected based on the results of both internal and external audits - calling clients, auditing calls, etc.)

Average percentage of completion of tasks for IPR within the required time frame;

Average satisfaction score with face-to-face training based on feedback questionnaires.

IV. Staff development:

The share of key positions in the company (determined individually) filled based on the results of the last personnel commission by successors with the readiness level RN (ready now) and RN+1 (ready in a year);

HR analytics includes various methods of analyzing and processing data necessary to increase the efficiency of services. More details on the topic can be found in the article.

From the article you will learn:

What is HR analytics

HR analytics is a process in which information processing and business analytics (BA) techniques are used to evaluate HR data. It is often called talent analytics. Data mining refers to the practice of mining databases to create new information. There are several main goals: providing insights and identifying key parameters.

Goal Features:

1. The first purpose is to provide information about the company's own operations that helps manage employees. Insights can ensure effective achievement of goals.
2. The second key function: helps to identify information, determine the optimal performance indicators of the HR service that the company should maintain. It provides models for predicting the ways in which a company can achieve a high return on investment (ROI) in human capital.

The objectives of HR analytics are to make the most efficient use of volumes of data about human resources, collected by most companies. Firms often have data such as employee demographics, educational records, and so on. A detailed analysis allows us to extract important knowledge from them that will help in further work.

Download documents on the topic:

Recently, a number of companies have been using big data in HR, when a large amount of information is structured and placed in tables. In this case, it is necessary to use special programs that allow calculations and analysis. It is rational to apply the methods if the organization employs more than 1000 people.

Why determine HR indicators

Personnel decisions are often based on professional instincts, as well as intuition, which must be taken into account when determining HR indicators. Hiring often depends on the contact that the recruiter was able to establish with the candidate for the position. The problem with instincts and intuition is that bad practices can take root. For example, injustice may go unnoticed. A clear example is the wage gap between women and men. Managers may think they are paying the same unless they look at the exact data.

HR analytics helps improve labor productivity,predict successful models. This eliminates some of the errors in decision making. For example, workload management will be more effective when data is used to show which departments or groups are already carrying excessive workload, and which are only allowing themselves to take on more responsibilities.

Experts have proven that analytics contributes to the rapid growth of an organization. Research conducted by MIT and IBM has shown that a high level of use of HR analytics can result in: increased sales, net operating income, and sales per employee.

Main areas of application of HR analytics

The areas of analytics are vast, and the indicators that an organization needs to focus on depend directly on the industry and the nature of the business.

Here are some examples of key HR indicators:

1. dismissal rate in the organization;
2. time of hiring workers;
3. staff turnover rates for different groups;
4. income per specialist.

The given indicators and other similar data are used to improve business efficiency. The key areas where information can help include:

Recruiting- analytics provide answers to questions about finding ideal candidates for open vacancies. For example, the information is used to identify the qualities of candidates that produce the best results. In addition, you can compare the data of applicants who remain in the organization and find the same denominators among them.

Health and Safety- HR analytics allows you to identify problem areas. The data points to roles, as well as places of work and other similar factors, that have high rates of accidents.

Employee Retention- thanks to the information, you can learn more about retaining specialists who are valuable to the organization. You can use HR analytics to identify aspects that increase engagement personnel.

Talent Gaps- information helps to identify gaps in the company. For example, some departments have more qualified employees than others, and this can greatly hinder the overall performance of the organization.

Sales efficiency- this area of ​​HR analytics helps to understand the details of how to improve performance. You can see that certain talent helps you perform better, and programs give immediate results.

Analysis of HR performance indicators

Every organization should conduct an analysis of HR performance indicators from time to time in order to promptly identify existing problems. Especially often it needs to be carried out during a crisis, when it is necessary to cut costs with minimal impact on the team, or rather its loyalty and motivation.

Indicator No. 1.

Indicator No. 1. Maintaining budget for employees

HR analysts always use this indicator in organizations where work is assessed using KPI. It is calculated as follows:

Sat = Zf: Salary × 100%, where:

• Sat - main budget compliance indicator;
• Zf - actual costs for company employees;
• Salary - planned costs for specialists.

If it is noted that the budget has been exceeded, still reduce labor costs last. But, for example, reduce the cost of the social package more boldly. Approach the costs of personnel training in a differentiated manner, using HR analytics methods. Highlight those that are undesirable to cut and those that can be sacrificed. Let's say, you need to plan funds for “Increasing your professional qualification level.” And fund general theoretical seminars when such training is really necessary.

Indicator No. 2.

Current financial return on employee expenses

Ei = Op: Zf, where:

• Ei is an indicator of the effectiveness of investments in employees;
• Op - production volume, calculated in monetary terms;
• Zf - actual financial costs for employees.

You will see exactly how many products the company produces for each ruble spent on personnel. The formula is easy to adjust, taking into account the specifics of the company's industry. If in a company HR affects production costs as well as sales, it is better to set a KPI that is tied to gross profit. In this case, HR analytics will be tied to simple calculations. The calculation formula is as follows:

Ei = (Pv - Zf) : Zf, where:

• Ei is the main indicator of the effectiveness of investments in employees;
• Pv - gross profit;
• Zf - actual costs for employees.

Indicator No. 3.

Labor productivity

This parameter shows exactly how much production is produced per employee. The higher the score, the better. It is calculated like this:

Pt = Op: Tz, where:

• Pt - labor productivity of workers;
• Op - production volume;
• TK - labor costs

For ease of calculation, replace labor costs the average number of employees, assuming that they all worked the same number of days in a particular period. An HR analyst must take into account all the nuances, enter data into reporting documents in order to subsequently calculate the parameters.

Indicator No. 4.

Staff turnover

Personnel turnover allows you to evaluate the work of the entire service and the HR director in particular. The formula for calculating fluidity looks like this:

Tk = Su: Co × 100%, where:

• Tk - exact indicator turnover;
• Su - the number of dismissed workers;
• Co - the total average number of employees.

Monitor whether replacement costs exceed specialist the price you pay to keep him. If they exceed, then turnover among unique specialists can result in increased recruitment costs for the company. How to avoid this? Identify the risk group among employees. As a rule, these are key managers of the company. Now turn uncontrollable turnover into manageable turnover.

In the risk group, identify the most disloyal people who may leave the company. Find a replacement for them in advance. Analyze the labor market and invite specialists of similar skill levels for interviews. Finally, ask if the candidate will be willing to take the position when it becomes available.

Indicator No. 5.

Absenteeism rate

Absenteeism is the absence of specialists on site during working hours. It doesn’t matter if it’s for a good reason or because of absenteeism, but the analyst must count them. If the indicator is high, the staff does not strive to perform their duties efficiently. It is calculated as follows:

Ap = Dor: Up × 100%, where:

• Ap - indicator among personnel;
• Dor - number of days of absence;
• Until - the total number of days in the period.

Indicator No. 6.

People's job satisfaction

Measure satisfaction with a survey. The formula is:

Lp = Cl: Co × 100%, where:

• Lp is an indicator of employee loyalty;
• Sl - number of loyal specialists;
• Co - the total number of respondents.

HR-environment conference dedicated to working with personnel, employee development and solving problems in the field of HR. Anton Lukyanov, head of the Yandex HR analytics group, spoke about the basics of working with data in HR and shared his experiences.

What is a "graph"?

All industries generate data. Thanks to data, we can better understand the internal client, optimize processes, and take a different look at employee interactions. These processes can be described using graph theory.

A graph is an abstract mathematical object that consists of vertices (points) and edges (lines) that connect them. As an example, you can take the Internet - many sites are connected by links. The basis of Yandex's business is to quickly crawl the Internet graph, analyze the content and provide a relevant answer to the user. Another example is Yandex.Taxi: by finding the shortest distance between points on the map, the service helps the passenger get to their destination.

Yandex company graph: dots - employees, lines - interaction between them

An abstract representation of objects in the form of graphs was proposed by the mathematician Leonhard Euler, who solved a popular problem in the 18th century: how to walk across all the city bridges of Königsberg without passing over any of them twice?

Data sources and requirements

In order to build a graph, you need data. An organization can use the following as data sources:

corporate mail

employee meeting calendar

tasks in the tracker

internal PBX calls

Trackers are especially popular in IT companies. They allow you to set tasks, assign those responsible for their execution, and attach files. Trackers greatly simplify collaboration within teams. In Yandex, this tool is used not only by developers, but also by other departments.

HR departments use the tracker for:

coordination of vacancies

hiring

adaptation

training

HR analytics is built on the basis of this tool.

Data requirements:

Quality. The data must be free of errors and omissions. The indicators must be brought to a general form.

Completeness. Nowadays, working in several information systems simultaneously has become the norm. Data taken from only one system will not be complete.

Interaction Analysis

The dots and lines on the graph represent employees and their interactions. The larger the diameter of the dot, the more intense the employee’s interaction with colleagues. The thicker the line connecting two employees, the more intense the interaction occurs in this pair.

Yandex employee interactions

It is important to take into account and correctly interpret all data. The intensity of employee interactions directly depends on their position and assigned tasks.

What conclusions can be drawn and what can be applied in organizational changes:

Effectiveness of interactions. If groups differ in the intensity of interactions and they have an objective indicator like KPI or a performance review is conducted, then it is possible to compare factors from the graph (intensity of interactions) and predict which actions lead to better performance. For example, an excessive amount of communication can overload managers and affect their burnout.

Management style of managers. The data will tell you whether a leader is using micromanagement or total control. Using data from the graph, you can tell the manager whether his management style is suitable for each specific task and what actions can improve employee efficiency.

Communications within the team. For example, team members interact well with each other, but have virtually no interaction with other colleagues. Such a team can make a project that was already in the company, data about it is stored, but the team does not know about it. As a result, company resources will be wasted due to insufficient communication with other departments. Or vice versa, when internal communications in a team are poor, this leads to a delay in project preparation deadlines.

A complete picture of employee interactions allows you to automatically create a list of colleagues for conducting a 360-degree survey. In this case, the data that the manager will receive based on the survey results will be complete.

Collaboration between teams

Metrics

Some metrics from graph theory can be successfully used in organizations:

Density/Sparseness. A graph is called complete when all its vertices are connected by edges. In the real world, an example of a complete graph could be a small startup: one small friendly team in which everyone knows each other and interacts effectively. The departure of one employee does not lead to the loss of essential information, and interaction is not disrupted.
As a company grows, density breaks down. It's hard to imagine a company with thousands of employees where all the employees know each other. Such a company is characterized by the risk of losing significant communications within the team when one employee leaves.

Distance. This metric is reminiscent of the famous theory of six handshakes. The smaller the distance between employees, the greater the cohesion, the more people know each other.

Centrality. This metric is well illustrated by the Game of Thrones character graph: after the unexpected death of one of the central characters in the work, the writers have difficulty with the storyline of the secondary characters.

Bridge. This can be illustrated by the example of Belgium, a country with two official languages, where a small bridge of people speaking two languages ​​connects “monolingual” fellow citizens. In a company, people on this bridge are carriers of very important connections. Their departure is a big problem for the company.

Application

In project-based companies, projects are started and closed on a weekly basis, it is important to keep track of expenses. It is necessary to correctly collect and transmit data for economic calculations. This process can be improved through automation. Yandex did the following:

We built a company graph.

By analyzing an employee’s behavior in each specific month, HR analysts see what project the employee is currently in. There is no need to correspond with service managers. This saves business and analysts time.

So far this is a pilot project that is showing interesting results. Many research projects in Yandex are subsequently rolled out into production.

So, the classic path of analytics

Data. Source, completeness and quality.

Visualization. How to look at this data and how to use it. Ideas about metrics, which in the case of HR analytics can be taken from graph theory or invented independently.

Reporting, which shows monthly what is happening in the company. For example, the intensity of interaction between managers.

Metrics, which are used for inferences.

Predictive Analytics- launching an automated process that saves time.

What else you need to know about HR analytics

Kevin Wheeler, President and Founder of Global Learning Resources, Inc., in an article“The Downside of HR Analytics: 8 Little-Known Facts” highlights several important factors that relate to HR data:

Analytics is not a magic pill. Analytics is not a miracle cure. Data can help you understand a problem and perhaps choose a more effective way to solve it, but data is no substitute for empathy and human reasoning.

Understanding what exactly you want to know. You need to be extremely clear about what you want to analyze or measure. And make sure it's even possible.

Use the appropriate method. The method of collecting data can also be challenging. One use of analytics is to clarify a problem or look for possible causes.

Passive data may be better than requested data. It is much easier to collect passive data yourself than to request hard data from others. Collecting factual information based on the results of any actions and decisions is relatively simple.

Support is important. Effective use of data requires leadership support and a culture that values ​​data.

The goal controls the situation. There is a great temptation to measure everything, especially at first, when an analytics tool is just being implemented in a company. But it's better to focus on two or three key questions that you want answered. Then you will have enough time for more accurate data collection and full analysis.

The data is flawed. Recently, it has become common to put data on a pedestal and perceive it as pure information, in which there are no politics or opinions. But unfortunately, data analysis is influenced by opinions just like everything else.

The simpler the better. Take the time to make a list of what you would really like to know to improve your recruiting process; what data will help you improve the effectiveness of most sources or answer pressing management questions.

The problems of correct interpretation of data have also recently been discussedNassim Taleb said : “If you know how to work with Big Data, that’s good, but you need to be able to interpret it, filter out the nonsense and unnecessary noise that confuses everything. Interestingly, the only ones who know how to work with data are anti-terrorist services. They are able to avoid finding false correlations and narrow down the sample to specific suspects; they are looking for connections.You can simply fool a computer with data; statistical discipline is very important. Big data cannot tell us what is right, only what is NOT right.”

Image Source - HR Environment

Current and interesting HR cases in our Telegram. Subscribe to the channel!

Copying and any processing of materials from the site is prohibited