Why do we need business intelligence systems

Visualization in business analytics is very important. For example, the concept of these graphs is drawn together with Infographer.

The problem with specialized software for business analytics is that it stands as a plane - and is only needed when you have a large, very large company with a corresponding amount of data at your side. In general, there are not so many specialists on the market now, who have tested such systems, have experience working with them and can plainly say what it is, why it is necessary and what it will give after implementation.

The first and foremost application of business intelligence is stupidly in the forehead to be able to build arbitrary reports while the heads of departments or members of the board of directors confer. One of my favorite examples is from which factory to supply vodka to stores: from a distant (expensive logistics and 2 weeks on the road), but cheap at the cost of production or from the closest (1 day), but expensive?
')
It is clear that you can dig deeper and respond in a week. But more often you need a faster simulation and not hundreds of disparate reports, but a unified environment where a financial specialist can simply take and obtain the necessary data without involving the entire IT department. There are also specific tasks that solve these and related systems. I'll tell you now.

More typical tasks

Generally, right away: we are not talking about specific solutions, but about what a company can get from implementing Business Intelligence and related systems. It is clear that when data begins to be collected and analyzed, it is logical to use them in many places. The first main question is the collection of data, the second is the methodology of what to do with this data at all. Let's start to understand.

• So, the first is reporting preparation. It's all quite simple: BI acts as a single data repository for all subsystems and is able to combine them, process them in dependencies and return the result. The question of verification also disappears: if you have ever reduced a report of a large production company (under a month of work), then you know that 20 days out of 30 is a check whether everything was done correctly. Because the price of a mistake is good if a couple of millions, and good if rubles.
• The second task is data processing and cutting regular reports for all departments of the company . Of the most unusual, we used the BI system to obtain population censuses, toss them, to supplement, reconstruct, hypothesize, check, process and cut thousands of reports for all territories, cities and villages .
• The third task is to speed up the processing . Previously, the report was made once a month, and now in real time (this is a very, very critical difference for many business areas).
• The same class of systems is used to clean Big Data data and, as a result, to reduce overall operating costs, increase revenues and increase business competitiveness. In general, this is done due to the fact that you begin to make better decisions. Mathematically sound, as above with vodka. From examples of the same class of tasks - retail in the US likes to complement client profiles (for example, due to the behavior on the company's website and favorite products, as well as information from social networks), so that later, for example, to offer baby food to pregnant schoolgirls. They say sales are growing on these proposals from 1 to 10%. Banks are eyeing transactional analytics in real time: paid for a ticket to Europe - the bank immediately offered you insurance. The departments deal with the recognition of faces by surveillance cameras, their comparison with the base of persons and profiles in social networks and the construction of optimal proposals based on extended information. In one large retail network, at least in several large stores, well-known thieves are automatically recognized by persons at the entrance by means of video analytics.

Of course, if the task is a one-time - please, a specialist from the IT department, the script is decided. If the task is frequent, but simple, then the macro in XLS. But for large geographically distributed companies, where the same logistics determines the profitability of the entire business, there may be up to millions of variables at the start. Accordingly, SAS, SAP, Oracle, RapidMiner, Prognoz, Tableau, Qlik, Microsoft, Huawei, EMC, JDA, IBM and others have different solutions for this.

In Russia, solutions based on open-source products, “sharpened” by an integrator for a particular company, are often used. There are SaaS models (for example, one production company from our cloud consumes this service for production planning and distribution of raw materials across production sites).

Company performance evaluation

Naturally, the meta-task of BI and related solutions is an assessment of the company's effectiveness. The effectiveness of the company is evaluated according to a certain model, which is determined by a number of objective functions.

Roughly speaking, there is data, and there is a methodology for evaluating this data. And the model of interrelations of various components with each other.

We collect the actual data values ​​using classic Business Intelligence systems. For many, this BI ends and it is often a problem to combine deliveries and production, because they are made in different places and systems.

Next, the data are pushed into the model of the enterprise. The classical methodology - the same MTP - was created as early as the 87th year, and has not changed much since then. In a sense, updates and forks are in assortment, but the principle is the same everywhere. In short, the company's activities can be decomposed into 4 components: the financial part, the client part, personnel and reserves (that is, personnel) and development strategy. Even state-owned enterprises are valued in the same way, only instead of getting profit there is getting into a given budget.

Charm - in detail. The fact is that when the company is watched by auditors or when something is discussed at the board of directors, a maximum of 20 reported figures are usually estimated, such as net revenue, turnover, etc. It is these indicators that are left to the shareholders in the reports and recommendations are given for them. To evaluate not on sensations - but for myself every day indicators in the form of numbers.

And BI allows you to take a report and not just get a line of “total”, as is usually done, but look at what each indicator is made up of. And then - to fasten to the model a variety of things that are recalculated almost in real time.

Indicators are cascading to people - and the motivation system is turned on. For example, if the shareholders decided that after 3 years the profit should be 20% more, then it is easy to build state A and state B. And the model of transition for 3 years from state A to desired B. At the end of the year (quarter, day) you can see operational performance and understand there digging or not. The performance model can decompose the entire transition process and there will be a strategic map on how to change. Each manager in the plan will have what to do.

Once again: there is a strategy in which it is agreed at the macro level what to do. And there is automation of operating activities - and we can work with it too.

Life sketches

A simple example: in the regions, the report is collected by hand and sent in the form of 1C export. To understand that there is a real dark forest, one can only say, good or bad. And what is good and what is bad - you have to go and dig. Now we give a tool for the head, so that there is an entire operational card. He immediately sees from the reports where it is good and where it is bad - and where his intervention is required, he watches the entire accounting level down to the bottom, if he wishes.

Here is another example. For the company, there are indicators that the parent western company is lowering; this is a strategy and operational indicators. There are executive motivation systems that allow you to reward those who dig in the right direction. Last year they decided to bring it down to the level of each employee - it took a powerful IT solution. Plus, to make a link between the data of the production department and the service department - for example, if you have produced a little, this year there is little service, and this is not a cant of service, but a reaction to the realized production volume. We prepared the project methodology on the product stack. Integration - visualization - and postscript to payroll.

At retail, such decisions often end up pouring out the detail into employees. For example, in a pharmacy above the cashier, from the inside, it can burn, “Today we have to sell another% N of vitamins,” and the cashier will happily offer them to everyone. Like, vitaminchiki not forgotten?

At the global level, the same company that brings bananas to us in Russia uses in its Business Intelligence data on the weather forecast for the year. Seriously, one year ahead they are trying to predict what will happen to the weather. They need by leaps and bounds - and as a result, they almost know exactly how much bananas will cost next year in Russia.

Another company uses a personnel evaluation system for a number of formalized indicators. For example, how much older a person is at the average job in this position, whether the data was poured to the left, how he went on projects ... As a result, the two main metrics are “there is a future” or “there is no future”. And the second - a separate one - is a threat to posts like the vice-president. And the employees quickly understood that HR is starting to send the most “hard-working" people to Europe to learn English. Personnel officers themselves see for each employee a forecast of his departure. The question is, of course, the choice of the correct model and data collection (the same skip systems allow getting information about delays). But in general, the bases for all these monstrous complexes have long been laid in IT. You just need to learn how to use them.

It is important to note that in Russia the process and the person do not always differ. We love to search using BI guilty. In the West, they first check whether the process is correct. And if the process is correct - even then they start thinking about what someone has missed. Change the process, not the person. This intercultural difference resulted, for example, in the fact that in one company the boss introduced BI only because it was difficult for him to dismiss people (and they only rested on the road to TC). They made an evaluation system of indicators, which itself at the end of the month printed out orders for dismissal the most ineffective. And correctly made it all legally.

More specific examples of tasks

Consider a simple example - the supplier announced a price increase for all products by 10% starting in 2015:



As a result of the simulation, an optimal procurement plan was obtained by the end of 2014 for each product line, providing maximum savings in view of future price increases and minimal costs for storing and immobilizing assets in stocks.

If your CFO wanted to get what-if analysis (for example, what happens if you need to switch to another supplier with known parameters), in theory this can also be done in BI or in a logistics optimization system.

The second case - imagine the simplest situation. The company has a distribution center (RC) and a store warehouse, where goods must be delivered from the RC. There are only two products for simplicity - cement and plaster. The weekly demand of the store and the volume of capacities that are at our disposal are known - these are trucks with a lifting capacity of 20 tons. All other reference data are shown in the diagram below.





We are faced with the task of determining the number of vehicles and their loading in order to maximize profits (in fact, there can be many criteria for optimization - this means reducing costs, minimizing delivery time, and others).

The optimization criterion chosen by us is formalized as an objective function:



In other words, we want to select such delivery volumes and (the number of packages of each product) in order to get the maximum profit from the sale of goods.

Here - the total cost of transportation of goods (fixed in our example), but, generally speaking, this is not necessary. We also need to comply with a number of restrictions. For example, the total weight of the goods transported by truck should not exceed its carrying capacity. The total volume of supply for each product should not exceed its demand for this product (according to the conditions of our task, this is not always the case in real life). There are a few points that are understandable to any specialist, but absolutely indifferent to a computer. For example, it is important to specify that the volumes of supply of each product are non-negative. Of course, the number of packages of goods must be an integer. Further the final system is compiled and solved. This is a classic programming, which is good to do with the help of Matlab or Wolfram's grandfather.

From our initial data, we found out that the second truck transports 18,000 kg of goods, and the delivery will bring 11,400 rubles of net profit. But at the same time, you can run a what-if analysis and are surprised to find out that if at any moment it is an inhuman effort to shove 19,680 kg into a truck, the expected profit is 13,300 rubles (1,900 rubles more considering the discount on surplus goods). And this is a reason to think about the fleet used by the company.

Implementation of SSP (KPI, etc.)

In banks, as a rule, the management model cascades well from the top level. There everything is linear, a lot of data is collected, and therefore the introduction of the ERPs gives a decrease in the mess in the branches. Easy to scale. Experience is, and the methodology usually brings a bank. We perform purely technical and integration work.

In the public sector, the question is now in its infancy, as in many other areas of automation. But the structures have grown to personnel evaluation. For example, 2 and a half years ago, metrics sets for all subordinate organizations were launched. And now this topic is relevant. Another area is university ranking, there are about 40 performance indicators. Here they did it simply: they collected data on educational institutions, counted, and then simply demolished the ineffective ones.

Similar systems SSP from mobile operators and retail, from the largest manufacturing companies. Oil industry people love it, their entire salary is considered so.

To make it clear how it works with automation, and not as before, there is a simple example - it used to take up to 6 months to agree on goals for store managers of one network. On 17 levels. Now - only 1 month.

In another company, 1,000 tops were rated by a variety of indicators. It turned out a report of 17 pages, with which they went to the carpet, how much each earned. And the head decided: really earned or deleted. Everything was decided by carpets: finances came from accounting systems, were transferred to paper, and also - on carpets - planned values ​​for indicators were given.

One more our solution works on passenger traffic forecasts and from this builds exactly the number of tellers that need to be replaced in order not to stand idle, but also not to wear out. By the way, there is one more interesting thing - the evaluation of the queue at the cashier on video analytics.

Overall in the market

Despite the crisis and the period before it, the Russian market for business intelligence systems has been growing in recent years (for example, our business analytics direction at CROC has grown almost 1.5 times in 2014, if we count in rubles). This is explained very simply: you need to identify ways to optimize business processes, reduce costs, identify new business development strategies. When a company is big, the price of implementing the solution is obviously less than those 8–13% efficiency that (based on world practice) can be won.

In medium-sized business, among BI-solutions, those that are used for reporting rather than forecasts are more relevant. Although open source is gradually penetrating all sectors of the business, and the range of tasks is expanding.

PS Just in case my mail, if you have a question not for comments, here: Brahew@croc.ru