Fall fast, take off high

The fact that it is useful to endure failures is discussed a long time and actively. Mistakes and blunders help you understand firsthand what ways you should not walk. It makes sense to adhere to this position for businesses of any scale – up to production mastodons with complex technological processes. For an innovative business, this is doubly important, provided that another dimension is added to the testing formula – time. The faster the experiment goes, the less resources will be spent on rejecting the wrong solution, as well as finding a successful one. The ability to act quickly and adapt to change is the key to the survival of companies in the face of uncertainty. The problem is that most Russian companies do not yet know how to create an environment and culture for testing hypotheses, which hinders business development.

Failure is a sign of progress

The creator of the electric incandescent lamp, Thomas Edison, literally suffered thousands of failures on the way to success. Moreover, he appreciated each of his negative results on a par with positive ones:

“I can never find what works best until I find what doesn’t”.

Even in areas related to complex technical systems, you cannot do without testing hypotheses and potential failures. Despite the achievements of mathematical modeling and computing over the century after Edison’s work, calculating some processes, although possible, is still often prohibitively expensive. For example, the calculation of gas dynamics processes can be extremely expensive and complex. The vast majority of companies in the world do not have such opportunities. In turn, an experiment, and not an accurate mathematical model, is a simple and cheap way to understand what will happen if, relatively speaking, the temperature is reduced and an additional valve is opened. In this case, the preliminary effect can be understood before the entire technological process is built under hypothetical changes.

In processes where there are many unknown values, it is much cheaper and faster to experiment than to try to derive accurate forecasts analytically. This is most suitable for industries where human behavior is decisive for the result: retail, media and culture, education, marketing and others. There are too many variables in such areas: how people behave in a particular city, in a given season, in given conditions, at a given moment in time. So, to decide on the construction of a shopping center, you can turn to analogies, take data on the foot traffic of a specific place, information on the income and purchases of residents of the area, but the exact result will still be clear only when this specific shopping center is opened and visitors begin (or do not begin) to come to it.

The importance of innovation, finding more effective and smart solutions, constant progress cannot be overestimated. There is evidence that business innovation in the country (which is measured in patents received, research, high-tech exports and similar parameters) is directly related to GDP growth. This is doubly necessary in a situation where business in almost fire conditions replaces high technologies. “Mobilizing intensification of R&D” is named by HSE researchers as one of the most successful import substitution strategies in the long term.

As an example, one metallurgical company conducted an experiment to use video analytics to analyze the clogging of incoming scrap, which it uses to produce finished products. It turned out that if you install a video camera and control the process of unloading cars with scrap, then you can automatically assess the quality of the incoming raw materials and, in case of clogging, reduce the total cost of purchased raw materials, thereby saving on costs

Another example is the introduction of recommendation systems of assistants to the operator, control is based not only on regulations and current indicators, but also on historical data. An experiment was conducted in one fertilizer company. The system analyzed more than 30 parameters coming from the instruments, as well as historical data – how the process changed when the parameters changed in the past – and gave a hint to the operator. Without such a system, the operator set control values based on his experience, the introduction of such a system would increase production by 0.5-1% and reduce unit costs for raw materials.

How to learn to experiment not being afraid to make mistakes?

The fear of mistakes is familiar to all cultures. However, in Russia, a historical component is added to this, deeply embedded in the cultural code. The “initiative is punishable” formula is securely sewn into the heads of our people, including those who work in technology industries, which are in heavy need of innovation.

Recent data show that attitudes towards initiative people have even been deteriorating recently. So, according to the study of the SuperJob service, over the past 10 years, Russians have somewhat cooled to initiative colleagues: 10 years ago, 71% of respondents positively evaluated enthusiasts, today their number is 68%. Russians are also less likely to note a positive attitude towards the initiative from the management (38% in 2010 and 33% now) and work collectives (35% and 32%, respectively). One can assume why. The average employee sees a passionary with a new idea and reflects: “My own work does not become less, they will not give me a bonus, but I will also be guilty if everything fails”. Therefore, if a company needs to develop constantly, it will be necessary to create not only opportunities for experiments, but also a full-fledged culture where a negative test result is not punished.

Russian companies in general do not know how to do this. The culture of Soviet research institutes did not flow into the commercial business of our time, where the emphasis is on the effectiveness of fulfilling current tasks. Of course, the development of such a culture involves the observance of a number of conditions. The first of them is to look soberly at the staff and understand whether there are people in the team who can conduct experiments, how strong their competencies are. The experimental team will need a healthy mix of hard skills, as well as soft competencies.

Hard skills include:

understanding of the subject area;
experience with key AI algorithms, understanding the “materiel”;
understanding of the infrastructure landscape: sensors, network, server capacities for storage and processing; understanding architecture construction;
experience (not only via seniority, but also some specific “chevrons”, such as past crises, special projects, and so on).

Important personality traits include:

initiative;
sensitivity (understanding team tasks and reactions);
honesty (bluntly reporting problems and errors).

Often, employees offer to introduce or pilot a particular technology that is “fashionable” at the current moment. But you need to understand that not all technologies give the same economic effect for different industries, and also strongly depend on the level of digital maturity of the enterprise. It is also worth noting the low level of knowledge in the subject area, for example: in production, they propose to introduce technology based on understanding only “mathematics”, without understanding how a specific business process or unit works. A combination of both skills and knowledge is necessary for a successful launch.

In general, experiments should be treated as a full-fledged tool for the development of a company, which in an ideal scenario should be organically integrated into the structure of making strategic, tactical and operational decisions.

As a rule, in companies with a high culture of innovation (for example, Sibur, Yandex, Gazprom) – regardless of whether they have a separate R&D unit – employees are allocated a separate time for an experimental task and allowed to create their own interest groups. Simply put, management needs to create an atmosphere of trust and acceptance first.
Testing does not have to end in a positive result. Conventionally, it is impossible to set a task with restrictions like: “Well, we need everything to work out, otherwise why are we now spending money and time on it”.
Then you will need the rules for conducting experiments. Employees must understand the entire roadmap: who to contact with the initiative, how to draw it up and submit it, who to involve in the initial assessment. The criteria should be clearly defined whether the idea goes into the experiment or not.

How to calculate budgets for the experiment?

The fail fast concept that came from IT can help here. In information technology, experiments can be carried out quite cheaply and easily – IT specialist allocate a separate environment, write code, and run the test. So the state of the system will show whether it works or not.

On the other hand, the principle itself gives a clear idea of the strategy of the experiment – make a mistake or not, but do it quickly in order not to waste resources, quickly find all the problems and reject all the wrong strategies.

The formula is as follows:

Result = (resources + people) * Time

Accordingly, the less time spent, the cheaper the experiment.

Of course, with all the variety of models for prioritizing experiments (WSJF, MoSCoW, ICE, RICE, KANO and others), there is no universal pattern. For companies of different maturity levels, the absolute and relative cost of experiments will differ. However, to evaluate, let’s apply the analogy.

Suppose you have a limited amount of funds – say 10 coins. Your goal is to produce 100 chests. However, using the available resources in the traditional way, you consistently achieve results in only 90 chests, that is, your plan is not fulfilled.

This is not so critical, but in such a situation, a logical step would be to allocate part of the resources (say, 1 coin) to experimental production methods or innovative technologies. If the experiment is successful, you can increase production to 130 chests. In the worst case, if the experiment does not produce the expected result, you will produce 80 chests, which is a slight reduction from the planned volume and will not lead to significant losses. It is important to be able to assess the risks adequately. If the potential damage from an unsuccessful experiment can greatly affect the overall result, for example, reduce production to 50 chests, you should refrain from such actions. Experiments must be deliberate and have a clear plan in case of both success and failure.

If, after careful analysis, potential losses turn out to be acceptable and do not undermine the main strategic directions of the business, you can and should take risks.

In conclusion, experimentation is a muscle that needs to be pumped. The more experiments a company sets up, the better it learns to do them and the cheaper, faster and more efficient they pass. However, for this you need to decide and start changing consciousness within the company.