Alexei ChernobrovovConsultant on Analytics and Data Monetization

How not lost a fortune on Data Science

Today, Data Analytics is needed in almost every private company and budget organization. But an experienced Data Scientist is quite expensive in the labor market and not every company can afford to have such a specialist on a permanent basis. In this case, it seems reasonable to involve a third-party consultant in organizing the processes of collecting, processing and interpreting corporate data in order to extract valuable business information from this. However, not everything is so simple - in reality, about 85% of Big Data projects fail [1]. It's hard to tell the exact share of data analytics failures, but the overall figure is impressive. In this article, I will tell you how to avoid such a fate and make your data analytics project successful with the help of a third-party consultant - practical tips for business “on the other side of the barricade”.


Prerequisites for the success of a data analytics project: determine the initial state

PMBOK, the most popular set of knowledge on project management, does not accidentally separate initiation and planning into separate groups of management processes. It is at these stages that the fundamental concepts and the Charter of the project are developed and approved, and its content (Project Scope) is also determined [2]. In this case, all interested participants (stakeholders) should be identified in detail, describing the nature and degree of their involvement in the project. The customer (investor) or potential user should have a clear understanding of the results that the data analytics project will bring them, as well as what efforts (not counting the fees of the consultant involved) this will cost them. In the absence of this, the project is doomed to failure in advance.

Since the budget of start-ups and small businesses for R&D activities, which include consulting services for data analytics, is quite limited, you should understand where and in what form your information arrays are involved before you engage an external Data Scientist. The cost of the analyst’s work directly depends on the nature of data storage and presentation. For example, if the enterprise has databases, cloud-based CRM systems and other repositories of corporate information, the analyst will not have to spend his time (and your money) on organizing the processes of collecting information arrays.

You should also independently determine the maturity level of your business processes using the CMMI model [3]. It makes no sense to start a data analytics project and pay an expensive specialist if you have not reached at least CMMI level 3 - chaos cannot be analyzed (Fig. 1).

Fig. 1. Levels of process maturity according to the CMMI model
Fig. 1. Levels of process maturity according to the CMMI model

What and from whom to expect: consolidating responsibility

Once you understand your initial state, determine what you want - what exactly do you expect from a data analyst. It is the customer himself who must clearly identify the goals and results of the data analytics project. For example, Data Analyst is committed to deploying the Tableau cloud BI system [4] for you and setting up continuous monitoring of the 10 most important business indicators (total profit and revenue, conversion, detailed expenses and revenues for various items, effectiveness of advertising channels, etc.).

It is important to understand that Data Scientist is not the only participant in a data analytics project. You, as a customer and a future user of a product created by him (BI-system, statistical models and recommendations based on them for changing production processes), are directly involved in project activities. At the same time, one should not exert excessive pressure on the consultant by checking any of his actions or independently understanding each mathematical formula and SQL query. However, if you ignore the questions of the analyst or are not interested in the intermediate results of the project, you can get at the output not what was originally expected. You can also significantly exceed the project budget by going beyond the time frame or by giving the consultant complete freedom to choose tools. If the Data Scientist, carried away by the task, chooses a too expensive package of statistical modeling or an "overly sophisticated" ETL system, you will have to pay for it. And the breadth of functionality of the analytical tool is not a guarantee of the successful implementation of the data analytics project. Therefore, it is very important to determine in advance the procedures for interaction with an external consultant, having precisely specified the nature and frequency of communications. For example, a brief report by e-mail once a week and a personal meeting 2 times a month with the customer. You should also consolidate the responsibility for the implementation of the project on the customer side, creating a temporary team of its own employees, with whom the consultant will work. For example, such a team may include a product manager, marketer, database administrator, and other specialists involved in the processing and use of the analyzed information. Flexible project management methodologies (Agile) and methods supporting them (Scrum, Kanban, Lean, etc.) [5], implemented in most modern IT tools for remote interaction (Trello and other similar tools), will help organize the work of such a project office; managers).

Finally, when concluding an agreement with an external data analytics consultant, be sure to sign the NDA (Non-disclosure agreement) - confidentiality and non-disclosure agreements.

How to choose a consultant

Many beginners and experienced professionals communicate with each other in the subject forums on data analysis, Data Science, Big Data and Machine Learning, for example, Data Science CentralKdnuggetsSmartData Collective, Cross Validated section on Stack Overflow. You can also look at the Kaggle and Boosters competitive platforms or offline conferences, festivals, championships and other similar events (DataFest, meetings from Yandex,, Avito and other data-driven companies) in search of Data Professional’s.

Having a profile education with a candidate is not a guarantee of an excellent result. A much larger role is played by practical experience in the field you need, for example, marketing, telecommunications, HoReCa, etc. Nevertheless, a qualification in mathematics, computer science, systems analysis, Data Science, Data Mining or Business Intelligence confirmed by a diploma or certificates will be a very significant factor when choosing a data analytics consultant. The recommendations of colleagues, competitors and customer reviews from past projects are another important criterion for finding the right Data Scientist.

When evaluating a candidate, pay special attention to his working skills in conditions similar to yours, for example, whether he has encountered similar volumes and storage systems, and how successfully (quickly and efficiently) he has solved similar problems. Finally, in addition to technical competencies (hard skills), do not forget about the universal human qualities of a reliable employee: working capacity, adaptability, stress tolerance, sociability, attention to detail and other similar soft skills. To work effectively, even a temporary freelance consultant must meet your corporate values ​​and industry requirements.


To summarize the above, we highlight the following key success factors for a data analytics project involving an external consultant:

  • the customer has an adequate understanding of the initial state of their business processes, corporate data and the results of the analyst’s work with them;
  • precise definition of the goals and results of the project in clear wording and measurable indicators;
  • a detailed description of all the nuances of the project before its implementation in the Charter of the project and other documents (Project Scope, calendar plans, budget, procedures for the interaction of participants, a list of stakeholders, etc.);
  • a high degree of customer involvement in the project implementation process with constant communication online and offline;
  • detailed assignment of responsibility to the consultant and staff members of the customer for each indicator of the project;
  • fixing all initial agreements and interim results arising in the process of work in documentary form;
  • the consultant has relevant experience working with similar tasks, information systems and data volumes.

In the absence of the above factors, the data analytics project will most likely fail and replenish the piggy bank of negative human experience in Big Data practice (Fig. 2).


Fig. 2. Top 5 reasons for the failure of Big Data projects [1]
Fig. 2. Top 5 reasons for the failure of Big Data projects [1]




  2. Project management processes
  3. CMMI
  4. Tableau BI platform. Official site
  5. Top 7 project management methods: Agile, Scrum, Kanban, PRINCE2 and others