In 1966 (yes, no typo) the Harvard Business Review (HBR) published an article on How to Buy/Sell Professional Services. This was written in response to a rapid increase in their purchasing by management. A recent read of this article led me to question:

• Do the same principles apply for buying/selling analytics as a professional service? And,
• How do they translate to the procurement of analytics from external providers?

Digital Transformation and Analytics are in the strategy of all of today’s big and small businesses. With the advent of AWS, Azure, Google, Atlassian, Salesforce and other cloud-based vendor services, this combined with the implementation of Enterprise Resource Planning tools such as SAP, Dynamics, Oracle ERP, etc., there has never been more data to manage, govern, curate, control and interrogate.

With so much data on the minds of today’s management, and the desire to leverage it, there are many more businesses offering analytics services. But can the principles of yesteryear be applied to decision making today?

The answer is a resounding Yes. The following text translates three key objectives of Professionals Services from the 1960s into the day-to-day of an analytics professional in the 2020s and, importantly, provides a framework for management to consider when purchasing Analytics Professionals.

What aspects of the HBR article maintain relevance today for those looking to purchase analytics skills?

The article makes mention of the key aspects to consider when purchasing/selling Professional Services. The buyer must be confident the seller will:

• Minimise uncertainties involved in managing a business
• Fundamentally understand the business’ problems
• Be capable of rendering the service

In delivering professional services in the form of analytics services, these are exactly the key aspects we address in today’s world, albeit with a slightly different skillset. To the uninitiated, how do these translate and what are the activities that an analytics professional will employ to deliver each of these?

Minimising uncertainty

Statistics is concerned with the use of data in the context of uncertainty and decision making in the face of uncertainty

Wikipedia definition of Statistics

Looking at the Wikipedia definition of statistics above, there are parallels with the HBR article’s themes in the first bullet point (minimising uncertainty). While analytics seeks to employ statistical methods that account for variation and mitigation of its impact, there are more simple techniques that bona fide data scientists deploy daily to reduce uncertainty.

Uncertainty can arise from poorly managed data; for example, in a database. On more than one occasion, we have found that businesses simply do not manage the inputs to their data sources to draw the simplest of conclusions. Indeed, most end users compare data entry with putting information into a black box without ever having to worry about it after that point in time. Quite often we will set out with one goal in a project, but the best outcome in such a scenario is a process change that begins to treat data as an asset. Without a data specialist, the data going into the database will remain unknown, and by its nature it will increase uncertainty for decisions makers. Uncertainty in data isn’t restricted to databases – how many times have our readers found an error in a spreadsheet?

Uncertainty can be minimised by codifying repeatable and refreshable analyses that would generally be reserved for the spreadsheet professional. Codifying an important decision-making analysis will lend itself to being governed by a code review, containing metrics that are signed off and stored in a git repository (where all changes can be traced via versioning controls). Changes can be managed from development, to test and into production.

Analytics can address many other issues where uncertainty can lead to

• Impacts on worker safety;
• Changes in productivity;
• Poor planning and execution;
• Poor process compliance;
• Poor decision making; and
• Levels of assurance and audit.

Understanding business problems

Understanding business problems is undoubtedly the “achilles heel” of the analytics industry. There are countless stories of the data scientist focussing on the intricacies and sophistication of a model and its outputs, while not considering the broader picture (see HBR article section on extrinsic, persuading by method). Quite often a data scientist will go away with data only to reconstruct a microsystem (and excitedly, too) without understanding the broader business problem.

For most data scientists, the business problem can be easily forgotten and sometimes approached with malaise, as for a data scientist a mathematical concept can be more enticing than a simple process change. Hence the “achilles heel”, akin to where the professional has a toolkit with a sledgehammer in it and uses it, when the client would’ve been better served with a tape measure. Indeed, the creation of a new process, rather than the build of a machine learning model, may present the best outcome.

A truly confident Analytics Professional will have the tools in her/his back pocket but will focus on the biggest and most meaningful business problem first. In many cases, we have found that on commencement of a project such as a systems-modelling approach, the interviews and information gathering provide the greatest insights to the client over and above the outputs of the intended model. Likewise, we see simple business reporting solutions and data quality assessment generating 80% of the value of complex analytics solutions.

An experienced and pragmatic data professional who is willing to ignore her/his data science skills for an alternative outcome will focus on the business problem for the best result.

Capabilities for rendering the service

What are the skills of a capable data professional in this rapidly evolving field? As a firm, we have an extensive history of working with a large number of data professionals both in coding solutions and solving complex problems for clients. As a firm, we see many graduate CVs submitted for review.

Our experience is that the best professionals are confident with physical modelling of the real world (Physics) and advanced Mathematics, Engineering and Statistics. These are very difficult courses to excel in at a tertiary level and they are generally focussed on mathematically solving real world issues.

Of course, experience is also key: professionals need to be exposed to different data sources, structures, coding languages and modelling methods, as well as a broad set of clients and industries. They need to know when to sit up and leave the toolkit, but also when to roll up the sleeves and get their hands dirty on a sophisticated model.

Capabilities must include understanding the value of the business, getting to grips with its manageable uncertainties, detailing its processes and meeting with its stakeholders. All this is done by excellent data scientists before interacting with its technology.

The HBR article refers to the “The true professional” – does one go with the best salesman or the professional who can sell? In our experience, analytics are best sold by the latter as key risks and intricacies of techniques are truly understood by this professional. The only real issue with this strategy is in finding the true professional. The current market is so tight for genuine, quantitative professionals that the purchaser may only ever get to meet the professional salesman.

Wrapping up

In summary, an open mind and due diligence are the keys to choosing the right provider of data analytics services. A team with a strong quantitative background is foundational for an analytics project, and this must be combined with one that can work closely on the core business problems to deliver value.

Any partnership between seller and purchaser must begin with the understanding that the initial problem the firm is facing may not be the problem that is solved. This is most likely to happen with flexible and confident analytics professionals often willing to recommend a process change over some model outputs. Adhering to this framework will provide the tonic that purchasers of analytics need.