"If you do not know how to ask the right question, you discover nothing."

– W.Edward Deming

“Why won’t that project deliver on time?” “How can I improve staff retention?” “Why is no one is buying my product?” “Who should I be targeting to attract more investors?” “What are our profit margins likely to look like next year?”

Business challenges – every organization has them. From HR, to supply chain logistics to customer profiling and demand forecasting. In the digital age and the Internet of Things, businesses are able to acquire data. Whether those datasets are big or small, structured or unstructured, data is the basic building block of everything we do in analytics: the reports we build, the analyses we perform, the decisions we influence, and the optimizations we derive. But it’s not necessarily about having more data – it’s about choosing what to interpret that brings real value. As Einstein famously said, not everything that can be counted counts, and not everything that counts can be counted. It’s about identifying the right piece of the puzzle to play with – and that depends on the business goal in question.

The translation of problems to technical solutions is often a stumbling point, as business insight desired often gets lost in the technical translation. By framing business challenges in the correct way, technical data scientists can ensure correct analyses are carried out and the relevant processes defined to solve the challenge; ensuring leaders are equipped with the insights they need to optimise the business ecosystem.

1. What are the current stock levels of each product that we sell?
2. What do staff productivity levels look like today?
3. How many copies of my publication sold last year?
4. How many people stayed at my hotel last month?

Descriptive analytics to the rescue! The aim here is to answer questions which look at ‘moments’ in historical datasets e.g. total number, average of a spread of data. The output of such an approach is usually the traditional business intelligence dashboard or spreadsheet report with clear, digestible stats and facts about the business in hindsight.

But whilst digesting the raw data and figures in the dashboard business leaders also often want to gain insight into why these historical events happened in the first place.

1. Why has there been a surge in stock levels?
2. What has caused staff productivity to drop by 40% this year compared to last?
3. What factor lead to publication X selling more this year compared to last
4. What is the reason for more people visiting the hotel in the summer compared to the winter?

These are all examples of where a diagnostic analytics approach can be used to define use cases, which answer why something happened so steps can be taken to improve or avoid a situation in the future.

So great – your business can conclude trends in moments of historical data and understand why these things may have happened. But that’s often not enough. In today’s agile markets, business leaders need to be continuously innovating to gain foresight and remain competitive. Data scientists do this by looking back at historical datasets and identifying trends to predict future events, which can inform the on-going business strategy – those who do not remember the past are condemned to repeat it.

1. What publications are readers likely to buy if they also read magazine X?
2. What factors determine whether staff are productive or not?
3. Who is most likely to buy my product, considering they also buy X product on a regular basis?

Predictive analytics are required to accommodate this type of business challenge. Technical analysts fit models to datasets to identify ‘rules of thumb’ or relationships between different variables. Once a model has been identified, different variables can be tested against the model to determine the influence of that variable on the particular model in question. Use predictive analytics when you need to understand something about the future – understand trends, detect clusters or exceptions.

Often businesses get too caught up looking backwards and not enough time taking action on the analytics they see. Business leaders also want to know how to optimise elements of the business, given known constraints.

1. What is the most efficient way of producing X, ensuring seamless collaboration across the business ecosystem, given the resource and tools I have?
2. What is the best way to minimise the production of waste, given the existing processes?
3. Given current resource levels, how can we best deliver the project on time?

Prescriptive analytics are required when the business wants to find the best combination of input that either maximises or minimises the output given constraints and is usually the most valuable kind of analysis for an organization. Data scientists simulate multiple outcomes and choose the best to communicate to business stakeholders. The results of this type of analytics are usually delivered to stakeholders in a written report with step recommendations on the best course of action to take.