With the mainstreaming of advanced data analytics technologies, companies today can risk becoming too dependent on the outputs they receive from the analytics tools, which could serve biased results unless solid data analytics models are applied to the way in which the data is interrogated.
While data is a friend, and the only valid way for companies to strategise based on fact, data analytics tools can only deliver the outputs they have been asked for. If the pool of data being analysed is too limited, or there is no end objective or purpose for using the results after the scientific methods have been applied to the data, then the whole exercise is virtually futile.
It is seldom enough to drill down into a limited data repository and base broad strategic decisions on the findings. In effect, this would be like a novelty manufacturer assessing only the pre-festive season sales and concluding that Christmas trees are a perennial bestseller. Common sense reveals this will not be the case, and Christmas trees won't sell at all in January. But, in the case of more complex products and services, trends and markets are not as easy to predict. This is where analytics comes in. Crucially, analytics must look beyond specific domain insights and seek a broader view for a more objective insight.
Comparisons and correlations
A factory may deploy analytics to determine which products to focus on to increase profit, for example. But, where the questioning is too narrow, the results will not support strategic growth goals. The company must qualify and complement the questioning with comparatives. It is not enough to assess which products are the biggest sellers - the factory also needs to determine what products are manufactured at the lowest cost, and which deliver the highest return. By bringing together more components and correlating the data on the lowest cost products, highest return products and top sellers, the factory is positioned to make better strategic decisions.
Where the questioning is too narrow, the results will not support strategic growth goals.
In South Africa, many companies do not approach analytics in this way. They have a set of specific insights they want, and once they find them, they stop there. In this siloed approach, the results are not correlated against a broader pool of data for more objective outcomes. This may be due in part to factors such as the time and cost required for ongoing comparison and correlation, but it is also due to a lack of maturity in the market.
In mature companies, data sciences are applied to all possible angles or queues and information resources to produce insights to monetise or franchise the data. It is not just a case of finding unknown trends and insights - the discovery has to be purposeful as well.
Questioning extraneous data
There is an additional resource that needs to be applied for data sciences, ie, extraneous public (on the Web) information, such as survey statistics. This data enables comparatives to be done against the company's internal data and is also a source of external or global "run of the mill" trends.
For example, a fleet operator looking to determine which vehicles are most efficient ahead of a further purchase might start by analysing the performance of the company's own vehicles. The company might assess fuel consumption, maintenance costs and distance travelled over the past six months, and based on this, correlate the three insights to identify the top five vehicles in the fleet. But, by stopping at this point, the company has already 'biased' or constrained the data to its own vehicles and within its own parameters. There could be a variety of reasons why the company's own vehicles performed in the way they did, and these could skew the accuracy of the results.
To increase the accuracy of the outcomes, the data should now be compared with a broader pool of extraneous data - from another dealer or from independent research, for instance. Once the analysis moves beyond the subjective, constrained data, it becomes more objective and increasingly relevant as part of an effective, purpose-driven approach to data analytics.