Big Data, Computing Power & Analytics: 4 Obstacles to Successful Implementation
We've got big data, the computing power and mathematical techniques so why is it so hard to embed analytics in to business processes? Perhaps we should be refining demand rather than increasing supply. Some of the key blockers may be
Not all data is useful data - a lot of it is the waste product of manufacture or fuzzy proxies for attributes we already measure. Often, we have the response (e.g. the customer action) or the explanatory factors (the system at the point of the action) but not both… or they lack the rigorous matching required by the analytical techniques.
No substitute for understanding the business – beyond descriptive techniques many investigations search for useful causal relationships. This requires a firm idea of how the business serves customers and makes money combined with the ability to express these in mathematical terms. It’s this combination, business knowledge and analytic ability, that can often be the sticking point.
Embedding analytical capability – as analysts are a scarce resource it is often tempting for large organisations to group them in to a centre of excellence. However, it may also separate the new team from both (i) the local processes and systems that produce the internal data and (ii) an understanding of how the business units generate value.
Implementing results is the hardest bit – nothing changes unless the results are implemented back in to the business. Although the insight may be powerful the change to exploit it may need significant resource. This requires a management that views analytics as an essential component of the business mix as well as prioritising delivery.
So we do have the data, computers and techniques but
- Skilled people in these analytical techniques are rare
- Those who can develop innovative analyses that improves the business model are rarer
- And those who understand the techniques and results well enough to pick the winners and implement them are the rarest of all
Great points from Jackie Penny & Alan Salamon, high quality data that matches the underlying physical model is a prerequisite (the preparation can be long and boring but has to be done) and its our creativity and knowledge that will define good questions - not just the analytics
As someone else commented, to get value from big data and powerful analytics you have to know what question you want answering and why. This also relates to identifying flaws in your current strategy. Sometimes when you identify 'the question' you find you don't need to generate a big data related project. BIg data and deep analysis can be a solution looking for a problem and a crutch on which to avoid leadership in decision making.
Great question. Quality data analysis has to be underpinned by excellent data model. The business data model and physical models must be aligned in order to get meaningful outputs . In large organisations where data is collected from multiple sources/systems (some third party) and also media entry points, it is a challenging arena. Skilled data and process analyst resource pools need to be cultivated to bring complementary skills together. Business/technical language barriers have to be bridged. Mapping processes is relatively simple; but it becomes complex where data analytics has to be sub divided and linked to key value hprocess steps e.g combined data split out to customer journey/entry/contact points. Of utmost importance is the understanding of what problem is to be solved. Without a clear objective, data and process analysis will not have a focus.
What I'm finding is that analysts can't always produce because the questions they're asked to answer are not always the right ones...
We ask a lot of analysts, we expect them to understand the business and the very technical data we buy in. The major stumbling block however, in many organisations, is the IT which is slow to change and adapt to new inputs - the tail wagging the dog - and often stopping implementation.