Big Data - the devil is in the detail
Big Data is an amazing source of new opportunities for businesses. But it can only be a source of innovation and a means to stand apart if we pay attention to weak signals.
In 2015, 8,000 billion gigabytes of data were collected around the world, and this volume is set to double every two years. Big Data can be defined as “all technologies, methods and practices intended to store and rapidly analyse extremely large data sets”. It is an amazing source of new opportunities for businesses.
Big Data as a source of optimisation
Let’s take a simple example. Thanks to Big Data, the supermarket chain Target can now predict when one of its customers is expecting a baby, and it can offer the customer personalised discount vouchers. This practice is not limited to marketing. For example, SNCF is fitting its rails with smart sensors that allow it to better plan its maintenance schedules... nearly 30% of French farmers now collect data on how their crops are ripening to adjust their use of fertilisers in each square metre of their fields... and Toulon rugby club, RCT, uses GPS trackers to analyse its games...
However, a surprising trend is emerging: the use of Big Data is tending to standardise practices and analysis. Indeed, if we all collect the same type of data, it is highly likely that we will all come to the same conclusions and take the same action. So how can we use Big Data to innovate and stand apart from our peers?
The detail that makes all the difference: weak signals
The answer lies with what are known as weak signals. This refers to information that could easily be overlooked among the mass of data collected because it occurs infrequently or deviates from the norm. This information is, however, crucial. While Big Data can identify typical behaviour and produce general conclusions, weak signals can pinpoint an isolated factor that could in time lead to a major event (new trend, problem), but which needs to be investigated before it can be deemed useful.
For example, Big Data means I can see that high-tech product sales on my website peak at the weekend. The conclusion is clear: I should promote this product category at the end of the week. However, if I notice that a particular employee starts to use the company’s electronic tools less (fall in number of e-mails sent, intranet not used often, etc.), this is a weak signal that needs to be analysed so I can make the most of the information.
How can a weak signal be used?
1 - Make hypotheses about the weak signal
First identify the potential causes of the variation in activity. Is the person lacking motivation and could this feeling spread to the rest of the company? Or is it an individual case requiring a personalised response?
2 - Monitor changes in the weak signal
Changes in the weak signal then need to be monitored in light of these hypotheses. Is the decline in use confirmed for the employee in question? Has it extended to other members of staff?
3 - Confirm or reject the hypotheses
If the weak signal is confirmed, an investigation on the ground and individual interviews could provide me with more detailed conclusions and allow me to react accordingly. The weak signal could even provide information about a situation I had not previously considered, such as the gradual uptake of new tools among staff, pointing to new requirements and new ways of working.
As you can see, the biggest part of the challenge of weak signals is not identifying them, but dealing with them. Without monitoring changes in the weak signal or investigating the situation on the ground, it would be impossible to identify the actual problem or resolve it.
While overall, big data sets can identify major current trends, tomorrow’s trends can be found in weak signals, which can put you a step ahead of your peers. However, the two complementary approaches should not mask one important point: regardless of what information is collected or how it is processed, value is only created by analysing it and acting accordingly.
- "Traitement des Big Data et importance des signaux faibles", Horizons décisionnels
- "Analytique RH : Big, Small ou Smart Data", BiHRdy
- "Big Data : signaux faibles et lutte contre la fraude dans l'assurance", ZDNET
- "Big Data : à quoi sert le traitement des signaux faibles ?", ZDNET