Five thoughts on data analytics

Don't let politics get in the way
When creating actionable insights from social and digital monitoring tools, there are various stumbling blocks. Edward Rhodes, vice-president of sales at Linkfluence, said it was about finding the right organisations to work with to help capture the right datasets while Richard Maryniak, global chief insight and innovation officer at Black Swan Data, pointed to internal politics within businesses being the biggest obstacle as it was vital that teams worked together. "To get true insights from data you can't just employ data scientists – if you did, you'd fail," he added.

Social media surprises
Social listening is a top-down exercise, but for bottom-up learning you can create an algorithm so it understands the connections between conversations and then you discover the unknown unknowns, claimed Maryniak. "Because you're not asking the question, you're surprised at what comes back," he added. "For the first time, we're finding answers to the questions we didn't know to ask."

Looking back, to help you look forward
When Nuffield Health wanted to increase the number of new joiners signing up for its gym membership it looked at historical data to help forecast. While this gave it 83% accuracy, it couldn't identify individual drivers. So it created a decision tree learning algorithm and identified three factors – marketing spend, awareness and consideration. The analysis showed that Nuffield was falling short on consideration and so its comms needed to drive that. The resulting ad led to a 57% increase in spontaneous awareness; 5% increase in brand image and 2% increase in consideration.

Think about the ‘why'
Ray Poynter, managing director of The Future Place, made many a market researcher – and journalist – feel sheepish when he mocked the much repeated line that data tells you the ‘what’ but market research is needed to tell you the ‘why'. He scrutinised the audience, questioning whether we always even need to know the ‘why’ before identifying those occasions when we do. These are when we want to: change minds; change an attribute or feature to change behaviour; design something to meet an unmet need; and know if a pattern will continue. Sometimes the ‘why’ is a ‘nice to know', rather than a ‘need to know'.

Beware the bias
The idea that artificial intelligence is rational is wrong. "Everyone assumes AI is unbiased but the problem is it’s humans writing the code and collecting the data, so bias filters through," said Bethan Blakeley, director of Honeycomb. "It is damaging society."