Big Data vs the right data in pharma decision-making

Tom Lancaster sets out how to get the most from the wealth of information available and points out the pitfalls of using poor source data or asking the wrong questions.

My dad is a retired statistics professor who views Big Data askance. It’s not because of the massive hype the topic receives, which Gartner says, in its recent report, means it is headed towards the ‘trough of disillusionment’ and it’s not because my dad came of age in the era of the slide rule.

He’ll assert that it doesn’t matter how much data you have if you don’t know what question to ask – and this remains the age-old challenge for all researchers.

He’s right, and pharma teams should take heart. Big Data is not always the right data. Voluminous data sets that are collected, maintained and analysed represent a potential treasure trove for marketing and other departments. Yet the emphasis placed on analytics drawn from vast data sets – and the amazing gains that technology has enabled in speed and visualisation of it – obscures several correspondingly oversized problems:

1) If you’ve asked the wrong question to begin with, a massive sample size is meaningless. As one expert related: “if you don’t know what you’re looking for, you can find just about anything.”

2) If your data governance – data capture, storage and curation – are inconsistent, it’s difficult to assert that the follow-on conclusions drawn from it are representative, let alone accurate.

3) Even if you have massive data sets that allow teams to see unsuspected patterns, correlation is not causation. This is something drug makers must watch with care, lest teams suggest patients abandon treatments when causality is not there, as in recently identified increased melanoma rates seen in patients taking erectile dysfunction drugs.

4) If your data are dirty, for example criteria asked of a patient are imprecise when turned into multiple choice questions, or records are missing information, or asking questions that are obsolete, your conclusions can be flawed. A Toronto physician-turned-data-scientist bemoaned the challenge of deciphering medical records saying ‘smoking = 0’ when trying to figure out how many cigarettes a patient smoked, pointing out that where humans could easily understand what this meant, computers would struggle.

Pharma brand managers, market insights teams, and marketing leaders are starting to see through the hype and know that they need a better approach. Here are three ways to drink from the Big Data firehose without your conclusions getting washed away:

1) Boost data literacy, and not just by hiring brilliant data scientists. Dozens of universities are working to deliver data science-savvy graduates to meet the brisk hiring demand, and that’s fine. They are needed to help connect the dots between the technology platform and actionable information. However there needs to be emphasis on empowering information end users, so they can understand and apply the internal treasure troves for meaningful insights that can achieve business goals. Having data doesn’t mean you know what to do with it.

2) Explore the questions to be asked first. Use a smaller, yet meaningful, sample size of 50 to 100 individuals, to test focused questions, both closed- and open-ended, where comments can help refine the most meaningful questions to ask for the business decision at hand. Perhaps conduct a few in-depth interviews to drill down on certain aspects surrounding an issue. This is particularly helpful when market circumstances dictate immediacy.

For example, a large pharmaceutical company was days away from finalising a major acquisition when key opinion leaders flagged a minor side effect that they felt might impact prescribing behaviour. The pharma company had access to reams of data about the drug and the side effect, but not the specific answer, from likely prescribers – an answer delivered in that moment, in the context of the present debate. By using an on-demand market insights approach to frame the question, within three hours, the company had market data showing that the side effect in question was not an issue.

3) Make technology your friend. Leverage mobility among prescribers; remember most physicians use their mobile devices at work and four out of five physicians are expected to own a tablet this year. This involves more than simply enabling web pages for mobile device browser access. Dialogues should be made shorter, taking just minutes to complete and asking from four to eight focused questions, so that actionable results can be made available in hours, not weeks.

With the rise of physician online communities of various kinds, targeting and accessing prescriber expertise and opinions has never been easier.

An April 2015 Forrester Research report explained the Big Data challenge as ‘businesses are drowning in data but starving for insights’. For pharma teams, and any stakeholder trying to optimise resources, staying afloat means knowing the difference between the Big Data they have, the right data they want, and building an ongoing process to help the one deliver the other.

About the author:

Tom Lancaster is senior vice president of technology strategy at InCrowd, an on-demand market intelligence provider to the life sciences.

Read more from InCrowd:

The best route to physician engagement