Managers need to “stop making excuses” for their flawed big data and work to ensure that aggregate information on which high-level decisions are made is complete, accurate, and understandable.
So says Thomas C. Redman, president of Navesink Consulting Group and author of Data Driven: Profiting from Your Most Important Business Asset. In a Harvard Business Review bog, Redman outlined how managers can improve their information management practices to make better high-level decisions.
“An error in a customer database only fouls up a transaction or two,” Redman wrote. “But aggregated bad data can send a decision awry and hurt the company for years.”
He first advises managers to ask themselves whether they really have all the data they need to make an informed decision; whether they understand what the data really means; whether they trust the data in making a particular decision; and whenever they do trust the data, whether that trust is justified.
Incomplete data are a common problem, particularly with key performance measures. Redman cites an example of when a company decides to expedite a multi-step process, but queue time — the time spent between steps — is not measured, making the recalculation of the entire process difficult.
“This is a big deal: When the queue time is high, creating better interfaces between steps can yield huge reductions in cycle time,” he wrote. “If the queue time is low, the actual work within steps must be made quicker.”
Misinterpreted data due to erroneous definitions or misunderstood algorithms are another problem, Redman wrote. So are data that can’t be trusted because they may not be accurate enough.
“It’s particularly disheartening when a big data analysis yields a stunning new insight, but the big data team spent 90% of its time cleaning up the data,” he wrote. “You know they did the best they could, but you’re uncertain they got enough of the errors. Can’t trust the data, can’t trust the analysis, can’t trust the insight!”
While many managers accept flawed data as just another uncertain variable they have to contend with, Redman dismisses that thinking is “sheer folly, particularly since it doesn’t have to be that way.”