Why prioritising 100% data accuracy is a mistake

Matthew-Teale-002Now more than ever, it is constantly drilled into marketers that decisions must be data driven and the prerequisite for this is a good database. “Good” tends to mean a customer database which is clean, up-to-date and containing the relational database structures that allow for a single customer view.

But while maintaining data quality is a laudable goal for any organisation, we would question the wisdom of obsessively prioritising this goal.

Our argument is that brands waste precious time chasing the chimera of data perfection. Had they acted on insights from their untidy database, brands could have allocated time more wisely to fixing business problems.

Essentially acting on insight from a database which is 85% accurate is always more efficient than waiting until your database is 100% (which it never will be) and then implementing the changes once this goal is achieved.

There are numerous hypothetical scenarios this could apply to. A company delivering bespoke cakes does not require a single customer view database with 100% accuracy to know that its deliveries are running low and packing is not working correctly.

Instead of correcting obvious deficiencies, the business ploughs on trying to improve data quality and their bottom lines suffers as a result. The equivalent of a DJ who spends all their time alphabetising and rearranging their record collection, that they never get round to playing any of them.

The above example outlines the importance of a round table discussion with different departments as opposed to gaining insight from perfect customer databases. A discussion like this is likely to reveal the key priorities for the organisation and mould the strategic direction. In a data age, it is easy to forget that some solutions can be found more efficiently through open dialogue and collaboration.

The metrics that are most meaningful are often available in the absence of a single customer view. Delivery lead times, availability to purchase, delivery accuracy, customer reviews, these data points are all easily accessible and will identify most problems that need to be fixed.

These metrics are straightforward to impact and are easily accessible. Focusing on improving these metrics will mean organisations do not lose perspective and get bogged down in the details.

There is also the problem of death by data. Various companies have been led to believe collecting every bit of data under the sun is a desirable goal. This can result in organisations spending vast amounts of time and resource collecting completely superfluous data that is never used.

However, this can be avoided by communicating with different parts of the business to establish what data is necessary to collect and putting in the place processes to achieve this. This will avoid amassing large amounts of extraneous information that has no bearing on business objectives.

The reality of working in the data business is that data quality is never going to be 100%. Having in place the correct data quality protocols is important for organisations but the highest quality data is never useful after the fact.

What businesses should strive for is balancing good data quality protocols with performing the necessary analytics and deriving insight from the data that exists. Prior to analytical work being carried out, good analysts will always clean data and identify any outliers/investigate spurious patterns.

This is far from a rally cry to down tools and stop working on improving data quality. It is a reminder that perfectly accurate data is an unattainable goal and that it is better to implement changes based on nearly accurate data than it is to vacillate until a higher level of data accuracy has been achieved, by which time you may have missed the boat.

Matt Teale is head of data science at Metrix Data Science

Print Friendly