File:Dice 2in1 d6.JPGThose days when you need to make an important decision can be trying at the best of times. The old saying “Garbage In, Garbage Out” is never more relevant (particularly if an AI tool is making automated decisions).

In one of my previous roles, if performance wasn’t going quite to plan, I knew exactly what my boss would say to me, so I used to get in there before him and ask myself the questions first… 

  • Is the input data right, so can we trust the outputs?
  • What actually is our current performance?
  • What is causing this performance level?

Only then is it worth thinking about what to do in order to improve performance.

But what if you can’t rely upon the data? If you are close to the problem, then it may be evident where data cannot be trusted. However, if you are remote from the problem, perhaps trying to develop strategic plans for the organisation, then you will not know when data is inaccurate and may blindly assume that, since the data is complete, valid and plausible, that it is correct. This is not actually the fault of the analyst, since they have no better information to use, but local staff may partially be to blame if they know the data is wrong but don’t do anything about it.

If people start mistrusting the data in the corporate system because it doesn’t agree with their version of the data, they may then decide to create their own spreadsheets that they keep updated, ignoring inputting into the corporate system (that enables others to also make decisions). This is a classic! What makes it even worse is when this person moves into a new role and their successor quickly realises that they cannot trust the data in the corporate system, is not sure they trust the spreadsheet (which they would have developed differently anyway) so they create a new spreadsheet. You now have three (or more) versions of the data which may not agree with each other 🙁

In some organisations lots of time and energy (and money) is spent gathering data from different sources to come up with alternate views of what the performance actually is. This can then result in lengthy discussions (perhaps arguments) about current performance levels rather than spending this time and energy working out how to improve the organisations performance.

Sometimes, data quality problems are easy to spot, the wrong house being demolished or a space probe failing. Sometimes they are far harder to spot, such as a strategic business plan not being able to adequately evidence ~£600m of projects or organisations losing 30% of their revenue due to poor data.

So what can organisations do to minimise the risk of data quality problems?

First, they need to understand that poor data quality does not have to be ‘normal’ – changing beliefs and resetting the view of what is acceptable is critical.

Secondly, data quality should be managed in a holistic way, similarly to other aspects of quality. ISO 8000 provides a range of standards relating to data quality and data quality management.

Third, like improving health and safety performance, the organisation needs to recognise that it will take time and sustained effort to improve data quality levels.

Why bother? Well, the benefits of improving data quality management align closely with other quality management activities by removing waste/avoidable cost from the organisation and avoiding poor decision making.

If this all sounds challenging and leaves you wondering where to start, two useful starting points can be providing data quality management training for key team members and assessing the maturity of your approach to data quality management. The first of these will improve the skills and awareness of your team, the second will provide recommendations of where improvements should be applied and should help trigger debates within the organisation over which actions to prioritise and the data areas where most benefit will be secured.

A follow on blog post ‘Does your data trust you?’ takes an alternative look at this challenge from the perspective of the data itself.

Want to find out more? Then click on some of the links above or get in contact for a FREE introductory consultation.

Tagged on:             

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.