Business intelligence is now accessible to companies of all sizes and sectors. I run a small business and we have incredible reporting and analytical capabilities that allow us to model every aspect of our operation. We can create dashboards of website traffic, article distribution, revenue performance, social media analytics and a whole raft of insights from internal and external data that relates to our business.

The same applies to any business, of course. There are no real financial barriers to business intelligence implementation nowadays.

However, with this ease of implementation comes over-confidence. It’s far too easy to "plug and play" without thinking about the quality of the underlying data.

So where can you start with managing data quality in a fledgling business intelligence solution so that you can start creating trust in the findings?

It’s tempting to put every piece of data under the microscope, but I would personally urge an organic, iterative process by adopting some of the following steps:

1. Identify your most critical business intelligence reports or dashboards. Most businesses will only use a fraction of the reporting capabilities they create. Ask managers what information they find the most useful and what actions they are hoping to take with these insights.
2. Make sure you have clear definitions of all the data items found in the report and a lineage of where that data originates from. That alone will help improve the quality of your final insights.
3. Focus on key data items that are used in the grouping process. For example, product names, manufacturer names, sales regions and other category-type values. Mistakes in these can seriously skew your reports and hide errors. In one retail sales dashboard there was a product with a prefix of "café" and another with "cafe." A simple error that caused the sales performance of a multi-million product line to be misrepresented.
4. Identify any calculations or data values that are created by composite values. These are often overlooked and it only needs one small error to signficantly impact an entire report. In one telecom company, for example, a data entry mistake meant that one building was listed as having hundreds of expensive equipment items. Because it was too cumbersome to delete the mistake by hand, the data entry worker left the error in there but notified local workers. When this data was used to create an asset utilisation dashboard, it showed that the region had spent far more than other areas and this impacted budgeting until they spotted the mistake.

These are just some simple starting points to begin building some business intelligence capabilities on firmer foundations. They are not exhaustive by any means, but if your organisation is like many companies that employ no data quality measures whatsoever then they will be a valuable starting point.