During a data quality assessment, one of my clients discovered that a large chunk of data that ultimately fed into their business analytics engine was sourced externally. After examining the contracts surrounding this data, I found that 100% of it failed to possess service-level agreements (SLAs) for the quality of information expected.
I see this problem time after time in organisations of all sizes. Data just isn’t factored into service-level contracts. As a result, any defects get passed down the line as service issues, unplanned administrative costs and decreased profitability.
What if you operated a restaurant but failed to hold your food suppliers to adequate service levels for timeliness or quality of produce? Your business wouldn’t last long.
What to include in a data quality SLA
Here are a few simple ideas to help you get started as you consider SLA requirements for external data.
You could start the first section with a clause around timeliness. If you’re receiving credit information about companies or customers, for example, then you need to be sure that the information is timely and fit for your downstream analytics processes. Consider:
- Does your data supplier include this in the contract?
- What additional timeliness clauses should you add?
- How will you measure that they're being adhered to?
- What's the process for issue management?
Next, you could add a section on completeness.
- Which fields must be populated?
- How many records are you expecting?
- How do you know information is missing (and how will you be notified)?
- What's the tolerance for missing information? (Missing information can have a huge impact on the quality of analytics and decision-making.)
Validity and format correctness could be additional clauses to add. For example, if there are date fields included in the data extract, what are the valid formats? What precision is required for financial information – two decimal places, one, or rounded up?
All of these need to be specified in your supplier SLA so that you don't create excessive processing and additional manual tasks downstream. There are, of course, many more data quality rules that can be added – but the need for them depends on the quality levels you need to support for your particular business.
Bake rules into the process
Once you have your data quality rules locked down, negotiate contract terms with the supplier and build the rules into an inbound data quality monitoring process to validate that service levels are being applied. This doesn’t have to be complex. Just creating a simple reporting dashboard that looks at some key metrics is a great starting point. But take steps to think about how your data is flowing into the organisation – and eventually into your analytics layer – so you can ensure its quality.
The moral here is that you need the supplier to build good quality management into the data upfront. The key is to make sure that all third-party data has an SLA process as part of the contracting phase. Baking this into a data governance policy makes good sense so that the whole organisation can start to adopt the same standard.
How does your organisation address supplier data quality? I'd really like to hear from you. Comment below and tell me about the types of approaches you're taking.
Learn why SAS is a Leader in the 2016 Gartner Magic Quadrant for Data Quality Tools
