Data quality explained
Data quality is a hot topic in any agency, especially when you're about to start a new integration project to connect a website or app with a back-office CRM or other system, usually meaning a uni- or bi-directionally data sync between multiple data repositories. Data quality plays a major role in many areas of our work; marketing campaign planning, informing decision making, optimising operations and business processes in general, and is becoming an increasingly popular topic of conversation.
Wikipedia suggests that data is generally considered high quality if it is fit for [its] intended uses in operations, decision making and planning. They use this diagram to explain the different and most common quality dimensions used by traditional encyclopaedias (format, authority), wikipedia (style, completeness, timeliness, objectivity, credibility, relevance and readability) and Web 2.0 (usefulness, reputation, involvement and accessibility), so a good overview of the approach from yesterday and today.
The important part of the Wiki description for us is ‘intended uses in operations’, as quiet often we find that data is structured perhaps to support old processes, or maybe it was previously fit for purpose but hasn’t kept up with the pace of change or new processes.
Good quality data is sensibly structured, easy to understand and consistent. It is often said that if data or logic can’t be communicated easily in an Excel document without annotations and notes, then the data quality isn’t good. But whilst data is of course about rows and columns of raw data, data quality is often measured on how data is used, so more focused on processes. For this reason, a review of business processes is a good idea, to make sure that their data enables you to work how you'd like to, not necessarily how you’ve always worked before.
See some examples of data quality, and how this can impact your organisation:
One area we often find significant room for improvement is with product data, especially when importing or synchronising data from stock, fulfilment or order management systems. Hurdles we often have to overcome include;
- Product variations as separate products and SKU’s instead of for example a single product with variations – ie SKU001-blk
- A single ‘size’ field used for different types of measurement; S M L XL, 120mm, 14Kg, 12v, size 10
- Incorrect use of fields for example using ‘title’ for post nominals resulting in Dear OBE Smith
- Manually specifying prices in multiple currencies making it difficult to quickly adjust exchange rates
- Products being managed in Salesforce or another CRM that don’t support multiple images, variations, or semi-complex categorisation.
Undoubtedly the most common data type we work with is personal contact information, specifically that from customer relationship management (CRM) platforms like Salesforce, Sugar CRM, ACT, MS Dynamics and Zoho – or even Excel or in .CSV format.
If the data is being used for marketing or providing customer services, then consistency, completion and segmentation are key to having good quality data. Only when data / contacts can be segmented can you truly send targeted communication, one message isn’t suitable for all!
One of our most surprising discoveries was finding an email marketing database that didn’t have an email address for every contact, maybe the result of a mass compliance data purge perhaps, but worrying that potentially hot leads have not been contacted using the appropriate channels. Receiving inappropriate / mis-targeted mailings is a result or poor data quality, or sometimes laziness.
This data set is often subject to lengthy data cleansing and de-duping processes in an effort to achieve really good quality and therefore useful data.
One question we love to ask is ‘why do you capture that data?’, answers typically include ‘because I was told to’, ‘we’ve always done it this way’ or ‘that’s how the system is setup’, but quite often not the answer we’re looking for. There should be a clear reason for any data to captured, and a clear process for how it’s going to be used. If you’re not sure why data is gathered we recommend reviewing your data sets and processes to learn about what data is captured and how it’s used, or better still, how it should or could be used.
A data mapping process will help to identify problems, a detailed review will highlight gaps where important data is missing, bloat or unnecessary data, confusing or contradictory data, or legacy data that has no use.
It’s really important that you have confidence in your data quality before thinking about automation, otherwise you’ll miss valuable opportunities, potentially damage existing relationships or worse, fail compliance.
You should easily be able to extract a list of attendees from last years symposium and segment them into members and non-members for example, if not then you’re not quite ready to start automating your processes for next year.
Whether you’re looking to automate a process after a user registers for an event, like sending reminders and followup surveys, or perhaps wishing to keep in touch with loyal customers, you can be more agile, targeted and efficient if your data quality is fit for purpose.
We couldn’t finish an article about data without mentioning GDPR, the reality however is that much of the legislation has always been in place under older DMA, ICO and data protection guidelines; GDPR has simply reminded us of their importance.
Some of the common compliance issues we come uncover with data include:
- Being unable to trace the original source of your data
- Knowing which version of your policies a data subject has accepted
- Where data is stored, especially when using cloud services, ie Office 365, Sharepoint, Dropbox
- Not including offsite backups in risk assessments
- Data security not being taken seriously
- Knowing exactly who has access to your data and how it’s used from department to department
- Treating all data equally
- Fragmented processes and policies
- Outsourcing responsibility
- Keeping legacy data.
I wrote this article whilst working at Granite 5.