With all the talk about the accuracy of algorithms, our research shows decisions can be improved by up to 28% with high-quality data – leading to better targeting and more sales. But the trouble is many companies don’t have high-quality data; in fact, 44% of the projects we receive at Artefact involve cleaning up data-sets that are incomplete, partially corrupted or uncharacterised.
Here are five steps brand owners and other organisations can follow to improve the quality of their organisational data:
Take time to identify the problem areas
Instead of trying to solve problems, start by understanding them. Very often, data quality issues are deeply rooted in departments. So, it’s important to use an “issue-driven path” to define problems, collect and understand the depth and root causes of every data problem.
As an example: one client blamed bad data quality for the poor results of their CRM retargeting use case. But when we looked at the scope of the case, we found they were using the wrong Google Analytics activation group. The problem wasn’t data quality but a misunderstanding of the different data-sets.
Designate a data steward to solve problems
After identifying where the data quality problems lie, identify a data steward who can find ways to solve these challenges, manage, control and monitor operational data.
A data steward implements all necessary data action and remediation plans. They assure the quality of the data produced (gaps between different sources and systems, completeness, freshness, etc.) and document all data (sources, types, quality, reliability…) so it can be securely stored and properly accessed and used.
It isn’t easy to find a qualified data steward. One way to accelerate the creation of the role is to start with an external consultant, then train an internal resource to take over.
Build a data single point of contact (SPOC) team
Although the data steward can manage data quality, you need support from every department in your organisation to ensure the success of all data quality projects. A strong network of data-savvy SPOCs can provide vital and timely information, documentation, and insights to the data steward about what’s happening in each of their departments.
The data steward should also work with your data SPOCs as “a feature team”, inviting contacts from relevant departments to work in agile sprints with clearly defined objectives. It is important that SPOCs keep their departments in the loop so that the full data journey is transparent, understandable and easy to access.
Adopt a use case-driven data quality approach
When it comes to data organisation, most marketers have a backlog and don’t know where to start. They should define their sprints based on the use cases they can improve or unlock with better data quality. Here’s why:
– You need to prove that data quality is not just “nice to have”. By reporting business successes and showing related sales or savings in all your departments, you’ll create momentum around data quality.
– You need to be concrete and pragmatic. You can’t view data from a lofty point of view; to solve data quality problems and organise sprints around use cases, you need to get down to the operational level and engage with SPOCs and all stakeholders.
– You need to understand the current and future data needs for each use case. That way, instead of just solving for the current problem, you can set up data quality for the long term, ensuring the data steward is aware of future data needs from all SPOCs to keep relevant departments updated. That’s crucial, as many brands have data quality projects due to short term approaches focused on a single data quality problem.
Acquire data quality tools to sustain and facilitate this approach
Once you have a community of data quality experts across your organisation, empower them with the right methodologies and tools.
However, don’t equip your organisation with specific tools for detecting, assessing, cleansing and remediating data quality issues at first. On the contrary, start with manual but well-defined processes to restore knowledge of your data heritage and boost your understanding of the root causes behind your data quality problems.
Then prioritise a data domain (client data, sales data, etc.) that’s crucial for your business – but which doesn’t have too many complex issues – and define cross-team processes to detect when there’s a data quality problem and how to solve it. For instance, you could build an in-house dashboard to regularly verify the completeness and validity of your client data, comparing benchmarks such as email addresses or number of website visits.
Once you’ve achieved a more advanced level of data quality maturity, you can invest in tools such as Talend or Attacama that can help you automatically detect, profile and build remediation plans. Dedicated tools like these plug into your data ecosystem (data lake, data warehouse, etc.) and turn it into a real control tower for data quality.
Left unattended, bad data can kill companies’ growth ambitions – but it doesn’t have to be this way. Marketers can easily start to clean things up with a few tactical and pragmatic measures. The most important thing is to start. As Chinese philosopher Lao Tzu said: “A journey of a thousand miles begins with one step.”