The Real Meaning of ‘Bad Data Quality’ and Why it Matters

We know that data quality has a significant impact on businesses.

The likes of Gartner, Harvard Business Review, Forbes and IBM have written about how costly bad data can be, and talent teams are not exempt from that impact. By our calculations, recruiters spend at least one day each week—or 20% of their time—doing manual tasks, including looking for data, cleaning up mistakes and checking for duplicates. Beyond the wasted time and effort, and the gradual decline of the recruiting team’s morale, this has knock-on costs. Lost candidates due to negative experiences or faster competitors, increased time to hire, and continued pressure on existing employees.

There are more negative impacts. 70% of organizations are increasing investments in talent analytics, but only 12% feel like they're getting results -Gartner-. This is largely due to the lack of good data. In talent teams, recruiters lose trust in their software, which leads to low adoption, automated processes are more at risk of handling candidate records incorrectly -hello compliance issues- or not at all, and you can forget about unbiased, accurate machine learning initiatives.

But what does “poor data quality” really mean?

Is it missing or incomplete records? Is it incorrect information, or data stored in the wrong format? How about out-of-date data or no history at all? Does it mean duplicates in one system, or inconsistent information across many systems? Is it about access to the data?

Not all data quality issues are equal. Once we know what we mean by “poor data quality”, we can identify the root cause, understand the impact of the problem, and act accordingly.

So let’s get more specific. You can find many definitions of data quality online, such as this one. Here’s a synopsis of the seven types of data quality, tailored to Talent Acquisition.

What causes bad recruiting data and how do we fix it?

Aside from accessibility, the other types of bad talent data can be caused by people either typing in the wrong information, or not typing in anything at all when they should. This is going to happen, people are busy, and most of us aren’t trying to intentionally mess up data and make our lives more difficult. Machines also make mistakes. Integrations might not be robust enough, parsing technology could be causing issues, or enrichment engines might not have high enough match rates.

We can do a few things here to fix data quality, and each could be an article in their own right. For now, let’s bucket them into two themes:

Choose a talent acquisition provider who values talent data quality and builds solutions for it.

• Look for solutions with deep integrations, so that the correct recruiting data is shared across systems at the right time.
• They should have solid underlying technologies in a dedicated talent data platform, such as accurate multi-language resume parsers, enrichment engines that can augment profiles from a wide range of sources, and data standardization methods (often using machine learning) to normalize and maintain clean data.
• A user-friendly interface helps. This incentivizes recruiters to maintain data quality, for example with autocomplete fields and in-app nudges, easy deduplication and re-enrichment features, as well as features that compensate for poor quality data, such as fuzzy matching on search and filtering.
• The system should have good reporting capabilities. You can’t fix what you can’t measure, and being able to track the state of your data is the first step to acting on it.

Choose a partner who can help you implement and maintain high quality data flows.

• Work with a partner who can help you set up the ideal data flow, which will help talent teams work in the most frictionless way, across your systems, and optimize for data quality. This includes setting up automated workflows, ensuring integrations are in place, and that enough, correct, up-to-date data is in the system before you onboard your users, so they see value from the start.
• With data flows set up, the partner should help you train your teams to use the systems in ways that respect the data flows, and crucially, help them understand why it’s beneficial for them to maintain data quality by doing so. Our research shows that data quality is more important during the adoption phase than general use, so it’s important to get your team thinking about and interacting with data in the right way.
• Once the system is running, ask the partner for a regular data quality status, for example during quarterly business reviews. This gives you extra visibility beyond the in-app reports and a focal point to act to data quality issues.
• You should have the option of cleaning up your data retrospectively, because data quality can deteriorate over time. This could mean running ad hoc projects with your partner, such as a bulk parse of previously unparsed resumes, re-enriching contacts, and revisiting the integration and data flows set up during implementation so they are optimized as your needs change.

But it is worth fixing?

In short, yes. The longer answer is, it depends—on your goals, and the type of talent data quality you're talking about. The aim is not to get to the nirvana of perfect data quality, as at some point the effort to fix it will outweigh the cost. The aim is to have high enough data quality, so your teams can work effectively to hire the best talent in the shortest time. For example, arguably, your contact records don’t need to be 100% complete. As long as the key fields are fully complete—and assuming they adhere to the other data quality types—your teams can still contact candidates and have the context they need to convert them.

So first, work with your team and partner to uncover the pain points caused by poor data, and prioritize them based on how much negative impact they are having.

At Beamery, we focus on data quality, through our Talent Data Platform, which connects, completes and contextualizes information across the Talent Operating System. For a deeper dive into data quality, check out our ebook on Data: Moving from Defective to Dependable.