Achieving good enough data quality in your knowledge base – by Michelle Knight
Michelle Knight | August 6, 2021
Also, it gives you a return on investment (ROI), such as reduced time spent on queries and handling requests upon the first contact. On the other hand, bad self-service KB data means an organization’s customer support waste time responding to routine questions and have less time for complex customer requests.
Knowing how to determine what counts as good enough data quality and how to measure and improve it increases customer and support staff usage of an organization’s self-service KB.
This article discusses the following topics to achieve good enough data quality in your KB:
- What is good enough data quality?
- How does bad KB data quality impact customers?
- How does good KB data quality impact customers?
- Four ideas on achieving good enough KB data quality
What is good enough data quality?
Good enough data quality lets customers (readers) use and trust the KB’s article’s data, facts, or statistics collected for reference or analysis. This information may or may not contain numbers. Instead, article data may contain steps, text, hyperlinks, or images that form a knowledge entity.
A knowledge entity translates into either a true or false statement for a particular business context. For example, when this article states KB, it equals knowledge base, a true condition. Information has good enough data quality when the most significant knowledge entities meet enough truth conditions required to do business. In the example, good enough “KB” data quality keeps the article readable while saving the writer time.
After identifying business requirements and critical knowledge entities in a knowledge base article, rate the data quality along six different dimensions:
- Accuracy: KB article has true entities with correct formats. For example, this article breaks down into five different sections with five corresponding headers.
- Completeness: The KB base articles contain all the facts necessary for an entity to be true. For example, if an article instructs a customer to open a folder on a computer desktop, the full name of that folder appears (preferably with a screenshot).
- Consistency: Any place in an article that repeats data for an entity does so, in the same manner, to make that condition true. For example, if a writer starts to sequence a list of steps, the following steps will have a number.
- Timeliness: Critical KB entities match the current state of the truth condition. For example, if a reference’s URL changes, the URL changes to match the new address.
- Uniqueness/Deduplication: Each knowledge entity has a single meaning with the same attributes. For example, a navigation bar helps a user move around the KB. Therefore, duplicating the “navigation bar” concept to describe how a user searches would make the “navigation bar” entity false.
- Validity: The knowledge entities in an article make sense and satisfy the purpose and context. For example, an entity of a chocolate chip recipe represents invalid data in an article about categorizing a KB article.
The knowledge entity’s criticality and context determine how and when to fix KB data quality issues. For example, a missed period at the end of a sentence may be ok added later; a misspelling on a company logo in the KB header demands immediate attention and resolution.
How does bad KB data quality impact customers?
Bad KB data quality wastes time, jams resources, and discourages customer use of the KB. Take the Emergency SOS function on the iPhone, for example.
The SOS function, which is turned on automatically upon purchase, calls 911 and your selected comments to know about trouble. Unfortunately, this feature turns on very quickly by mistake, as my mom learned.
As I receive an emergency text from my mom, she embarrassingly searches the internet on how to turn it off. Finally, she opens the Apple support document and sees, buried down the page:
After following the steps several times, her phone will not end the SOS message. After five emergency SOS messages later, my mom texts, “I am fine. My cell phone just went nuts and is calling 911 continuously. The phone doesn’t work.” I get another SOS message from her after that.
My mom looks again on the internet and finds this response in the Apple community forum:
My mom turns off the Emergency SOS, and my phone falls silent.
Look back at the Apple KB article and find it lacks adequate accuracy, completeness, and timeliness. For example, the article forgot to mention which buttons to release quickly and how long to hold the side button.
This bad data impact spans beyond mere embarrassment. The 911 call centers say that 10% of the total 570,000 (57,000) calls happen accidentally, and most of these come from the iPhone.
How does good KB data quality impact customers?
Good enough data quality in KB articles, serve customers and organizations well. A customer gets a timely and trustworthy answer, and the organization enhances its reputation.
The U.S. National Library of Medicine (NLM) exemplifies good quality as shown in the screenshot below or at this link:
The NLM KB article, geared to the public consumer, explains what the library can and cannot do accurately. The KB article lists consumer and research portals, opening to the correct sites when clicking the link, making the KB article complete and timely. In addition, the article presents valid information about each resource that its visitors can verify.
As a result of good data quality, including its knowledge base articles, the NLM captures the ear of health professionals and general consumers. The Medical Library Association recommends the NLM. The American Medical Association (AMA) partners with the NLM and cites NLM studies and standards.
Users search the NLM “billions of times per year by millions of people.” The most recent statistics, last reviewed July 28, 2021, show over 21 million web page views per day. Consequently, customers value the NLM KB, and the organization takes care of the simple questions.
Four ideas on achieving good enough KB data quality
Achieving good enough KB data quality within a reasonable time and budget can be straightforward. Check out the ideas below:
Idea 1: Write a KB data quality strategy
A KB data quality strategy guides managers, technical writers, and customer support about aligning its construction and usage with the business strategy. The KB data quality strategy informs what data customers need from KB articles, what critical knowledge entities to include in data quality monitoring, and how to assess and improve KB data quality. Such a strategy does not need to be complex.
For example, the state of Oregon has the data strategy: “A better Oregon through better data.”
This data strategy document guides different Oregon departments in governing and managing data, including data quality. So, various Oregon offices can reuse this data strategy document when coming up with a KB data quality strategy to create better customer self-service.
Idea 2: Start small with easy victories
Choose a section of your knowledge base or a few critical knowledge entities to review for data quality and update as needed. For non-critical KB issues, highlight them and come back to them at a later point.
For example, you can check hyperlink accuracy in a few articles that you choose based on your KB data quality strategy. In addition, automated tools exist, like the World Wide Web Consortium (W3C®) link validator, to quickly get to these easy wins.
Idea 3: Improve KB data quality incrementally
Improve KB data quality incrementally, along with content lifecycle (CLC) tasks. For example, during the research phase, ask questions about the level of data quality required by the KB and develop a quick test for critical data quality elements in the content. Then, plan for resources during the planning phase and run the test when revising article contents.
Idea 4: Regularly monitor data quality of KB articles
KB data quality can degrade over time. Hyperlinks change, factual accuracy may need revision, and information added to complete a KB article. Planning to monitor article data quality through sampling, reviewing previous content issues, and batch fixing issues makes sense.
You want to avoid devoting an entire day, week, or month to getting data quality in your knowledge base articles to an acceptable level. Instead, streamline data quality tasks regularly, so you find and fix factual issues before your customer sees them.
Good enough KB data quality supports the business, making documentation reliable and valuable for customers. Also, it increases the value of a self-service KB, having your customers coming back and using it. Finally, it saves your customer support time from answering routine questions and frees other resources from accidental errors.