Best Practices To Improve Data Hygiene for Small Businesses

Common Data Hygiene Missteps for SMBs 

SMBs don’t have to solve all of their data problems at once, Schreiner says, and if they have to start with spreadsheets, that’s OK. They can take incremental steps to improve their data hygiene without delaying daily operations, and they can start by having a clear problem to solve. 

“‘What data do I need to solve that problem, and then, what can I do to clean it up?’” Schreiner says. “You don’t want to have a cleanup exercise that you then have to repeat for all of the new data coming in next month or quarter. You want to get ahead of that and have governance.” 

There are automated tools that SMBs can use to standardize their data, he adds. For example, if U.S. states are spelled out and written as abbreviations in your customer relationship management system, you can automate turning those inconsistent inputs into all two-letter abbreviations. 

EXPLORE: Read CDW's latest research report on AI and cyber risk. 

Another misstep is not properly understanding your data to know which data needs to be more readily accessible to your employees. Customer files may have a long storage life compared to last week’s sales data. But if you don’t understand how long you need to store something and whether or not it’s immediately useful, that can impact your storage costs and degrade the output of the AI tool you’re relying on. 

“You need to be thoughtful about what do you need and for how long. So, it starts with defining those things so you have a good understanding, and then you can put a strategy in place,” Schreiner says. AWS offers an intelligent-tiering function that can automatically sort data into the appropriate storage based on your retention policies. For example, data that needs to be accessed within the hour lives in higher-cost storage, while data that won’t be reviewed in the next year or so is archived in the lowest-cost storage. 

Practical, Easy-To-Follow Data Hygiene Guidance for SMBs 

While there are several industry-standard data governance frameworks that SMB leaders can seek out, Schreiner recognizes that many of them can be rather complex and overly technical, which can be intimidating for an SMB owner or a supporting IT professional. 

He offers more simplified guidance to help SMBs get started: 

1. Assess your business’s current state. Review where there may be duplicates or inconsistencies in your data. If you have not already done so, create a profile of your data: What do you have now, and which data is most impactful to your business? Tools such as AWS Glue DataBrew can help clean and normalize data in a more accessible way. 
2. Methodically clean your data. “Standardize your data as best you can, especially on the high-value data that you know you need to make decisions on, so that you can increase your confidence with that data,” Schreiner says. 
3. Set preventative measures so that your data doesn’t fall into disarray. For SMBs further along on their journey that have likely completed the first two steps, Schreiner suggests setting measures that will improve data consistency moving forward. 
4. Use automation to monitor your data. Automated tools can help detect anomalies in your data and alert you if conditions change over time. 
5. Clearly document the data about your data. “Where’s the data coming from? What’s your confidence level? What does it mean? Is this a customer record? Is it a person? Is it a company?” Schreiner says. “Having context about your data is very valuable when you’ve gotten to this stage, because then you can feed that into AI or into a data analytics platform to help it understand the data it has access to and help you develop better solutions or insights into the data itself.” 

Starting with this general base will help SMBs tackle what can be a very daunting aspect of data governance: unstructured data. While structured data lives in an easily located database or a spreadsheet with neat columns and rows, Schreiner says a lot of valuable information lives in unstructured data that can be found in emails, a locally saved PDF on a desktop or in online reviews that may be useful to improving an AI tool’s output.

Unstructured data is another category of data that SMBs should understand, down to how dated something is. For example, a marketing price list PDF from two years ago would not be relevant now, so it should be decommissioned or archived. 

“Unstructured data becomes a new area for value creation inside of SMBs, and AI is definitely helping get more value out of those sources,” Schreiner says. For example, an SMB leader asking an AI tool, “What are the top 10 things my business can do to improve customer satisfaction?” based on data extracted from hundreds of customer emails would help to potentially change the business’s competitive posture. 

The Interconnected Nature of Data Hygiene and AI 

Schreiner notes that most of the SMB customers that he talks to are interested in AI, which eventually leads to conversations about data. “They go hand in hand. Data quality isn't a one-time cleanup project,” he says. 

When friction arises in an AI project, Schreiner says, it’s almost always connected to faulty underlying data. 

“To avoid that disappointment, focus on the problem that you want to solve: Is it an important problem? How are you going to measure that you solved it?” Schreiner says, adding that SMBs should measure their averages before and after AI use for proper tracking. 

DIVE DEEPER: CDW research uncovers where organizations are in their AI journeys.

He also recommends for AI testing to take place in a real setting, not under the most ideal circumstances. When companies test AI with a perfect data set, they shouldn’t be surprised later when AI gets used in a real setting with imperfect data and performs less accurately. 

“I’ve seen it far too often, but it’s usually because the data that they’re testing with isn’t a very good representation of reality,” he adds. “I want customers to be realistic about the quality of their data and how it directly correlates with much value they’re going to get out of AI and how much they can trust it as they continue to do more complex tasks.”