In the realm of Big Data, it can be easy to become overwhelmed by the amount of information there is to test and dissect. Yet, complicated as the process might be, there is an increasingly strong impetus for forward-thinking business leaders to mine and analyze their data for any points that might be actionable and valuable to their company.
To this end, research reveals that 53% of companies are planning to implement policies that help them improve their Big Data analytics. Though most intend to do so to improve warehouse optimization, others are taking steps to find points and places within their data that reveal ways to improve customer interactions and predict future maintenance issues.
With so much information at our fingertips, determining where to begin with data testing can be a challenge within itself. How do you begin and in what pattern do you continue? Is it best to organize the data in rows or columns? Do you apply filters and only test a set percentage or do you consider the entire lot? Then, once you get going, what is the proper protocol for ending the testing?
These are just some of the questions that today’s analyst might be wondering when faced with the time-consuming task of finding value within a seemingly endless sea of numbers and letters. As such, today we’re taking a look at five strategies that can make the data testing process as simple and successful as possible.
1. Keep information and templates consistent
Especially when data testing within an agile environment, there are often instances where not everyone involved in the project is on the same page, literally. If you’re making changes on your end and only communicating those via emails or online messaging, it can be easy for team members to lose track of the updates, perform additional testing on top of your efforts, and cross many lines in the process.
It’s essential that everyone on your team is working from within the same template. In some cases, this document can take on other names, including requirement document. Regardless, it is essential that every team member is looking at and reviewing the same version, not one saved on your personal desktop that you continuously open, access, save and close without communicating to anyone else what you adjusted.
When a single template is used as a reference, every team member is made aware of updates in real time, thus helping to avoid duplicate work. In some instances, a central requirements document might not be available, and in these cases, additional workarounds will need to be set in place to ensure that as lines are coded and bugs are identified, all team members are made aware of the changes. This is where it can help to partner with a software testing service, such as that provided by QualityLogic and others, to ensure that testing and development efforts are performed in a secure, flexible environment.
2. Create a robust test plan
While it can be tempting to skip this step and dive straight into the process of data testing, it’s important to create an initial test plan to ensure that all team members are working toward a common goal. Within this plan should be set parameters on precisely what data testers should be looking at and testing, along with which metrics constitute a passing or failing grade.
When one takes the time to create a simple test plan, it can help to streamline efforts and create clearer dialogue around expectations, time limits and individual duties. Not only should testers have access to this plan, but it should also be made available to in-house programmers, business leaders, managers, test leads and more.
As a general guideline, such a plan should include sections on what to test, where to test and how much of the data to test. Concerning what to test, one person might want to only test that the data is entered into the correct format. Another might want to verify that is is actually accurate. Still another team member may want to see that the information is both correct and entered the right way. Communicating these viewpoints at the forefront is essential, as is coming to a general consensus for moving forward. If this isn’t achieved, one tester could be passing a row of data as approved simply because it follows a set template, while another is pouring over every detail to make sure all points are accurate.
Then, the plan will need to establish where the testing will take place. Often, a tester will utilize a snapshot of the environment as a metric against which to perform all testing activities. Where is this snapshot coming from? Does it originate with the dev.com site? If so, this is what you will use to ensure that the data values you’re checking are valid.
This is where SQL statements can come in handy, as a snapshot can change over time and require further testing updates. For instance, you may be testing against a snapshot of an employee list, wherein one person is listed as present in the company on a temporary intern basis. When you refresh that snapshot, the person may no longer be working with the organization. In this case, you’ll be able to receive a new test case if the SQL statement was in place prior to the change.
3. Know how much to test
There is no set rule for determining how much of a dataset a company should analyze. Some might be comfortable reviewing 10% of one, whereas others may require testers to take a look at a majority of the file. In this case, take a look at the data as a collective unit and consider the amount of variety present within it. Can you break it up into sections based on information provided? If so, can you then consider which sections can be dismissed?
Consider, for instance, if you are testing your company’s payroll system. You might find that you can break the data up into benefit types, company codes, retirement plans and other employee-specific data points. You may examine all or some of these categories and find little to no variation in the data past a certain step. This is where you’ll know to stop.
4. Know when to take notes
Detailed and accurate note-taking is a vital part of the testing process. Yet, it might not be feasible for someone to track every change or concern as they come along. Still, if something pops up that appears to be an abnormality, a tester should note it right away. A large part of the testing process is looking for diversions from the norm. In other words, testers should be on the lookout for data that simply seems out of place or doesn’t quite fit into the template.
Keep in mind, however, that it can be easy to fall into a rabbit hole of investigation that could ultimately end up leading nowhere. In fact, recent research shows that data scientists report massaging and cleaning data to weed out irrelevant or complicated information is the most time-consuming part of their jobs.
Testers should give themselves a reasonable amount of time to chase down answers and make notes of odd formatting, spacing or other elements, and then move on to continue the testing process. If you spend more than 15 minutes examining a particular data point only to still come up empty-handed and scratching your head, find someone else within the testing team to give the data a second set of eyes.
5. Keep the bigger testing goal top of mind
Though it’s important to look out for oddities, remember to keep the bigger goal in mind. If you’re comparing your system’s data against a snapshot or data extract, for instance, you might not care as much that the data was physically entered in a certain format, given that it’s able to be pulled accurately. For instance, if your testing reveals that someone in your company is still receiving health benefits though he or she is marked as retired, you might immediately spring into action and spend half the day investigating the error.
Yet, if your extract matches the current system, you can continue with testing. The fact that the health benefits data hasn’t been updated on one end isn’t a concern to your testing method. Rather, you’re only seeking to confirm that the data in the system matches that on the extract and if both ends check out, you can go onto the next.
More Accurate Testing Starts With a Clearer Vision
Though there can be myriad points to be investigated in a data testing environment, it’s important to ensure that all team members are up to speed on the process and understand what is expected of them. When employees communicate with each other, work off a common, shared template and reference a central test plan when performing work, chances are high that their efforts will be in concert with one another, not working against each other.
This streamlined process is only helped by testers who are taking accurate notes, checking inaccuracies in a timely manner and staying true to the task throughout the process. Though this list isn’t exhaustive, it serves as a strong jumping-off point for novice and experienced testers alike to make their testing efforts as successful as possible.