This content is part of the Essential Guide: Everything to know about overcoming digital transformation challenges

Seven mistakes digital transformation projects must avoid

Every organization managing big data, employing IoT devices or pulling from a disparate set of information streams is looking at ways to put all of the data they have gleamed from their various digital transformation projects to good use. However, turning the basic data obtained from various digital transformation projects into actual insights is not a particularly easy task. Many pitfalls, obstacles, and misconceptions can cause digital transformation projects to fail. Here are seven common mistakes that businesses make when implementing digital transformation projects.

#1 Digital transformation projects that don’t have a well-defined objective

Without dedicated support from business decision makers and a clear purpose for the digital transformation project, any artificial intelligence or analytics attempt is doomed from the start. According to Sai Devulapalli, Data Analytics Product and Business Leader at Dell-EMC, “I’ve seen too many Hadoop projects that are embarked upon and then stall because they are not tied back to business outcomes.” For a digital transformation project to move forward, it should have at least two clear use cases in the event that one doesn’t pan out. It should also have sponsorship from lines of business within the organization that will directly benefit from the outcome of the project. Keeping the end goal clearly in mind is essential. “Don’t build a data lake and go fishing for use cases. Stick with the scope that’s been created.”

#2 Organizations haven’t realized that a problem will yield to analysis

Executives and upper tier managers are accustomed to making decision based on business intuition. That’s all well and good until a better option comes along. Data Scientist Neeraj Madan offers the following observation. “On various data science engagements, I have noticed that the clients are not always aware of the business problems that can or cannot be solved using data. I always encourage our clients to find solutions to their business problems using empirical data. Your intuition may give you a starting point. But if you are going to make a decision based on the intuition, it better be backed by evidence or data.” Working with a trusted data advisor and using appropriate methodology such as CRISP-DM helps separate the agenda-driven narratives from objective reality.

#3 Digital transformation projects use dirty, disorganized data

The insights gained from analytics are only as good as the data on which they are based. Sadly, data quality is often substandard. Alex Sabatier, Global Account Manager for Enterprise Sales at NVIDIA, notes that the first challenge is to clean and label the data. “Businesses want to create a data lake, but they end up creating a data swamp because they don’t know how to organize it. If it’s not a true representation, you end up with a false model.” It’s very hard for business decision-makers to learn to trust analytics and AI when it doesn’t align with the real world.

#4 Data science teams don’t follow the rules

Failure to follow best practices within a data science project and within the surrounding organizational hierarchy can spell disaster. Juan Vasquez, Communications & Data Analyst for the Operations Innovation team in the LA Mayor’s office, reveals his findings from a major analytics initiative. “I was surprised by the importance of following processes. If you don’t go through the right steps, you end up missing the outcome or the outcome simply doesn’t matter.” Following procedure applies to more than just the technical side where skipping important prep work such as cleaning and organizing data leads to poor quality analytics. It’s just as critical not to leave stakeholders out of the process. If the people who are funding, approving, and using the system aren’t in the loop, digital transformation projects may be abandoned, or adoption may be too low to achieve the desired ROI.

#5 Businesses expect too much from their data

While some organizations don’t realize the power of data their digital transformation projects have acquired, others overestimate its ability. Consultant Dr. Meltem Ballan, reveals her struggle with clients who want too much, too soon. “They think we have a magic wand and can change their organization very quickly with insights from the data. But they often don’t have the database to pull that data.” Her job is to patiently and clearly show clients where things aren’t working so they can come to more realistic conclusions and explore their options.

#6 Internal IT resists the assistance of outside experts

According to Ballan, consulting firms who come in to spearhead a data science project also find it important to clear up another misconception “Some people feel threatened by even having a consultant come and lend assistance. We aren’t there to take their jobs. We are there to improve their business.” Once the consultants pack up and go home, those employees need to know how to keep the analytics running, demonstrating their continued value to the organization. It makes good business and career sense to cooperate with consultants.

#7 Organizations misinterpret data based on wishful thinking

Even with good data, folly comes into play when humans start chasing the next shiny object instead of using common sense. According to Data Skeptic Kyle Polich, “Queries are one thing, interpretations are another.” Just because an online retail website gets a lot of hits between midnight and six AM from unemployed people doesn’t mean this is a new business opportunity. “It’s easy to deceive yourself and think that’s a target market to pursue. I wouldn’t expect any other audience to be online at that time.”

Making smart decisions starts with choosing the right digital transformation projects, taking time to prep the data, bringing help on board when needed, and taking a good hard look at the results to separate insights from the noise.

App Architecture
Software Quality
Cloud Computing