Top 5 Reasons For Data Science Project FailurePiyanka JainBlockedUnblockFollowFollowingMar 6Whether you are a seasoned data scientist or a business executive with a significant investment in analytics, chances are you’ve seen the powerful impact of analytics as well as the failures.
So what drives an analytics project’s failure or success?Data Science projects fail when they produce no actionable insights.
Even a seasoned analyst’s efforts to coax insights from the data can be futile unless a number of key factors come together to increase the odds of success.
Let’s talk about the top 5 reasons analytics fail.
1.
Starting with data instead of the question.
The most common misunderstanding about analytics is that if you look at data hard enough, you will find insights.
Staring at daily dashboards in the hope that insights will miraculously reveal themselves is often overwhelming, confusing and unsuccessful.
Successful analytics start by identifying the question you’re trying to answer from the data.
For example, if site conversion is an issue, instead of studying your website data hoping to find reasons for low conversion, narrow down your efforts to a specific question.
In this case, it might be “How can we increase conversion from 23% to 26%?” This approach allows you to focus on finding actionable drivers of conversion that can have impact.
2.
An exploratory approach to analytics.
Once you have identified the question you are trying to answer, do you explore all the data at hand in the hopes of finding insights or do you identify which data to study by using hypotheses as guard rails?The exploratory approach often fails to find any insights and if it does, is a lengthy process.
On the other hand, using hypotheses to narrow down both the scope of the project and the data set needed, leads to the answers quickly.
This process also generates secondary questions to ask data to further refine the insights.
In our example, the hypotheses might involve certain pages or experiences thought to be driving lower conversion.
These hypotheses are then used to identify the data needed to find segments of low conversion, and, once proven, address them.
3.
Weak hypotheses.
A data science project can still fail even when it begins with a business question and a structured approach for analysis if the hypotheses used to narrow down the scope of the problem are weak.
Weak hypotheses result from failure to follow due process with the right stakeholder.
4.
Unengaged or absent stakeholders.
Successful analytics projects deliver actionable insights that are then used to make changes in the product or customer experience and ultimately drive business results.
Any successful data science project, therefore, requires active engagement of the right stakeholders — the decision makers and owners of the business processes involved in the problem being studied.
Working with the wrong or absent stakeholders leads to weak hypotheses, long cycles to analysis and wasted or no insights.
In the conversion example, the product manager responsible for conversion, the product dev team making site changes, as well as the analyst need to be involved to make sure any actionable insights are acted upon promptly.
5.
Inaccessible or bad data.
Lastly, analytics can fail even after following a hypothesis-driven structured approach with involved stakeholders if the organization doesn’t have easy access to clean and reliable data.
The data needn’t be perfect for successful analytics, just cleaner and with fewer data issues.
Data maturity is thus a prerequisite for analytics maturity.
Thank you for reading my post.
I am passionate about using data to build better products and create amazing customer experiences.
I write about my learnings at LinkedIn and at Forbes.
To read my future posts, simply join my network here or click ‘Follow’ or join me on Twitter.
About Piyanka JainA highly-regarded industry thought leader in data analytics, Piyanka Jain is an internationally acclaimed best-selling author and a frequent keynote speaker on using data-driven decision-making for competitive advantage at both corporate leadership summit as well as business conferences.
At Aryng, she leads her SWAT Data Science team to solve complex business problems, develop organizational Data Literacy, and deliver rapid ROI using machine learning, deep learning, and AI.
Her client list includes companies like Google, Box, Here, Applied Materials, Abbott Labs, GE.
As a highly-regarded industry thought leader in analytics, she writes for publications including Forbes, Harvard Business Review, and InsideHR.
.