Light My (Analytics) Fire: Five Steps to Get Started with Your First Analytics Project
“Analytics” is among the most prevalent buzzwords in business today. Everyone wants to make data-driven decisions, get insights faster, and generally improve business functions. If you’re trying to get started in analytics, I recommend beginning with a small pilot project. Do this to get things moving, build confidence, and generate momentum. Starting small (and getting one win) creates a kind of muscle memory that you can leverage to further either a more formal analytics function or more individual analytics projects.
Bottom line: Taking the first step is often hard. Start small, but make sure to start! In the interest of getting started, I have put together five general steps below. Think of these as a roadmap to victory. Looking at the data science process as a series of steps will put you in a good position for future success.
Data Analytics Defined: A Word about Terminology
When starting on your analytics journey, try not to let the terminology overwhelm you.
From working within various business intelligence and analytic functions over the years, I’ve come to realize that “analytics” can mean many things to different people. I define the term as simply an intentional, structured effort to analyze data followed by a process to use the results of that analysis to derive relevant insights. These insights then inform business decisions.
You can think of analytics as covering various terms: data analytics, data science, and business intelligence — just to name a few. If you proceed using the steps below, your choice of terminology won’t matter. The results (and your happiness with the outcome) will be the same.
The Big Picture
Think of an analytics project as a machine with five simple parts. Each component works a bit differently, but all must work for the machine to function. What’s more, there is an order to how things work. Each of the five steps below represents one of the parts.
Step 1: Identify the Business Problem
Analytic project success means beginning with a meaningful, well-defined business problem.
Identifying a business problem has been my first step for every project. Consider that answers are only worthwhile when you have a meaningful question. Formulate a question for your analytics to answer. The more concrete and business-relevant the question, the better!
Your business problem mainly depends on the nature of your industry, your company’s strategic goals, and/or your departmental initiatives. Isolating and carefully stating your business problem will likely focus on either making something good even better or making something bad hurt a little less.
If your business problem is vaguely defined, resist the urge to keep going. Instead, talk with both your coworkers and stakeholders to firm up the business problem. Once you have a business problem defined, pitch your problem statement to someone whose opinion you value.
Example: a (Relatively) Bad Problem Statement
“Management wants to ‘move the needle’ on a few major KPIs to get some quick wins. Work your analytics magic! Let’s see what you find.”
This type of problem statement will doom you to inevitable failure. Which KPIs are most important? What is “management” expecting? Would a 5% improvement equal data science success? This problem statement needs more specifics.
Example: a (Relatively) Good Problem Statement
“The supply chain VP wants to know what potential effect a change in vendor will have on our ability to ship products overnight to our biggest customer, XYZ Company.” The VP wants to consider another vendor for cost reasons but would not do so if more than 5% of shipments could no longer be sent overnight.”
This problem statement has a lot going for it. A specific, named person whom you could interview (the VP) wants to know about something quite clear and quantifiable. Particularly nice here is the “what is X’s effect on Y?” being clear right from the start. You also know that the VP has a specific testable threshold in mind (the 5%).
Step 2: Know Your Stakeholders, and Define Success
Talking to your stakeholders and incorporating their feedback increases your chances of project success.
Stakeholders assist by generating momentum. Call this momentum “enthusiasm,” “buy-in,” or whatever you like. Stakeholders also provide you with their expectations. Knowing what someone expects helps you manage (and sometimes reign in) those expectations.
Pinpoint the stakeholders in your organization who need to be connected to your project. These stakeholders may be decision makers, project sponsors, recipients of the project’s insights, or a combination of all three. Once you know names, it’s time to start talking.
How do you find the stakeholders, obtain their definitions of “success,” and get momentum for your project? Interviewing. By this, I mean conversations. During the early project stages, your talks will resemble the question and answer format of an interview. Later, your interactions may resemble status updates, information sharing, etc. The important thing is to communicate!
After completing Steps 1 and 2, you will know two vital bits of information: 1) what the problem is and 2) who is connected to the problem.
Step 3: Inventory Your Data
Data is the lifeblood of analytics, like fuel in an engine. Without it, you aren’t travelling anywhere.
Understanding your data comprises a few simple steps:
1. Do you have the data points needed to solve the problem identified in Step 1?
a. Note: The stakeholders from Step 2 can often help with data sourcing.
2. Are the data in a format that makes analysis possible?
Cataloguing your data is a crucial step. Having insufficient data is a nonstarter. Imagine defining a key metric as numerator / denominator, only to realize that you don’t have reliable data for the numerator. Determining data format is equally crucial. Format will control how much effort is required to analyze the data.
If you answer “yes” to both questions on this step, you are ready for Step 4.
Step 4: Assess Your Skills and Choose Your Tools
Bite off only as much as you can chew. This means an honest evaluation of both your people and your tools.
For your five-part data machine to operate, you will need a few skills:
1. Domain expertise to properly understand the business problem identified in Step 1,
2. Technical knowledge to source, clean, and prepare the data from Step 3, and
3. Tool-specific abilities to visualize, aggregate, or otherwise report the data.
The presence or absence of these skills will determine your options. If your data exists in a SQL database, you will need query skills to extract it. If your data requires cleaning prior to use, you will need data wrangling skills to produce a clean dataset. Success in Step 4 usually comes from harnessing different talents on your team. Even if you completely lack one of the necessary skill sets, you always have options. These include training existing staff and/or bringing in a pro to get you over the hurdle.
Data analysis is not tool-dependent, but it is knowledge-dependent. I’ve seen sharp individual analysts, find breakthrough insights using only Excel. Yet, I’ve also seen teams of analysts fall short of finding insights, despite having access to cutting-edge big data platforms and costly business intelligence tools.
Step 5: Learn from Your First Attempt
A huge part of doing data science well is the process of refinement. After your first project concludes, evaluate yourself. Congratulate your team on the positives but learn from the negatives.
You defined a business problem, worked effectively with stakeholders, managed your data, and took on a project appropriate for your team’s skill. You directed that energy into an analytic project and came up with some results? What now?
One of the biggest mistakes analytic practitioners make is failing to evaluate themselves. What did you do well? What could you have improved? How did you measure business impact for your analytic solution? Confronting these questions and documenting your experience will help you seal the gaps in your process and will make your next analytic project easier to conduct, better integrated with your stakeholders, and better aligned with measurable business impact.
Maybe reiterate your “bottom line” from above as a wrap-up here
Remembering the bottom line: Look at the data science process as a series of steps. Doing so will help you overcome the hardest part, which is getting started in the first place. There’s no time like the present to get out there and take your first step to success!
