Previous Related Posts
Identifying an Ambiguous Problem
The past two Data Duel blogs deemphasized technical ability and touted soft skills as crucial for an analytics career. My goal is to bring the discussion out of the theoretical and into a practical example from my own career where I applied those four soft skills in an analytics context. In fact, the lessons I learned in the following example are ones I regularly utilize 6+ years later.
To define terms, “Ambiguous Problem” is one which no one clearly defines and for which no one provides a clear solution.
Let’s go back to 2014. I’m working at a manufacturing/distribution company. Growth is starting to explode, and I’m working as the org’s only data analyst. Reporting into the SVP of Sales, my desk is on the sales floor in the middle of ringing phones and reps busy entering orders into our system.
After a few months I notice something interesting. Nearly every rep has an Excel spreadsheet called the “Sales Catalog” up when they’re on the phone, showing various items for sale. Sometimes the row says “In Stock” and sometimes in red it says “Out of Stock”. I also heard grumbling – the sheet isn’t right. They’ll tell a customer “Yes that item is in stock” but when they go to order it, the system denies the request. Yikes, not a great experience for the customer or rep.
The process to correct data errors was also bumpy. Excel only allows one person to update a shared network file at once, and that person is the SVP’s Executive Assistant. Reps would call or ping the EA, telling them what update to make to the Excel file. Then all the reps had to close & re-load Excel to get the up-to-date information.
As I noticed these issues stacking up, I heavily leveraged empathy and curiosity to understand what the reps wanted to accomplish and why we ended up in this rather inefficient place. I talked to multiple people across the organization – sales reps, sales managers, systems – to make sure I had a grasp of everything.
These conversations let me take an ambiguous problem and define it:
Reps can’t get accurate and timely in-stock data to their customers.
Crucially, no one told me about this data problem or how to solve it. It was up to me to define and solve it
Developing a Solution
With the problem defined, it was time to work on a solution. This is where organization became crucial. As you may have noticed, there isn’t a single solution to this problem. Instead, I needed to break it down into sub-tasks:
(1) Figure out where the true ‘In Stock’ data in the system is stored
(2) Create SQL script to retrieve that data
(3) Get that data into Excel for the sales floor (It’s OK to keep something in a format familiar to them, even if it’s not fully optimal)
(4) Make the report better! (Wouldn’t it be cool if instead of just saying ‘out of stock’ it said when it would be back in stock?)
(5) Discuss V1 with leadership & iterate as needed before launch
(6) Launch new tool with training/documentation
Each of these steps was non-trivial. I had to dive into our database and really understand how the items moved into and out of stock. I had to figure out how to write an accurate SQL script to replicate those movements. I had to figure out how to connect SQL tables into Excel and create a reliable pipeline. All while making sure I kept a similar form-factor for the sales floor to maximize adoption.
In the midst of completing each step, I made sure to understand the accuracy needed. “Good Enough” data wasn’t necessarily clear. For instance, I added in some buffer to what ‘In Stock’ meant due to how fast-moving the data was. Items went into and out of stock quickly. I wanted to minimize scenarios where my document said ‘In Stock’ and yet the system didn’t let the customer put in an order. Additionally, I needed to hedge on when an item would be ‘Back In Stock’ — more on this in a later post!
Critically, I also went through development cycle with leadership and other trusted Sales team members to make sure what I made would match their needs. They would see an early draft, give feedback and I would fix that before starting the cycle again. This is again where empathy came into play – I needed to understand their problem and make sure what I created actually solved it, rather than assuming.
Analysts provide massive value by identifying and solving ambiguous data problems. I learned that early on with this Sales Catalog example. I liberally applied each of the four soft skills to go from problem identification to problem solution:
Curiosity: Dug into what the reps were trying to do and what problems their existing solution created
Accuracy: Determined the tolerance of “Good Enough” data, both due to database limitations and hedging where I would prefer inaccuracy to rest
Organization: Broke down the problem into sub problems, which built into my final solution
Empathy: From start to finish, I made sure to listen to many voices across the team – both in understanding the problem and making sure my solution actually made their work lives easier
This same cycle has served me well time and time again in an analytics career. If you can proactively discover analytics problems and solve the important ones, you’ll quickly provide value to any company lucky enough to have you aboard.