BY JIM SAWYER – CHIEF SCIENTIST
We have a rule of thumb around here: You have to consider at least 3 options.
Any meaningful problem you are considering, any design decision you are contemplating, any data study you are conducting, any software feature you are creating… all of them need 3 options. I guess this can even apply all the way up to life choices and career decisions as well. What are the 3 options you considered before making your decision?
It’s easy enough for Data Scientists to come up with “the” way to attack a quantitative problem. We’re smart and creative and innovative – it’s what we do. Then our immediate reactionary temptation is to charge forward, to dive in headfirst and start slinging SQL, or build us some fancy models, or crunch and munge and grind some data – all because we are intellectually driven to get moving, to get at the answer. We want to do something. We want to act.
What we should do is pause and reflect: What are my other options? How else might I approach this problem? What are the pros and cons of these alternatives relative to my first idea?
This often comes into play when a client makes a direct request to our Data Science team: “Can you create a spreadsheet showing the 3-year trend of sales for product line X in markets A, B, and C?”
Well yes, we could do that. But what are our other options?
Let’s start here: Why is the question being asked? What is the business question that we are trying to address? What else could be folded in to the analysis to help generate insight into that specific business question?
For example, why those particular markets? Would it be helpful to compare against other markets, or perhaps against a view of all markets? How about other product categories? How about other metrics such as profit margin or units sold?
And who is asking? Who is the end consumer of the information? How will it be used? How will the results be distributed? Does it stand alone or will it become a part of a larger analysis or presentation? If so, what are the objectives of the larger endeavor?
Then there’s the delivery mechanism. You asked for a spreadsheet. But is a table of numbers the best way to convey the information? It might be, depending on the audience and the purpose. But would it create more value if a chart or other graphic were used instead? And what type of visualization – would a simple line chart suffice, or are there other creative treatments that would deliver the results with more of an impact?
Or is it just as simple as fulfilling the request exactly as asked? That’s an option too.
In fact, “do nothing” can also be a viable choice once in a while. Is the thing “good enough” as it is? Does this problem absolutely need to be solved right now? Sometimes, what may seem absolutely critical to one person at one snapshot in time may be a much littler deal when reflecting on the bigger picture and the context for the work.
So the next time you find yourself facing a bright and shiny new problem or analytic study, and you are 99% confident that you know exactly the right way to proceed, please do me a favor and pause. Just for a minute.
What are your other two options?
The Other Two Options
BY JIM SAWYER – CHIEF SCIENTIST
We have a rule of thumb around here: You have to consider at least 3 options.
Any meaningful problem you are considering, any design decision you are contemplating, any data study you are conducting, any software feature you are creating… all of them need 3 options. I guess this can even apply all the way up to life choices and career decisions as well. What are the 3 options you considered before making your decision?
It’s easy enough for Data Scientists to come up with “the” way to attack a quantitative problem. We’re smart and creative and innovative – it’s what we do.[1]I know, we do a lot of other cool stuff too that the world really needs to know about. We’re so misunderstood. But stay with me on this. Then our immediate reactionary temptation is to charge forward, to dive in headfirst and start slinging SQL, or build us some fancy models, or crunch and munge and grind some data – all because we are intellectually driven to get moving, to get at the answer. We want to do something. We want to act.
What we should do is pause and reflect: What are my other options? How else might I approach this problem? What are the pros and cons of these alternatives relative to my first idea?
This often comes into play when a client makes a direct request to our Data Science team: “Can you create a spreadsheet showing the 3-year trend of sales for product line X in markets A, B, and C?”
Well yes, we could do that. But what are our other options?
Let’s start here: Why is the question being asked? What is the business question that we are trying to address? What else could be folded in to the analysis to help generate insight into that specific business question?
For example, why those particular markets? Would it be helpful to compare against other markets, or perhaps against a view of all markets? How about other product categories? How about other metrics such as profit margin or units sold?
And who is asking? Who is the end consumer of the information? How will it be used? How will the results be distributed? Does it stand alone or will it become a part of a larger analysis or presentation? If so, what are the objectives of the larger endeavor?
Then there’s the delivery mechanism. You asked for a spreadsheet. But is a table of numbers the best way to convey the information? It might be, depending on the audience and the purpose. But would it create more value if a chart or other graphic were used instead? And what type of visualization – would a simple line chart suffice, or are there other creative treatments that would deliver the results with more of an impact?[2]Quick sidebar on data visualizations: At a recent conference, I was incredibly inspired by a fascinating talk on data visualizations given by Julie Rodriguez from Sapient. You should buy her book. (And no, I don’t know her, so this is a totally independent pitch!)
Or is it just as simple as fulfilling the request exactly as asked? That’s an option too.
In fact, “do nothing” can also be a viable choice once in a while. Is the thing “good enough” as it is? Does this problem absolutely need to be solved right now?[3]Well, we’re hired consultants after all, so the answer is usually yes. Sometimes, what may seem absolutely critical to one person at one snapshot in time may be a much littler deal when reflecting on the bigger picture and the context for the work.
So the next time you find yourself facing a bright and shiny new problem or analytic study, and you are 99% confident that you know exactly the right way to proceed, please do me a favor and pause. Just for a minute.
What are your other two options?
Footnotes [ + ]
Tags: