Inductive vs Deductive approaches #statisticalThinking @muralCo #imagine2020

-
In this past month I've had a bunch of interesting interactions with strategists, consultants and statisticians. Quite tongue-in-cheek, but I thought I'd share this insight on #inductive vs. #deductive approaches to statistics :



Inductive vs Deductive

But more seriously, what does it mean to be inductive or deductive? I'm sharing my notes here, because it was a chance for me to experiment with the Mural service, having attended some of Mural #Imagine2020.

I'm planning to use #Mural on my next data science project and I can quickly see how this can be a great facilitation tool to work with problem owners and business leaders to bring design thinking and a clear analytical thought process to their ill-defined business problems.


Figure 1: Inductive

Let's have a hypothetical situation of a manufacturer of unique personal mobility devices. It's high quality, engineered product and battery-powered device that transport users through urban areas. Average retail price of USD$1,000 per unit. Think: electronic skateboard.

Here, we found out that there is an increasing trend of customer complaints. Customers are finding that the product doesn't meet advertised specifications and are unsatisfied with their purchases. Conversely, other competitors have been increasing market share by selling more products and have been stealing a march against us. In other investigations, we've discovered that the operations team runs out of important components in stock. 

This kind of inductive "bottom-up" approach (read from the left to the right) to derive a problem statement is great when you know something about the right hand side of the Mural. But you're still trying to understand the relationship between the moving parts.

Figure 2: Deductive

Same situation - this time we start with the overall problem statement and disaggregate it into its component parts. We use this when we have a very clear idea of the problem structure. Especially when you are using a mathematical relationship of a framework: Economic Value Added, Return on Capital Employed, etc.

In this case, you might see that once we disaggregated the overall problem, we've quickly honed in on the levers and identified some root causes. At this point we might pivot out and focus our efforts on confirming a new hypothesis: "Our electronic skateboards aren't price-competitive and aren't reliable because of poorly engineered and expensive motor components".




Comments

Post a Comment

Popular posts from this blog

Define: artificial intelligence

-
Image
Alas, that term "artificial intelligence" gets bandied about a lot in business and non-business circles. So what exactly do we mean by "AI"? And to make things more complicated - what do people mean when they say "machine learning"?  A no-nonsense guide to AI Quite simply: AI is a general label for all of the computerised systems that perform tasks in an approximation of human intelligence. You know that magic rice cooker? It might have a heuristic (fancy way of saying "a rule of thumb") on "how long to keep the water on the boil". It's not quite as sexy as "machine learning", but this approximation of how grandma used to cook the rice is as simple as saying, "use two cups of water for every one cup of rice".  The next level down is "machine learning", and I actually prefer to split this down into supervised learning, unsupervised learning and reinforcement learning. These are statistical methods where
Read more

Can you do this analysis in Tableau?

-
Image
If all you know how to use is a hammer, everything looks like a nail! I frequently get asked, "Izam, can you use Tableau to create a chart that looks like this ..." or "Izam, can you use Tableau to calculate this ..." Lately, I've been thinking about how the tools we use shape the way that we go about solving problems. And the above adage comes to mind: if all you know how to use is a hammer, everything looks like a nail! Excel and PowerPoint are great for writing documents and reports. Reports which go about answering the: "WHAT"  are we planning to do?  "HOW MUCH"  are profits? or "WHEN"  does it need to be done by? Tableau on the other hand, shines when we use it to explore the question ... "WHY"  is this trend happening? "How much are profits?" For example, with the Superstore sample data set, an analyst might be interested in building out a table that compares the key performance indicators of
Read more