If you eat food (or torture food labels) long enough, it will confess to anything. Data gives a ranked list of Carbohydrates.
How can you apply Product Analytics and Product Management best practices to make data-driven decisions about your food or diet? If you skipped over the previous article to avoid getting into the weeds, you’ve landed at the right spot.
Do you want to know optimal food choices? You can use this article as a cheat sheet for choosing carbohydrate sources when on a diet.
In the previous article, we understood the distribution of carbohydrates, Glycemic Index, and cost across a few carbohydrate-rich foods. Now, we will identify a success metric a.k.a. “goodness metric” and rank carbohydrates accordingly.
If you want to jump to the final result, scroll to the end of this article for a ranked list.
Thumbnail credits to Background photo created by freepik.
Possible Success Factors
We have now looked at the distribution of most measurements, we now want to define a metric of success for carbs based on the goals and measurements. A metric of goodness will be
Proportional to the percentage of calories from carbohydrates in a food item
More specifically, proportional to the percentage of carbohydrate calories from fiber plus starch.
Inversely proportional to GI
Inversely proportional to the expense
Inversely proportional to the percentage of calories from sugar
Let’s call this a goodness metric.
Goodness ∝ (% Cal from Carb)*(% Carb Cal from Fiber and Starch)(GI) * (Price per 100g) * (% Cal from Sugar)
It is not obvious how to use GL, since GL is proportional to #1 and #3, but we want the opposite different proportionality of #1 and #3.
If we rank carrot or watermelon (GI exceptions discussed before), they would rank low because they have high GI and low carb content. Is that an acceptable result? Yes, because they are not good sources of carb if they are low in carb.
Are GI and percentage of calories from sugar proportional? If yes, we can use only one of those. Looking below, we can see they are not correlated, given the low R-square.
How about “percentage of carbohydrate calories from fiber and starch” vs “percentage of calories from sugar”?
They are closely correlated. A correlation matrix is more numeric although less visual way to find this out. Below, I’ve attempted to visualize the correlation between the goodness factors we discussed earlier.
There are no scatter points above 0.50 for Glycemic Index, so it is rather independent of the other variables here.
The price per 100g food item only correlates with the price per 100g carbohydrate. Since the latter has a conceptual (not visible in the metrics) correlation with % Cal from Carb, let us remove the latter and keep only the former.
“% Calories from Sugar” is highly correlated with “% Carb Cal from Sugar”, “% Calories from Fiber+Starch”, and “% Carb Cal from Fiber+Starch”, so we should use only one of those. The foods we are considering have most of their calories coming from carbohydrates. So, we can look at the distribution either within carbs or outside. “% Calories from Sugar” has some zeroes, so we can instead go with “% Carb Cal from Fiber+Starch”
Let’s Call This A Goodness Metric
Let’s call this a goodness metric. However, similar to our analysis of proteins, price is not as important as the other two nutritional factors, so we will use Price to rank items that are relatively close in the other two measures.
Goodness Metric 1 = (% Carb Cal from Fiber+Starch)(GI)
Goodness Metric 2 = (Price per 100g)
The above visual shows us that for many food items, the Goodness Metric 1 is close enough that we can rank items within that band on price. I’ve visualized the bands using “ticks” on the X axis.
Other things to consider
You can define GI for a meal instead of using GI of a food item. A combination of food items can reduce the GI of the meal, hence meeting the goals. We did not consider this in the discussion. Independent on the fiber, protein, or water you add to a meal, you can lower the GI of the meal by option for lower GI carbohydrate food items.
Were These Metrics Necessary To Rank Carbohydrates?
I realized after months or years of keeping this article in my backlog that I was building compound metrics here to optimize the ranking. But the time-taking regression analysis and correlations did not help. The ranking for carbohydrates can be done using one measure. Just one measure. The Glycemic Index (GI).
The allure of compound metrics drew me into this long-drawn analysis.
Simplified Analysis Using GI And Price. Only.
Let’s stick to the simple metric scientists have identify for carbohydrates a long time back. Glycemic Index.
I’ve tabulated the same below.
Have you ever been tempted with metrics, data, or compound metrics to realize it wasn’t required? That you had a simpler alternative?
If you want me to add more sources to this list or use a different price source, let me know.