Why average as a metric is misleading and should not be considered for any kind of decision making ?
Did you know India’s per capita income for FY 2018-19 was Rs. 1,26,406 and for FY 2019-20, it is expected to be Rs. 1,35,048 ?
Per capita income can be defined as average income earned per person in a given area (city, region, country, etc.) in a specified year. It is calculated by dividing the area’s total income by its total population. It includes all people including children, elders or anyone who is counted in the census or population of the specified region.
That means “on an average”, every Indian ( including a child) earned approximately Rs. 1.35 Lakhs ( USD 1800) per annum last year. If that’s the case, “on an average”, the monthly income of an Indian would have been Rs. 11,254 ( USD 150) per month. Now, with that in mind, an individual would have earned Rs 375 per day “on an average”.
Since 2007, India has set its official threshold at Rs. 26 a day (USD 0.43) in rural areas and about Rs. 32 per day ( USD 0.53 ) in urban areas where people earning below these thresholds are defined as people living below poverty line (BPL) . As per RBI reports, Out of the total population living in the rural parts of 35 states and UTs of India, 25.7% of them is living below the poverty line. Isn’t it contradictory to say that on one hand, average income of an Indian Rs 375 per day where as on the other hand, 25.7% of Indians still live below poverty line ?
While the facts remain, the only thing that is making all of this look either good or bad is the law of averages. When you say “on an average" , you mean to say “typically”. What you may want to imply is that “typically” or “in general”, people are earning Rs 375/- per day based on the numbers available. But this does not paint the correct picture. In reality, “typically” does not mean average, it means “median”.
Outliers are single biggest influencers while calculating average of any collection of data points. Now, if you are an analyst graduated from a business school who has studied analytics as a subject, you can apply normal distribution curve, calculate distance from mean to get standard deviation. And then, arrive at many other conclusions. The only thing standard deviation can help with is to understand that if the distance from the mean is too big or not. Standard Deviation can help in understanding whether considering the average as a metric is useful or not.
Considering the above, According to the 2015 World Wealth Report, India had 198,000 high net worth individuals (annual income over USD 1 million) with a combined wealth of USD 785 billion. Assuming India’s population to be 1.3 billion people, 0.015% people generate more than half of total income of Indian population. This is what causes India’s per capita income to reach such a high number despite 25.7% population being below the poverty line.
In calculating averages, sample size matters a lot. Consider this as an example. In a room, there are 10 people with an average net worth of Rs 75 lakhs ( USD 100k ) . Now, enters Mukesh Ambani in the room whose net worth is Rs. 3.75 lakh crores ( USD 53 billion) . Suddenly, the average net worth in the room increased from USD 100k to USD 4.8 billion.
Unbelievable? Isn’t It a mathematical fallacy? You must be thinking it is crazy to put Mukesh Ambani in the same room to calculate average networth of middle class people. But that’s what we do everyday in our jobs. We calculate averages all the time. If you are salesperson, you always have to beat the average sales numbers without consideration of equality in the sales figures and other factors.
I am not sure of other industries as I have spent 19 years working in this so called IT Industry or Tech industry in the tech hub of India called Bangalore. Every year, when the time comes for appraisals and increments, HR heads of every company brings data to the table proving how much “on an average” the increments are being handed out this year. In that average calculation, HRs very smartly make sure to remove Flipkart, Amazon, Google, Facebook like companies but make sure to include the likes of Infosys, Wipro, TCS, CTS etc. and then, they provide the salary increment guidelines to the senior management to have 1-2% above the average. This number is then made public giving a false sense that the company is paying above average salaries or increments.
As I have been working in media technology company for last few years, I have noticed a very cool trend of publishing average viewership numbers. As per the latest report published by BARC ( the company that measures TRPs for Indian TV channels), due to COVID lockdown in India, the average TV viewership duration has increased to 4.48 hours per day from average viewership of 3.46 hours per day. This may ignore a convenient fact that most of the times, people leave the TV on and no one is watching. And secondly, the sample size of BARC is just 1,88,000 people.
So, if you notice the numbers in these examples and may be the examples you’ll see in your life, averages are calculated and used only when they convey a positive indicator and create a false sense of pride and growth. Most of the organisations have been using averages for the purpose of marketing and managing stakeholders. But for real business planning and strategies, average is the number that should be avoided all the time. If you building a business strategy based on law of averages, I am sure you are just trying to raise money from VCs or trying to prove that the work you have done is great.
What should you do if averages should not be used ?
There are so many mathematical ways to determine the success or failure of the business. And one thing is for sure, to determine business health, depending on simple math is dangerous. But for the sake of simplicity, one of the better indicator is Median. Median is the number that lies in the middle of the dataset. Outliers do not have any impact on Median. Median gives you just a bit but somewhat better picture of the data you have. As it ignores outliers, you can safely assume that if median number looks good for your business, then you are doing good “typically”.
Otherwise, the better way is always to create clusters of datasets. Histograms can help you create these packets of data or various smaller datasets. Now, by segmenting the data, you’ll be able to create buckets of the origins of data and then, perform separate business analysis of each bucket. Once you do that, you can drive a strategy either to improve each bucket performance, or to kill some buckets and focus on buckets that are doing better.
Consider an example from the media industry itself. People have started watching a lot of content online. Apart from YouTube as a popular content source, Netflix, Amazon prime video, Hotstar, Sony Liv, Zee5, Voot, MXPlayer, and plethora of other apps are getting to the market. Each of the apps are claiming that they are winning a huge market share as they are growing. Now, if you are working in one of those media companies, instead of looking at average session duration of a viewer or average viewership duration in a day, look at the median viewership duration. This will give you a realistic view of how much time “typically” people spend on your platform. Then, create the bucket of users based on different session durations. By doing that, you’ll be able to profile the interests of the users on your platform. May be your platform is more famous for short form content rather than movies or may be its popular only for Cricket matches or only for episodic content.
While there are so many statistical analysis techniques that are applied in every business, the basic technique of calculating an average is too common at every level in any professional environment. Because of the law of averages ( and subsequently either law of small numbers or law of large numbers), management teams arrive at quick conclusions that can affect livelihoods and earnings of people and their own business goals too.
In a nutshell, If the average number looks good, call your marketing department to make press releases and announcements on social media. Otherwise, ignore any kind of averages in life.
No one has ever liked to be just an average in life.