Calculate the number of queries answered by Google per second.

I would start by defining what a query is, based on that brainstorm through all the factors affecting the number of queries, and come up with a mathematical equation by assigning weights to factors and also making assumptions approved by the interviewer Once the equation is finalized I will substitute the factors with some assumed numbers to come up with a final answer.

Defining Query: In this context, a query is any search request or question that a user types into Google's search engine. Each query represents a specific search, whether it's a word, phrase, or question, that the search engine processes to return relevant results.

for this, we can start with population access to the internet divide them into cohorts based on internet usage per day, and downsize the number to seconds. 

Super Active users: The users with a very high number of queries
Active users: Users who depend a lot on the internet.
Moderate Users: Those who occasionally use the internet.
Irregular users: Who barely use the internet. ( ignoring these for ease)

The equation can be:
No of queries per second = ((Number of Super active users* Average queries by Super active users per day) + (Number of Active users* Average queries by active users per day) + (Number of moderate users* Average queries by moderate users per day))/ 24*36000

assuming 50% of the population has internet access: 
=0.5*800 Crores = 400 Crores.
For most of this population, the internet is synonymous with Google except for China. 
China's population = 100 Crore, assuming 60% have access to the internet (slightly greater than average) 

Population using google: 400 - (100)*0.6 = 340 Crores

Assuming 10% are super active users - with 10 queries a day
50% are active users - with 5 queries a day
30% moderate users with - 1 queriy a day.

No of super active users = (0.1)*340 = 34 crores
No of active users = (0.5)*340 = 170 Crores
No of moderate users = (0.3)*340= 102 Crores

Total Queries per day = 34*10 + 170 *5 + 102* 1 = 1292 Crores

Total Queries per second = 1292 crores/(24*3600) = approx 1.4 lakh



 

Interesting.

Clarifying questions:

1. Are we incorporating queries that Google couldn't respond to due to user errors resulting in errors? (Assuming Yes)

2. Does it encompass queries from around the world? (Assuming: Yes)

3. Does it encompass queries that Google has only answered once, or those it has needed to clarify or redirect elsewhere?

4. Does Google include not just search engines but also platforms owned by Google, such as YouTube? (Assuming just search engine)

5. I assume queries can be posed via a laptop, mobile device, tablet, or phone with an internet connection.

Flow of the event: User searched/typed a query in the form of a question/photo/video/equation/statement on Google -- clicked on the enter/search button -- query answered by Google

Equation: No of people have access of internet * fraction of people having google on the device* fraction of people using google per second  

No of people have access over internet: Total count of people around the world = 8 billion

Assuming that out of this 70% population lives in developing and developed nations and 30% population lives in underdeveloped small nations

Assuming that in the developed and developing nations 80% of people have at least one smart device with access to the internet and 60% of people in the underdeveloped nations have at least one smart device with access to internet

Number of people with internet access = 8 Billion*70%*80% + 8 Billion*30%*60%

                                      = 448000000 + 144000000

                                      = 5.92 Billion

Since Google is one of the widely used search engines, we are assuming 60% of people have Google. 40% include people using safari(since it is a premium and expensive brand, only fraction of people will have apple device), other search engines

People having Google on the device = 5.92 Billion*60%

                                   = 3.5 Billion people have google

People using search engines per second:

Assuming we will have 3 types of users active/casual/passive users

Let us say active users are using google 8 times a day and are 30% population

Casual users are using 3 times a day and constitute 40% population

Passive users are using 30% of the users and might not use google each day

 

Google searches per second per day = 3.5B*30%*8+ 3.5B%*40%*3

                                   = 8.4 Billion + 4.2 Billion

                                   = 12.6 Billion/86400 seconds

 

                                   = 145k queries per second

To start, let's consider the global population and the percentage of people who have access to the internet. According to recent statistics, around 59% of the global population has access to the internet. As of 2024, this would amount to approximately 4.4 billion people.

 

Next, let's consider the percentage of people who use Google as their primary search engine. According to StatCounter, Google holds around 92% of the global search engine market share. Therefore, we can estimate that around 4.03 billion people use Google as their primary search engine.

 

Now, let's consider the frequency of searches. According to a study by Moz, the average user conducts around 1.2 searches per day. However, this number can vary widely depending on the user's age, location, and other factors. For the sake of simplicity, let's assume that the average user conducts 1 search per day.

 

With these assumptions, we can calculate the number of queries answered by Google per second as follows:

 

4.03 billion users * 1 search per day = 4.03 billion searches per day

4.03 billion searches per day / 60 seconds per minute / 60 minutes per hour / 24 hours per day = approximately 46,656 searches per second

 

Therefore, we can estimate that Google answers around 46,656 queries per second. However, it's important to note that this is just an estimate and the actual number could be higher or lower depending on various factors such as changes in internet access, search engine market share, and user behavior.

Clarifications:

- Mode of searches is Manual via mobile, desktop, app, web, home device. Ignoring any robo/auto API based searches.

- Geographically, we need to look at global population?

 

Analysis: We'll do a top-down approach.

Global Population = 7.5B

Less, Geographies excluded: China, Half of Russia (people here don't use Google as their primary search engine); Less 1.5B (remain 6B)

Less, Only 50% people are connected to internet (remain 3B)

Less, Google market share with Bing, AOL, other regional engines: Let's assume it to be about 70-80% = 75% (remain 2.25B)

We are left with 2.2B people who are potentially using Google search as their primary search.

 

Usage Frequency

We can segment them in their frequency of usage. I would say there are 3 types of searchers: Aggressive, Active, Passive

Passive searchers are those who don't really use internet to search on a daily basis (like old people, kids, our mom/dads,  majority of rural population). Assuming this type represents the majority = 50% = 1.1B people

Active people would search once a day. Assumed to be 25% of population = 550M people

Aggressive searchers are super active searchers who might search 4-8 times a day. Assumed to be 25% of the total population = 550M people

Total searches per day= 550*1 + 550*6 = 3.8B searches per day

= 3.8B / 24 hours / 60 min / 60 sec = ~ 44,000 searches per second

A quick disclaimer: There are couple of really compelling answers but I am going to be very specific and share only short answer here.

Clarifying Question or Assumption:

• Any specific product under Google product stack? Let's assume Gmail in this case.
• Any specific country or region to focus? Let's assume Globally

• Average users per second = daily active users / seconds per day
• average users per second = daily active users / (hours per day * minutes per hour * seconds per minute)

• Based on my latest read, Gmail has 2 billion monthly active users, so I am thinking we can get this number by using this formula.

• Here are approx. 333 billion email messages sent every day in 2023 so far- Gmail holds 30% market share globally- 111 billion emails (Read, Search and write)-

  Total operations- 111 billion/86400—1.3 million queries per second

Clarifying Questions:

1. Do I focus only on google search and not include other searches like on gmail, maps etc. - concentrate only on main search

2. Platform web or app -> consider for app based searches

3. Any specific geography - consider India

4. Any timeframe of search I should consider - consider per second per year on average

 

Approach;

1. Population

2. Eligible base

3. Split basis frequency of usage

4. Arrive at a number

 

India’s population: 1.4billion

Eligible base: 

1. 600 million smartphone users

2. Internet penetration: 80% -> 480 million

3. Frequency of usage:

   1. High - 10% working professions -> 10 searches made per week

   2. Medium -> 30% in this bracket with 5 searches per week

   3. Low -> 60% in this bracket considering education, awareness, use case, other competing apps etc. -> 2 searches per week

 

Number of searches per year -> 52 (48*10 + 140*5 + 290*2) = 52 *(480+700+580) = 52*1760 ~8800 million searches per year

Global population = 7 billion

Internet penetration = ~ 60% of 7 billion = 4.2 B

Google search engine doesn't work in China(~1B population), therefore remaining= 4.2B - 1B = 3.2B

Dividing this 3.2B into different user segments:
1. Frequent users (~40% of 3.2B):

• Mostly young or working population(15yo-40yo)
• 5 queries minimum per day

2. Regular users (~40% of 3.2B):

• Mostly working population of 40yo to 60yo
• 1 query per day

3. Irregular users (~20% of 3.2B):

• Involves users of age segment >15yo or <60yo)
• Using Google search very rarely, hence eliminating this user segment)

Therefore, 80% of 3.2B use google search in the following manner:
Frequent users: (40% of 3.2B)*5 = 6.4B search per day
Regular users: (40% of 3.2B)*1 = 1.28B search per day 

Total = ~7.5B search per day = ~87K search per second

 

Assuming world population is 600 crore.

in this population  only 80 % has smart phones.

& google has a share of about 60%.

New population that can use google

=600cr*0.8*0.6

 =288 crore 

I have divided the people in three category.

 

Category	Search	percentage
Category1	30 per day	5%
Category2	6 per day	40%
Category3	1 per week	55%

 

Category1

288*5*30/100 =432 crore per day

Category2

288*40*6/100=691.2 crore per day

Category3

288*55*1/100=158.4 crore per week

Divide it by 7

=22.62 crore per day

Total=432+691.2+22.62

=1145.82 crore per day

=1146 crore per day(approx)

Now going for per second

1146 crore/(24*60/60)

=0.0132 crore

=1.3 lakh searches per second

feel free to give suggestions.

Calculate the number of queries answered by Google per second

First, when a guesstimate is posed, you should start by asking clarification questions. 

Clarification questions:

1. Are we calculating data for global scale or continental scale or country scale. 
2. What is the mode of search? Mobile or laptop or PC? 

I don't have any other clarification questions other than this, but please do enlighten me if you guys think there can be other clarification questions. 

Analysis :

Global population : 7.98B

My assumption : 7.5B

Access to internet : 63%

Online population : 4.72B

Assume google penetration is 70%(because most of China uses other search engines which is 1.5B population, this assumption might cover our mistakes in other assumptions) 

No. Of Google search engine users : 3.3B

I would like to segment my users using age groups for a better approach. 

0-10 Yrs : Kids, most probably can't use google search so I assume 0 questions from this group. They are around 10%.

11-30 Yrs : Most active group, teenagers and employees. Assuming 4 questions from them, since i myself pose more than 4 questions. Around 30% of the 3.3B. 

0.3*3.3*4*10^9/86400 = 46200 QPS

30-50 Yrs : Less active group. Assuming 2 questions. 30% of 3.3B.

0.3*3.3*2*10^9/86400 = 23100 QPS

50-90 Yrs : Almost passive group. Assuming 1 question. 30% of 3.3B.

0.3*3.3*1*10^9/86400 = 11550 QPS

Total QPS = 80,850 QPS. 

Sanity check : 6.98B questions per day with 90% of 3.3B, makes it around 2 questions per person which I think is ok. 

Please give feedback on my approach, I thought of active-passive approach but age groups approach seemed better. 

 

 

 

 

 

Is it across the world? Yes

There could be other questions that may arise, I would like to ask them as I go by if required. 

I would like to take a top-down approach. 

What's the world's population? - 7 billion

Assume that 60% of them have an internet connection. This gets us to a 4.2billion.  

I think Google is banned in China -> take out 1.5 billion -> that leaves us with 2.7 billion people. 

Google has 90% search domination. That means about 2.43 billion people use it daily.

 2.4 billion people: they're students(30%), computer-related working employees(15%), non-computer-related people(25%), non-working class or retired (30)

there are 

1. students (30% = 0.72 billion)
   1. 5hrs per day on phone/laptop: social media, videos, entertainment apps = 4hrs
   2. 1hr per day => google time => approx 5 queries
   3. 0.72 * 5 = 3.6billion queries per day
2. computer-related working employees (15% = 0.36 billion) 
   1. 10hrs per day on phone/laptop
   2. 2hrs => google time => 5 queries
   3. 0.36*5 = 1.8billion per day
3. non-computer-related people (25% = 0.6 billion) 
   1. 7hrs per day on phone/laptop
      1. 6hrs on videos, entertainment, social media, shopping e-commerce, shopping deals
   2. 1hr => google time => 2 queries
   3. 0.6* 2= 1.2billion per day
4. non-working class or retired (30% = 0.72 billion)
   1. 3hrs per day on phone/laptop, rest of the time on tv
   2. google time => 1 query per day 
   3. 0.72* 1 = 0.72 billion queries per day

 

📌 Clarifying Questions

 

1. When we say google are we focusing only on the google search engine or also its other products- Take its search engine only
2. Since the trend of queries will differ from hour to hour so can we go ahead assuming it is a uniform way:- Yes
3. Where is the search or queries taking place like on mobile phone, browser, some of the APIs that run and ask queries:- Let's take phone and browser
4. In which region are we finding this number:- All over the world

📌 Creating the Equation

Number of queries answered by google per second =

Number of queries by a person in a day / (246060)

 📌 Solving

 

1. The population of the world is 7 billion
2. Only 60% of the world has access to the internet making the number = 4.2 billion.
3. Google's search engine market share all over the world is 70% = 2.94 billion.

⭐ We come to the number of people who are using google = 2.94 billion ~ 3 billion.

 

Now this number consists of all three types of users

1. Aggressive users (these are working professionals):- 5 times on a daily basis
2. Active users:- one time at least in a day
3. Passive users (Like our mom/dad, people in rural areas):- thrice in a week ~ 0.5 in a day

Aggressive users are those who very frequently search on the internet like 4-6 times a day. they are generally working professionals from all types of countries lie developed, developing, and underdeveloped.

Since these people have access to the internet the location of these people will be higher in developed and developing countries compared to underdeveloped countries.

The division could be as follows:-

1. Person in Developed countries - 20% → 0.6 billion
2. Person in Developing countries - 50% → 1.5 billion
3. Person in Underdeveloped countries - 30% → 0.9 billion

For Aggressive users, the division would be greater in developed and developing countries because of the opportunity and education

For Active users, the division will be slightly greater in developing countries than in both the other types because of the growing and reading phase.

For Passive users, the division will be highest in underdeveloped countries because of the less education and opportunities.

 AggressiveActivePassive

Developed Countries	60% * 0.6 billion = 0.36 billion	30% * 0.6 billion = 0.18 billion	10% * 1.5 billion = 0.15 billion
Developing Countries	30% * 1.5 billion = 0.45 billion	50% * 1.5 billion = 0.75 billion	20% * 1.5 billion = 0.30 billion
Underdeveloped Countries	10% * 0.9 billion = 0.09 billion	20% * 0.9 billion = 0.18 billion	70% * 0.9 billion = 0.63 billion
Total	0.9 billion	1.11 billion	1.08 billion
Total Number of queries	0.9 * 5 = 4.5 Billion	1.11 * 1 = 1.11 billion	1.08 billion * 0.5 = 0.54 billion

The total number of queries in a day = 4.5 + 1.11 + 0.54 = 6.15 ~ 6.2 billion

⭐ The total number of queries in a second = 6.2 billion / (24 hours * 60 mins * 60 secs) = 71,759 ~ 71,800 queries in a second

 

 📌 Sanity Check

 

Since the number of google search engine users is 3 billion and the daily number of queries is estimated to be 6.2 billion.

So it comes done to approximately 2 queries per user per day and it seems to be in limit.

 

Clarifications:

• Searches answered for any specific location / region
• Searched specific to any specific platform (Web, Mobile - further Android or iOS, Desktop, Laptop, iPad, etc.)
• Any other specific customer segment, use type like B2C or B2B
• Is there any specific goal for the estimation

• Searches emanating from India only
• Would include only mobile platforms - include Android and iOS
• Would include all type of user segments
• Goal for estimation is to increase the usage

• We can assume that B2B usage for e.g. Google Search acting as a search engine for a website or portal or through news app or embedded in Google Maps
• We can estimate based on population of India which has access to internet - through mobile data this is high in India at roughly 450 million
• Further can divide the population into : 
  • power users - will access through their mobile, latop or desktop in office, use google search 25 - 50 times in a day, they are 5% of users so that will be 50x22.5M / (24x60x60)  ~ 5M / 360 ~ 14000 searches / sec
  • normal users - next 15% users who use explicitly or implicitly (i.e. through an explicit or implicit search - search through apps like News App, restaurants, petrol pumps around u) about 5 times a day so that will be 15% of 450M x 5 / (24x60x60) ~ 4000 searches / sec
  • infrequent users - next 80% users who use infrequently so lets say 1 time per day - 4200 searched / sec
• So as per our estimate a simple model for B2C search gives roughly 22200 searches answered / sec

 

1.       Assuming we are talking about Google search engine queries – because google has other search services like Youtube as well.

2.       Total global population is 7B

3.       China does not use google as its Primary search engine so let’s reduce  1B

4.       So we are taking the sample now to be 6B

5.       Assuming only 50% of world’s population is using internet

6.       This reduces the sample size to be 3B

7.8.       These 3B user are spread across developed / developing and underdeveloped part of the world

9.       Based on the this we can split that 3B sample size into a segment of 1B each

10.   Let’s assume people in the developed countries are querying google 5 time a day

11.   People in the developing countries on an average are querying google thrice a day

12.   People in under developed countries are querying google .5 times a day

13.   Total query count per day is 5*1+3*1+.5*1 = 8.5B

 

 

1. Clarify the scope

   1. Is it only a google search engine? Or does it also include google ads, youtube, gmail, etc?

   2. For the purpose of the estimation, I’m going to assume google search engine only

2. Main equation:  # of users who uses google search engine x average daily search use /  seconds per day. 

3. Breakdown unknown into simpler equations

   1. # of users who uses google search engine

      1. World population x % of population that have access to google x ratio of google use compared to other search engines

         1. ~8B population 

         2. Assume 30% of world does not have access to Google (countries where Google access is blocked or do not have internet access)

         3. Assume among the 70% of world population that has access to Google, Google is the primary search engine for 50% of the people.

   2. Average daily search use: 

      1. Assume 50% of population are heavy-use users who uses google on average 5 times a day. This population is likely students who relies on google search for research, and people who have jobs that require using Google to look things up

      2. Assume 20% of population are medium-use users who uses google on average 3 times a day. 

      3. Assume 20% of population are light-use users who uses google on average once a day

      4. Assume 10% of population do not use search. 

   3. Seconds in a day= 24 x 60 x 60

4. Calculate the #

   1. # of users who uses google search engine

      1. 8 Billion X 70% X 50% = .35X8 = ~3 Billion

   2. Average daily use:

      1. (.50)(5)+(.20)(3)+(.20)(1)+0= 2.5+.6+.2 times per day. ~ 3 daily use

   3. Seconds in a day:

      1. 86,400

   4. (3B x 3)/86,400 = 105K. Roughly 100k queries per second

5. Do a sanity check

   1. There is ~8B people world wide, ~.001% of the population is using google search at any given time.

1. Clarify

Are these search queries? Like if I go to google.com or google search bar in the browser and fire up a search

Does it include all search platforms like mobile, web etc - I'll assume both

Is this for global users or US? I can start with US 

Is this for consumer or corporate? Google search engine is also leveraged by corporates? I'll assume consumer.

So I will calculate the number of search queries that are requested from users in US answered by google per second 

2. Equation

Number of queries answered = number of queries sent (assuming google answer every search query it receives - no caching)

Number of queries sent per second = (# of users * # of queries per user per day)/(24*60*60)

3. Calculate

US Population ~320M, assuming 80 years of life expectancy and uniform distribution

	Population spread	Internet access	Google search mkt share	# of users	Browsing freq / day	Search/browse	# of search queries/day
0-12	48000000	dont search		0	0	0	0
12-18	24000000	70%	90%	15120000	4	10	604,800,000
18-65	188000000	70%	90%	118440000	10	10	11,844,000,000
65-80	60000000	50%	90%	27000000	2	10	540,000,000
							12,988,800,000

 

Clarifications:

1. Queries searched via Google desktop or mobile? Assuming overall here.
2. Google has many products like Google Search engine, YT, etc. Here, I'm assuming Google search engine only.
3. Assuming no geographical area here - so overall world population will be part of the calculation.

 

Calculation:

I will do the calculation following the top-down approach as it's the best and easy to calculate a number.

7.5B is the world's total population.

China has a separate search engine (Baidu) - Hence, removing 1.4B population from total = 6.1B

Market share of Google search engine with other search engines (Bing, wiki, etc.). It's 90% according to the research. Hence, population count remains = 5.5B

People having internet connection are 60%. Hence, count remains = 3.3B

Now bifurcate 3.3B users based on the usage:

1. High usage (people who used to search on an average daily 5-6 times) = 25% (IT professionals, other professionals) = 825M
2. Moderate usage (people who used to search at least one time a day) = 50% (group of rural & urban areas, other professionals group, group of students) = 1.65B
3. Low usage (people who used to search sometimes i.e. one time in 4-5 days -> we can assume 0.25 a day) = 25% (Elderly people, students group) = 825M

 

Let's calculate Google Search queries per day according to the user groups:

1. High usage = 825M*5 = 4.1B queries
2. Moderate usage = 1.65B*1 = 1.65B queries
3. Low usage = 825M*0.25 = 206M queries

Total queries per day ≈ 6B

Queries per second = 6B/(24*60*60) ≈ 69,444 queries per second

Clarification - 

Is there a specific time you want me to take or just consider an average number of queries per second? Consider average. 

Are we talking about Google search queries? Yes

And queries can be asked by a human or a bot. What are we interested in here? Human queries

Okay, and are we interested in a specific geography? No, consider global. 

And lastly, do you want me to factor in the average time it takes for Google to answer a query or can we assume it's instantaneous? Assume it's instantaneous.

Refining problem statement - 

Thank you. So, then our problem statement is to calculate the average number of human search queries made in Google per second, globally. 

Approach - 

Global population - 8 billion -> 800 crores

Internet penetration ~ 70% -> 560 crores who have access to internet

Whole of China doesn't use Google -> This gives 560 - 160 = 400 crores

Market share of Google search would be 60%. So that gives us 240 crores. 

Out of these 240 crore people, I'll classify based on usage - 

Low - 0.5 query a day -> 20% (projecting a rough estimate based on the composition and usage I've observed in my family)

Medium - 1 query a day -> 40%

High - 5 queries a day -> 40%

So, the number of queries asked in 24 hours = 48 * 0.5 + 96 * 1 + 96 * 5 = 24+96+480 = 600 crore queries. 

Assuming uniform distribution, number of queries per second = 600 crore / (24*3600) ~ 70000 

Estimated -> 70000 queries per second. 

Clarifying questions and assumptions

• When we say queries, I assume this is any search performed by a user using the Google search engine?

• Google provides search capabilities on third party sites. I am not considering that for this estimation.

Breakdown of problem

• Population of the world

• Population of the world which has access to Google

• Population which has access to the Internet

• Search engine market share split

• Searches done per user in a day

• Total searches done

• Searches done per second

• Population of the world
  • 7.5 billion
• Population which has access to Google
  • China is the largest country which does not have access to Google. Reducing the population of China which is approximately 1.5 billion, we have 6 billion people who have access to Google.
• Population which has access to the internet
  • In each country, there is a percentage of the population which does not have access to the internet. As internet penetration is going up, I am keeping the internet penetration as 85% of the population. This would reduce the total population considered to 85% of 6 billion which will be 5.1 billion. I would like to round it down to 5 billion.
• Searches per day
  • As online search is a very common action, I would take the average number of searches per day per internet user as 3.
• Total searches per day
  • Total searches per day would come to 5*3 = 15 billion
• Searches on Google
  • As Google is the largest search engine in terms of volume of searches, I would put the percentage of all searches that happen on Google at 90% of all searches which comes to 13.5 billion.
  • Evenly distributing the searches throughout the day, searches done per second will come to 13.5 billion/3600 which comes to 3.75 million searches per second.

Clarifications: 

- Medium - phone, app, web, home device - No API calls and no robo search queries.

Steps:

Step 1: Start with the global population: 7.5B

Step 2: Eliminate China and Russia as google is banned there, thus eliminating about 1.5B of the world's population. Leaving you with 6B.

Step 3: Reduce 50% from the remaining population because internet penetration is about 50% when taken across the world.

This leaves you with 3B.

Step 4: Approximate googles market share in the search engine space to about 70-75%. This leaves you with 2.2B customers that use google search.

Step 5: Divide this 2.2B population into user groups based on their daily search pattern.

Group 1: Passive - Those who do not search on the web, rural population, elderly, kids etc - approximate it to 50%.

Group 2: Active users: Those who do atleat one search per day - Remember, even if a person does one search they probably do it 1-2 times. Approximate to 25%

Group 3: Aggressive users: Those who do multiple searches per day - 3-8 searches per day. Approximate to 25% of the population.

Step 6: Calculate the no. of searches per user group per day.

Group 1: # of searches per day:  0 - as they do not search the web.

Group 2: # of searches per day: 0.25*2.2B*(2searches per day)= 1.1B

Group 3: $n of searches per day: 0.25*2.2B*(8 searcher per day= 4.4B

Step 7: Total no. of searcher per day = 0+ 1.1B + 4.4B = 5.5B

 

Step 8: Searcher per second = (Total searcher per day)/ (No. of seconds in a day)

=> 5.5B/(25*60*60)= 63.6K  searcher per second

 

Approach:

1. Estimate the number of google search users

2. Estimate the number of queries per user per second

3. Forumula= google search users * number of queries per user per second

Estimate google search users:

World population, 7.5B

Google is banned in china, so world population without chine is 6 B

Assuming the internet penetration of 70%, total internet users are 4.2 B

Assuming google search market share is 90%, total google search users are 3.8 B

Estimate number of queries per user per second:

Average number of hours of internet use per user in a day is 2 hours

Number of queries per user per hour is 1 queries 

Number of queries per user per day is 2 queries

Number of queries per user per second is 2/(3600*24) queries

Estimate the number of queries per second = google search users * number of queries per user per second

= 3.8 B * 2/ (3600*24)

= 7.6 MM/ 864

~ 87K queries per second

 

 

 

Clarifying questions: 

1) Do we need to estimate on specific device like Desktop/laptop or Mobile ? - Assume Mobile

2) Is it related to particular product of Google? - like Google search, Youtube, Photos -  Assume all 

3) Any particular time like day or night? - Estimate an average query per 24hrs (entire day)

4) Any specific Geography - Limit the study to US.

to Summarize -  We need to provide avg. query on any Google product at any time of day on Mobile in US.

In order to provide an estimation, i would like to make few assumptions:

1) US population is roughly 300Million, I am expecting avg. life in US around 75 years.

2) I am expecting Mobile usage between age-group 16-65 year age group.

Assuming linear age-split, i get 200Million population within 16-65 age group

Since US is a mature/established country, i am assuming 90% penetration rate in 16-22 age group and 100% mobile penetration in 23-65 age group, hence total number of handheld devices = 197.2M

I am aslo assuming that all of those devices are smart-phones and have access to internet through LTE or Wi-Fi (College/Business/home/hotspots).

I furthur devided my age-group into avg. usage activity per day. for simplicity purpose, i am assuming 3minute per new search

1) 16-22 Age - 

These users are in students/College. They spent majority of time on Desktop/laptop working on research/study/online classes, but do spend on an average 3hr/day on mobile phone on activities like games, Social media, Search (College materials, study, online classes, research) and conversation with friends over phone. 

 

2) 23-30 Age - 

These users are new professional who have recently joined a company and are the heaviest avg users on Mobile phone by spending on an average 5hr/day on mobile phone on activities like games, Social media, Search (Linkedin, Jobs, dating, restaurants, places to go, maps, Youtube videos, vacation) and conversation with friends over phone. 

 

3) 31-50 Age - 

These users are Married, newly or becoming parents. They spent majority of time on Desktop/laptop in Professional space, but do spend on an average 4hr/day on mobile phone on activities like Social media, Search (kids shopping, kids school, study, restaurnts, vacation plans, youtube, Maps) and conversation with friends/family over phone. 

4) 51-65 Age - 

These users are in higher postions in their professional career and parents to high school/University Students. They spent majority of time on taking meeting from Desktop/laptop or even phone but do spend on an average 3hr/day on mobile phone on activities like Social media, Search (Kids fees, vacation, maps, restaurants, youtube) and taking office meetings or family calls. 

 

Using above segmentaion, i have created a below modeling which provides roughly 3.4B query generated every day on any of the google product or ~39,491 (3.4Billion/ *24*60*60) query per second.

Age-group	Active Devices (In Million)	Activities					Number of Avg Searches/sec
		Games	Social Media	Search	Talk over Phone	Number of Queries
16-22	25.2	0.5hr	0.5hr	0.5hr	1.5hr	10	252
23-30	32	1hr	2hr	1.5hr	1.5hr	30	960
31-50	80	0hr	1hr	1hr	2hr	20	1,600
51-65	60	0hr	0.5hr	0.5hr	2hr	10	600
	197						3412

Clarifications:

Geography: around the globe. Scope: "Anything searched on google.com, even if the query resulted in no answers found"

 

Approach:

We will use a top down approach by first finding out the number of people using Google and then further dividing it into their purpose of use and finally multiplying it with the average number of queries for that use case.

Taking into consideration that China doesn't have Google, we are left with 5.5B population, taking into consideration the access to internet around 50% and that google's share is 90% we have total population as ~2.5B.

People use google for following reasons

• Working professionals: lets say 60% of the world's population is working and out of these only a 10% need to search on google. Taking the average search for working professionals as 10 per day.
• Students: a. School: around 15% of the population goes to school and around a 20% would be using google for their homework and other activities. Taking 5 searches per day for students. b. Universities: These 5% population is a heavy user (15 searches per day) and almost 80% have the access to google.
• Checking Internet: most of the population search google to check the internet. around 60%  would search google to check their internet connections. Average searches: 1 per day.
• Other casual searching: 40% uses google everyday with 2 searches per day on an average

 

Clarifications and assumptions: 

Google comprises of search engine, maps, gmail, GCP, GSuite apps etc. Assuming this question is for the core search engine. 

Assuming question is for working hours from 8am - 5pm. 

Define scope of search engine: Queries coming from browsers on desktop and mobile, and Google search app on mobile.  

Out of scope: Not counting queries between google services. For e.g, within google maps (whether web app or mobile app), if there is a search, that is not counted. 

Limit scope to US searches. We can then calculate RoW. 

 

Number of queries per second = Number of daily users x Number of queries/user/second. 

 

Number of daily users: Every smartphone user will search atleast once daily. Assuming that 90% of people > 16 years have smartphone. 

US population: 330M

People > 16 years - 65% of population. 193M users with smartphone. Round it to 200M. 

 

A good population of these users will also use desktop. Set of user who will use desktop but not smartphone is very low. 4-8 th grade students is a segment who are likely to use desktop but may not have smartphone. Lets aprk this segment for later analysis. 

 

Basically, we have 200M daily users for Google search. 

 

Now we need to find number of times an avg user would search. There are definitely differnt segments here: 

1/ Knowledge workers - may search a lot. Since its tied to their job. Doing research

2/ Blue collar workers - maybe less. They may use various apps, but not use search a lot. 

3/ Students - 8th grade to graduates: use a lot .

4/ Unemployed: May search less. 

 

Overall lets assume they search maybe twice an hr during normal working hrs. So thats 8 hrs.  16 searches per day. 

 

Total searches from day = 200M x 16 = 3.2B searches/day. 111K searches per second, since we are considering only 8hrs window. Assuming that google seach query response is very high: 99.99%. So all these searches are answered. 

 

Sanity check? 111K searches per second seems low, but the assumptions on number of daily users and searches per hr are reasonable. So its ok. 

 

This is for US. Now for world. Lets split world into developed countries, India, China and developing countries. 

Developed countries - we can just calculate assuming protionate to their population against US. US 300M - 111K queries/sec. Total developed work population - EU, Canada, Australia, etc → assume 3x US population. Assume same usage split. So its 333K /s from developed world. 

1.2B developed world → 444K/s

1.4B India. Asusme its half the usage. (1.4B/300M)*0.5*111K/s  

1.5B China -- Google doesnt play here. 

 

Remaining - 3B populaiton in developing world. Assume same as India

 

Clarifying Questions :

1. Do we need to consider the global population or in a specific reason like north America . - Global
2. When we talk about number of queries i assume we are talking about search queries. is that a correct assumption ? - Yes
3. Are we talking about queries across all devices like mobile, web etc ? - Yes
4. Can we assume that we are talking only about google web search and not include you tube and other google products. - Yes
5.  

Calculation

Total Search Queries per day

Number of internet users using google search* number of search queries per day / seconds in a day

Analysis :

World wide internet users : If i remember correctly there are around 5 billion internet users globally.

Google is the most widely used search engine across the globe and i believe they enjoy a market share north of 70%. I believe there are some restrictions on using google in china .

I am a heavy internet user being in the tech industry and i go to google at least 10 times a day to search for something. However i would say I am among the top 10% google user. So if i do an even distribution like top 10% - 10 queries a day, next 10% 9 queries a day i can calculate the average search queries per person per day.

Lets get to the math.

Total internet users : 5B

China and Russian has strict restrictions on google -

Total population of China and Russia - 1.2 b+ .3 b = 1.5 billion

Internet users in this region → if 5B people across the globe uses internet, the percentage of global population using internet is 5/7.6 → 65%

In China this percentage will be higher as they have more internet penetration compared to third world country. So lets say 80% of the population has access to internet.

So total internet users in russia and china → 1.2 billion

Lets say only 20% population from these areas use google - > 1.2*.2 = .24 billion

Total internet users in

Total internet users outside china and russia = 3.3 billion

75% market share in search - 3.3*.75 = 2.5 b

So total google users : 2.75 b

Average search queries per day per user - > 55/10 → 5.5

So total number of search queries per day → 5.5*2.75 = 15 billion

So total per hr = 15/24 = .6B

per minute = 600/60 = 10 million

Per second = 10/60 = .16million = 160k

 

Clarify - Is this for Google searches only or other Google products (which might have separate calculations to add to this [YT, GMAIL primarily])

Assumption - This search stat has to be found across all platforms and not just desktop/mobile.

Number of Google searches per second 

= # internet active users x avg number of queries per second x Mkt share of Google

= (# internet active users x avg number of queries per day)/86400 x Mkt share of Google

= (# number of people globally * % of internet penetration) x (weighted average of average number of queries across different demographics segment in a day)/86400 x Mkt share of Google

---------------------------------------------------------

# number of people globally = 7.5 B

% internet penetration = 60%

# total online users = 4.5 B

Type of internet users:

Global life expectancy = 75

Age group active online = 10-60 (70%)

Avg number of queries by 10-20 in a day = 5

Avg number of queries by 20-30 in a day = 8

Avg number of queries by 30-40 in a day = 3

Avg number of queries by 40-50 in a day = 2

Avg number of queries by 50-60 in a day = 1

Weighted avg number of queries = (5+8+3+2+1)/6 = 3

Mkt share of Google searches = 85-90% of global search across platforms

Num of Google searches = 4.5 *10^9 * 70% * 3 / 86400 * 85% = 92969 approx rounding to 93k QPS.

Sanity check:

On a yearly level Google does over trillions of queries (1.2 T in 2012). So extrapolating that to 2021, it could be 2.4 by year end.

So, if there are 2.4 T searches by 3.15 B in 365 days = 76k QPS approx.

Actual - As per latest stats, over 91k QPS on Google

Clarifying questions -

1) We want to calculate the data for entire world.

2) We are talking about all platforms - Mobile, Desktop, App, etc

Methodology -

No. of queries per second = No. of users * No. of queries per user per second

Data cleaning-

1) Competition like Bing (~30% market share)

2) Countries like China have their own search engine (~1.5 billion)

3) Only 50% population have internet access 

4) Older population, small kids have to be eliminated, literacy

User base - (8-1.5)*50%*70%*50% - 1.1 billion

Category of usage	No. of queries/day	%	Total queries per day
Low usage	2	50%	1.1*50%*2 = 1.1
Medium usage	5	30%	1.1*30%*5 = 1.65
Heavy usage	10	20%	1.1*20%*10 = 2.2

Total no. of queries per day = 4.95 billion

No. of queries per second = 4.95*10^9/(24*60*60) = 57,291

This is basis assumption that the loading time of Google page is less than 1 second.

Clarification 

• When you say number of queries answered by Google - do you mean search queries? (Google can potentially get queries from Gmail as well for inbox searches), assuming its only search 
• Assuming its a BAU day 
• Globally or for India - I'll take a stab at global 
• Number of queries per second = number of queries daily/86400(seconds in a day)
• Number of queries daily=  number of users x avg daily query per user 
• Number of internet users in the world = 50% internet penetration = 50% of 7.5 billion = 3.8 billion
• Minus Internet population of china = 1 Billion(google isnt present in China)
• We are left with 2.8 billion this will be MAU
• Assume day/MAU ratio of 50% = 1.4 Billion users
• I expect 70% of users to search at least once a day
• Google search has 90% penetration into search engine 
• We can split the user in high,medium, low and zero usage,
• Zero- I expect a third of users to not search(basic internet access like email check or browsing)
• Low - these users search on average 3 times a day, majority of users fall in this bucket at 50%
• Medium - more active, spend more time on internet 6 times , 15% of population
• High - power users - uses it very often 12 times , 5% of population

I: I am going to focus on queries in US. By queries can we stick to just those queriest that are resultant from someone actively typing something in the Google search bar vs. also including queries that occur behind the scene by let's say accesing a Google API on some other app or webpage, and also not including queries that go out to Google by clicking on some link on a webpage.

Interviewer: Sure.

I: Great! I am going to attempt a top down approach whereby let's start with US population and determining how many people use Google and how many approx queries a week. We can then total all those components up to get an estimate.

I: Also, let's assume mobile / desktop / tablets, etc. so basically all devices.

Interviewer: Sure, let's see how that works out.

I: US has 360 million people. It's safe to assume that 95% of the population use Google = 342M. Out of these, 20% kids (20% *324M = 64.8M), 50% youngsters / young adults (40% * 324M = 162M), and 30% older adults (30% * 324M = 102.6M).

Kids generally play games and watch TV / movies for the most part (since they have ample time on their hands and if parents are educated they will ensure kids spend most of their time reading) and not on devices/internet doing random things. So, let's assume per day on an average, kids have about 10 queries * 5 = 50 queries + 5 * 2 = 10 so total is 60 queries in 7 days. I have half the # of queries on weekends because kids are generally not studying or reseraching for school work during weekends.

Youngsters / Young adults go to school, have reserach, have work, etc. so they will have more queries during week and prob just ted bit over weekend. So during week per day 20 * 5 = 100 + 15 * 2 = 30 totaling 130 queries.

Older people generally read financial stuff (mostly in my experience if they are on internet they are logged into their investment accounts and readign through that vs. Googling stuff). During the week, let's assume 7 queries and weekend is the same. So 5 * 7 = 35 queries.

So for Kids: 64.8M * 60 queries = 3.8 billion

Youngsters + Young Adults = 162M * 130 queries = 21 billion

Older people = 102.6M * 35 queries = 3.6 billion

The numbers seem about right to me in the sense that kids and older people have about the same time / need to google stuff while youngsters actually have a higher need to google stuff so the spike in number doesn't shock me so far.

This means we have 3.8billion + 21 billion + 3.6 billion = 28.4 billion queries / week across US. That's 8.76 times the queries than # of people who use Google. That seems quite about fine to me because that means each of the 324M people query about 8.76 queries a week. In fact, that seems maybe a ted bit lower. One thing, that I could have done is consider certain % of people who don't have internet access but that would be negligible number considering we did account for mobile users as well. So I don't think that would have changed much. I will go with 28.5 billion queries / week.

What queries? Web Search queries

For which country  queries in scope? US google web searches

Here is my top down calculation approach

• Start with US population number

• Factor it  by Google search market share in US

• Calculate total searches by them daily

• Calculate the lower and upper boundaries of queries per second (qps)

  • Assume even distribution of searches during the day

• Calculate peak hour queries per second

How many queries per second?

• US Population is about 300 million and Google search share in US is about 65%

• So 2 out of 3 people in US search on Google which is 200 million US users search on Google

• I assume about 25% of 200 million are children, people without internet access, disabled people, very old people
• So about 150 million people search on Google in US

• I will assume all of 150 million people search at least once in a day which translates to

  • Total Daily Searches = 150,000,000

  • 150,000,000 / 24 / 60*60 = 1736 queries per second

  • This gives us the lower limit for searches per second

• Many users are likely to do multiple searches during the day

• I am going to assume 50% of users search at least 3 times a day which translates to

  • Total Daily Searches = (75,000,000 * 1 + 75,000,000 * 3) = 300,000,000
  • (75,000,000 * 1 + 75,000,000 * 3) / 24 / 3600 = 3464 queries per second
  • The upper limit for searches per second which happens to be 2 times the lower limit

• The queries per second range for US is 1736 to 3464

• We know that the searches peak during mornings and evenings with evenings being highest

• I will assume a 3rd of the daily searches happen during the evening 4 hrs from 7pm to 11pm EST when both East and West coasters are hitting google.com to search

• That translates to 1/3 * (75,000,000 * 1 + 75,000,000 * 3) searches during the peak 4 hours
• The peak hours queries per second becomes 100,000,000 / 4 / 3600 = 6945

• So the queries per second range for US is from 1736 to 6944 depending on which time of the day you look at it

Clarification: The scope is manual search via mobile and desktop browser, and is limited to Google search, not other applications like Map, etc.

Here are the steps I will take:

1. Calculate global internet penetration, i.e. the number of people who have access to the internet.
2. Calculate the number of people who have access to the google search.
3. Figure out the search market segments based on usage frequency per day.
4. Add the total of search queries in the world per day.
5. What percentage of the above total is run by google based on the Google search engine market share.
6. Convert search query per day to search per second.

• 3 billion x %30 (inactive users) x 0 (no of searches per day)=0
• 3 billion x %30 (infrequent users) x 1 (no searches per day)=0.9
• 3 billion x %30 (casual users) x 5 (no of searches per day)=4.5
• 3 billion x %10 (powerusers) x 10 (no of searches per day)=3

Clarifying questions:

There could be many queries to google like search, GMAIL, Photos, let's assume we would like to estimate the number of search queries answered by Google per second.

Here's how our high level equation would like: 

1. Number of search queries answered per sec = Number of search queries succesfully submitted by active users per sec

Let's try to break down  the high level equation further. (Number of Search queries per sec) + (% successfuly submitted per second) + (Number of Active users) 

2. Number of search queries by active users per sec = number of active users per sec on google search

Users can search either using Mobile, Laptop. Let's try to focus on Number of active Mobile users here.

Mobile users can be located in any Geography, here let's try to focus on number of active Mobile users in the US.

US population = 300 Million

Let'sassume 200 Million people have access to mobile phone with a data plan

3. Number of active users per sec  = Number of Active Mobile users in US with a data plan =200 Million

4. Number of Active Mobile US users on Google Search ?? 

What brings users to Google Search. Users get redirected to Google Search through Ads or direct search. Let's focus on Ads redirection here. I'm assuming 60% users are re-directed through Google Ads since Google Ads have more reach and competitive in terms of pricing as compared to other platform

5. Number of Active Mobile US users on Google search = 120 Million

What % search queries are succesful? Succesful query depends on network bandwidth, cellular carrier, cell phone cache. Given Mobile technologies today auto clear cache frequently by forcing restarts and prompting users to close applications. So Network badwidth is the biggest barrier for a succesful search query submissiong, which can vary based on location or activity or weather conditions. So let's assume 60% queries are succesful

72 Million succesful queries on mobile per second is being answered by Google in the US

Sanity Check: It would be absurd if I would have come up with a number 300Million succesful search queries answered in the US per second  on mobile as not every person in US has a cell phone given the homeless population. And not every user is on Google as there are other competitios too. So 72 Million seems fair

 

We have to first arrive at the total number of Google users of any frequency.

Population of the world is around 8 Bn.

Google has around 500 Mn users in India.

China, which is the most populous nation in the world does not use Google.

USA adds around 300 Mn Google users.

So, adding Google users from Europe, Africa, Americas and Australia we can estimate total number of Google users to be around 2.4 Bn

 

Now, Google searches per second= Google searches per hour/3600

Google serches per hour= Average of Google searches per hours across all user groups.

 

There can be broadly 4 groups of Google users:

1. Heavy/Power Users: They do 1 search every 3 hours. 

2. Medium Users: They do 1 search every 6 hours

3. Light Users: They do 1 search every day

4. Intermittent users: They do 1 search every 3 days

 

It makes less meaning to go beyond less than 1 search in 3 days as the users would be very less and insignificant.

 

Coming to the calculation part

 

 

	Power	Medium	Light	Intermittent	Total
Users	500 Mn	1000 Mn	500 Mn	400 Mn	2400 Mn
Searches per hour per user	0.33	0.16	0.04	0.01
Searches per hour for cohort	165 Mn	160 Mn	20 Mn	4 Mn	370 Mn
Searches Per Sec	45833	44444	5556	1111	96944

 

So total Google searches per second is 96944 globally

Clarifiying questions

- When we say Google, are we looking at Google search or Google maps or gmail or youtube or google search API embedded in products

- Is this for mobile, or desktop or both

Lets say we are focusing on search queries via Google.com which may be video, image or location.

Assumptions

There are 7B people in the world, with 1.5 B in developed countries and 5.5B in developing countries

Internet penetration is 90% in developed countries, and 50% in developing countries

Google has 90% of search market WW (accounting for large share of search in India, SE Asia and lower in china)

People primarily use internet to a) find answers b) consume content c) shop and all these have a strong search aspect to it.

In a developed countries, queries/ day/ user: 10

In developing countries, no of queries a day: 5

Total queries in a day = 1.5B* 90%*90%*10 + 5.5B*50%*90%*5 = 21.15B

No of queries in a second: 21.15B/ 24/ 60/60 = ~250K queries a second

I will assume that the question only relates to Google Search Engine, and will not be look at Google Maps and other services that might generate queries. I will also assume that I am required to calculate an average QPS, as estimating min/max will involve looking at population densities across the globe and taking at the very least weekends and weekdays into account, and that would be too much for a question that should be answered within few minutes.

 

Basic assumptions about Google usage:

1. Total population - 7B

2. Internet penetration across the globe - 50%

3. China is an outlier with a pen of around 80%, using mostly domestic internet services - therefore I will not be taking its population of 1.3B into account

4. Google search engine market pen in other parts of the world - 70%

Now we can calculate our total user base - (7-1.3)*0.5*0.7 = 1.995B 

Let's segment them into three groups:

1. Power users, generating 50 queries on average, daily - 10%

2. Average users, generating 10 queries on average, daily - 60%

3. Occasional users, generating only 10 queries a week - 30%

Now we can calculate average QPS using this formula:

avQPS = (0,1*1,995*50*7+0,6*1,995*10*7+0.3*1,995*10)/(7*24*60*60)*1000000000 = (69,825+83.79+5,985)/604800*1000000000 = 263 888 QPS

 

Edit - checked out the other answers, like how structured they all are! Found an actual Google usage statistics for 2019 - 63 000 qps. Compared to others, I've significantly overestimated average daily amounts of queries per user segment, which led to significantly higher qps estimate