Estimate the memory needed for Nest cameras for the fiscal year 2020.

Note: I was not aware of the populations of the countries initially outside the US. Does anyone have recommendations on rough estimations if we use global populations? Alternatively, I could change the answer to reflect US only.

1. CLARIFY: 
   1. Are there specific types of Nests you want to focus on (Outdoor, Outdoor IQ, Indoor or Indoor IQ)? You choose. I will focus on all.
   2. Should we account for all old / existing cameras or just ones purchased in 2020? You choose. I will focus on those purchased in 2020. 
   3. Are we focused on a specific region? Your choice. I will focus on global regions. 
2. NEST BACKGROUND: Nest is part of the Google and creates smart home products. Nest produces a variety of cameras - both indoor and outdoor. There are 4 types: Indoor, Indoor IQ, Outdoor and Outdoor IQ. Nest uses a subscription service for users to record home security videos: either $6/month or $12/month. More details on the breakdown of the subscription services / video recordings later.
3. EQUATION: # of Nest devices purchased in 2020 globally x storage requirement of each type of Nest subscription service = total Nest storage for 2020
4. BREAKDOWN UNKNOWNS: 
   1. # OF NEST CAMERAS: 
      1. Global population where Nest is sold: Nest is only sold in select countries listed below with their populations. The total population of these countries is ~865M.

         1. Country	~Population in 2020
            USA	330M
            Canada	38M
            Mexico	129M
            UK	68M
            Belgium	12M
            France	65M
            Ireland	5M
            Netherlands	17M
            Germany	84M
            Austria	9M
            Spain	47M
            Italy	61M

      2. Population with Internet: Most of the countries where Nest is sold has great internet access (assume ~90%), except for Mexico (~60%). Given that Nest requires Internet connection, we can assume the users are connected to the Internet:
         1. Mexico Population with Internet: 129M * .6 = 77.4M
         2. All Other Population with Internet: (865M-129M) *9 = 662.4M 
         3. Total Population with Internet: 77.4M + 662.4M = ~740M
      3. Total Population of Someone Buying Security Cameras: I assume most buying a security camera falls between the ages of 30-65 (i.e. 65-30 = 35 age years).
         1. Average Life Span: 0-80 Years old
         2. Total Population by Age Year: 740M / 80 = 9.25M per age year
         3. Total Population of Someone Buying Security Cameras: 35 age years * 9.25M / age year = 323.75 -> Round to ~324M
      4. Urban v. Rural Split: Assume that half our population lives in the city v. country. Assume someone living in the country is slightly more likely to buy a home security device (based on my life experience). 

         1. Area	Total Eligible Population Buying Security Camera	Likelihood to Buy Security Device in 2020	Total Population Buying Security Camera in 2020
            City	324M / 2 = 162M	10%	162M * .1 = 16.2M
            Country	324M / 2 = 162M	20%	162M * .2 = 32.4M

      5. Total Population Purchasing Nest: Assume that of the people buying a security camera in 2020, 5% of those devices will be Nest.   
         1. Total Population Buying Security Camera in 2020: 16.2M + 32.4M = 48.6M
         2. Total Population Buying Nest in 2020: 48.6M * .05 = 2.43M (Round to 2.4M)
         3. Assumption: We're assuming most users are buying only 1 device, which seems most likely. There could be users buying 2 (indoor and outdoor), but, here we assume average is closer to 1 device.
   2. STORAGE REQUIREMENT FOR NEST: Nest has two subscription services $6/month and $12/month, which offer different recording options for home security videos. Let's assume the Nest subscriptions are split 50% for each type of service (i.e. 2.4M devices / 2 = 1.2M devices per service type). In order to determine the total storage for cameras, I will calculate the storage required for each type of recording. 

      1. Service	Event Recording	24/7 Recording
         $6	30 days of events	N/A
         $12	60 days of events	10 days of 24/7 footage

      2. Event Recording Storage: 
         1. Let's assume there are 10 events / day for all users.
         2. Let's assume each event records 1 minute of video for a total of 100MB / video.
         3. $6 Service: 1.2M devices * 10 events / day * 30 days * 100 MB / event = 36B MB
         4. $12 Service: 1.2M devices * 10 events / day * 60 days * 100 MB / event = 72B MB
         5. Total Event Recording Storage: 36B MB + 72B MB = 108B MB
      3. 24/7 Recording Storage: Only the $12/month service offers 24/7 recording.  
         1. Per above, we assume 1 minute of video = 100 MB. 
         2. Minutes / Day: 24 hours / day * 60 minutes / hour = 1440 minutes / day
         3. $12 Service: 1.2M devices * 10 days * 1440 minutes / day * 100 MB / minute = 1.73T MB
         4. Assumption: The 24/7 event recording is unique from the event recording storage. 
   3. TOTAL STORAGE REQUIREMENTS: 
      1. Event Recording Storage + 24/7 Recording Storage = Total Storage Required
      2. 108B MB + 1.73T MB = 1.84T MB storage

Estimate memory needed for Nest cameras for fiscal year 2020

I died on this question during a mock interview  today - sam2 path - just didn't keep everything orgnaized.  Here is my attempt at a written version.  

Clarifying questions

When you say memory, I assume you mean storage?  Yes. 

When you say nest cameras are we talking about outdoor cameras and not doorbells with cameras built in?  I’d like to focus just on the security cameras – if I do both cameras and doorbells the approach will be the same but it will add a thread to do doorbells as well.  I ‘d like to focus just on one to make life a little easier.  is that OK?  yes

You say fiscal year 2020 – But I assume you just want annualized storage and that the fiscal part question is just phraseology – yes. 

When I think about estimating the penetration of cameras, I really have no idea about camera trends outside of the US.  Is it OK if I do the US?  Sure

I’m sure that nest is used in both residential and commercial settings.  But nest was intended as a residential product – I’d like to base my calculations off of households which would exclude business uses. 

Approach

I going to do a top down approach to get the number of cameras:   population –> households –> households with Cameras –> Cameras. 

Then I am going to do a separate thread for how often the get tripped and how much storage they use. 

330 million population

Households /2.5 = 132,000,000

When I walk up and down my street, I see maybe 1 out of 10 houses with a security camera out front

Let’s assume this is typical so million 13,200,000

I think of cameras,  I think Nest, Ring, Arlo I’m sure there are others but I think those are the big ones.  When I work with large datasets I really like using the empirical rule which is that large populations fall into a bell curve and that 70% of the population falls within one standard deviation.  I’ll assume that that these three brands make up the 1 standard deviation and that the other brands fall outside.  70% = 9,240,000 cameras from those three.  I think ring is the biggest So lets go (spitball)

Ring 50%

Nest 25% = 2,310,000 households with a camera

Arlo 25%    

When I bought my Arlo’s I could buy a 1 pack or 3 pack.  I’m sure that some people have more but lets assume they are outliers and the avg is 2.  So nest 4,620,000 cameras.

The cameras are triggered when ever someone walks past.  Mine get triggered a few times an hour during the day and way less at night.  I’ll assume I am typical.  Peak times 12 hours, off peak 12 hours.  On peak 4 times an hour that is like once every 15 min.  Off peak say 1 time an hour..

4x12 = 48+12  (60 triggers per day).  Each triggers records 1 min of high def video.   High def video is ~100 mb per min. 

So each camera generates 6000 mb of video per day.

I forgot there is a paid version and an unpaid version.  I don’t remember the difference of storage – so unless you have that info, I’d like to assume that paid will keep your data all year and free will keep your data for 30 days.  Using the empirical rule again those that pay for the service are on the outside of 1  standard deviation so 15%  

Free version - 4,620,000*.7*6000mb*30 days  580,212,000,000 MB free version

Paid version - 4,620,000 *.15*6000mb*365 days = 1,517,670,000,000 mb paid version

Total storage required 2 trillion MB per year or 2 EB per year. 

Limitations

We made a lot of assumptions and if any of them are wrong there is a cascading impact.  I think there are a few big one’s that we should be aware of.

1.       Empirical rule – we did a lot of smoothing using the empirical rule High users will be balanced off by low users so the 60 triggers per day – we could do a high medium low but I assume they will even each other out.

2.       60 triggers per day – this is a spit ball and could be off – it’s based on my experience as a model user

3.       10% of houses have a camera – this is a spit ball based on my recollection walking down my street.  Recollections can be wrong, and my sample may not be representative. 

4.       records for 1 min by default.  If the default record time is 30 sec’s it cuts the whole thing in half.

a.       I did not compensate for some videos being longer – but given the number of other spitballs – I don’t think it’s going to change the closeness to accuracy. 

5.       Storage free version is 30 days – if it was 7 days that would be a pretty big reduction (also if the free version doesn’t record at 1080p – 720p is like ½ the storage.

Clarification is important before answering this Google estimation question.

• Are we trying to find the total amount of data needed to store all videos from Nest users? -- Yes
• Are we looking globally or just in the U.S.? -- Just the U.S.
• What kind of plan/pricing does Nest use for storage?
• $6/month for 30 days of event videos, $12/month for 60 days of event videos + 10 days of continuous video

Okay great, at a high-level we want to solve the following equation:

• (# of $6/mo users)*(Total Data per 30 days of event video) + (# of $12/mo users)*(Total Data per 60 days of event video + 10 days of continuous video)

• I'm going to assume that most users have more than one camera. Having only one camera seems unlikely since I'd imagine these can be bought in bundles. Also for most people, one camera likely isn't enough. 3+ cameras is probably in the long tail of the market as well. So 2 per user seems about right.
• 1 of the 2 cameras is probably located on the outside of the house. There's an average of 2.5 people per household in the U.S. If each is leaving and entering the house once per day that's ~5 events per day for this camera.
• The other camera is probably in the living room or some place similar. I imagine when everyone is home these devices are smart enough to not go off constantly. Since they'll primarily go off when no one is home lets say this is also capturing ~5 events per day.
• That's 10 events per user per day.
• Let's assume it's 30 seconds per event video. That's 300 seconds or 5 minutes of event video per day.
• I know my iPhone videos are ~20MB / minute. This means there's 100MB of event video per day.
• Multiplied over 30 days this is 3GB per 30 days. Similarly, for 60 days this is 6GB.
• For the 10 days of continuous video, using the same size/min, we get 300GB

• I will take a top-down approach to figure this out.
• There are 330M people in the U.S. and ~2.5/HH. Rounding this to 3, this is ~110M households.
• I don't know what the penetration rate is for the internet connected home security camera market as a whole is.
• That said, I would think that it varies quite a bit by age of the household.
• I'll calculate by household age group the % that use a home security camera and of these the % that are Nest customers.