1. Clarify the question
   1. What is a refrigerator? Can we assume it is a storage device to keep food from spoiling? (Interviewer “That is correct”)
   2. Are there any constraints or goals I should know about? (interviewer: “You can decide”)
   3. Is the sales estimate for US only (interviewer: “Yes, that’s correct”)
2. Set goals
   1. Fridge should have strategic fit with google’s existing business and other products
      1. Nest (becomes another device within user’s nest household)
      2. Ad targeting (we can capture rich data on food consumption habits)
3. Define users 2. Consumer 1. Admins — Household members who stock the fridge 2. Users — Household members who consume items from fridge 3. b. Business

4. Define pain points/needs/user goals

5. Define Features

c. Implementation

6. Summarize/Evaluate

So in the end we have a new kind of Nest device. It integrates into the user’s Nest household like any other Nest device. Admins and users can both access the device through the app, albeit with different privileges. For example, we would not want an office employee changing the ordering preferences for the entire company. In terms of capturing data to improve ad targeting, I think this will be challenging because its hard to track an individual user’s behavior. For example, I may be a big milk drinker, but the fridge only knows that a lot of milk is being consumed within my household, not necessarily that I am responsible. So the the data is very granular on the household or employer level, but not the user level. Lastly, refrigerators are big ticket items and have long replacement cycles so I’m not sure we can sell enough to capture a lot of data. Let’s examine this now.

7. Sales estimate

My model will work as follows:

Let’s run the numbers…

• There are approx 100M US housholds
• I will estimate the replacement cycle at 10 years
• That means 10M units are sold per year
• Fridges are a competitive market and I think consumers are looking for trusted brands. Therefore I think Google will have a hard time breaking through unless they partner with Samsung, GE, etc. to provide an OS type solution.
• Assuming Google does this, I think they can get mabye 2% of the market in year one

So that’s 200K units.

Clarification

• What are we launching this product for? —> To enter into the smart home space
• Which geo is this product going to be launched in? —> We are launching this in the US
• What are our main goals with this product offering —> Main goal are getting adoption
• Who is this product offering for consumers or enterprise? —> Lets focus on the consumer users

Goals

• We are building a refrigerator to increase our portfolio of products in the Smarthome areas
• This will be launched in the US and we are focused on increasing user adoption

User Personas

Lets list down the user groups

• Families: Families with young children. Frequently use the fride and freezer for kids meals, school meals, snacks. Actively restock fridge every week
• Couples: Young couples that have an active social life. Order in, Eat out and entertain ad lot. Spurts in terms of produce purchases
• Students: Focus purely on low cost and stick to staples.
• Old people: Stable lifestyle in terms of predictability of meals/
• Health conscious people: Do a lot of meal prep/ cooking. Buy fresh produce frequently and cook quite often.

From the above segment of users, I would focus on the Families user persona because

• This group typically has a high level of usage of the fridge
• This group is usually strapped for time due to balancing kids, work, household chores

Use Cases

Some of the use cases for

1. Stock: Show stock levels of supplies in the refrigerator via alerts sent to an application or email
2. Alerts: Alerts when items are running low on stock or reaching their expiration date
3. Deals: Alert about deals on commonly purchased items from different grocery stores
4. Grocery Lists: Auto create grocery lists based on supply levels and recommendations of items purchases previously
5. Ordering: auto-order groceries from stores based on supply levels and historical buying patters
6. Recipes: Show recipe recommendations based on items in the refrigerator
7. Events: Schedule important events such s birthdays, play dates, sleepovers, activities and set reminders for them

Prioritization

I would like to prioritize the above use cases using the RICE framework (Reach, Impact, Confidence, Effort)

1. Stock: (6, 3, 3, 2) —> 27
2. Alerts: ( 8, 3, 3, 3) —> 24
3. Deals: ( 5, 2, 2, 3 ) —> 6
4. Grocery Lists: ( 7, 3, 3,3 ) —> 21
5. Ordering: ( 7, 2, 2,3 ) —> 9
6. Recipes: ( 5, 3,2 ,3 ) —> 10
7. Events: ( 5, 3, 3, 2) —> 22

Based on the prioritization I am going to focus on the Stock feature to begin with

Solutions

1. Separate Cabinets: Have separate cabinets for items. User to input what itmes were loaded into which cabinet. Build pressure sensitive base to track product weight and track the product’s expiry date. Use this information to alert on stock patterns. User alerted when supply is running low
2. Bags – create separate bags for storage. Each bag would have a unique bar code which would get scanned when loading / unloading and track product expiration based on the details of the product. Weight / quantity would not be captured in this case unless when items are loaded and kept in the fridge there is someway to record the weight of the item based on which the barcodes are being scanned.
3. Track staples: Build special compartment for staples such as milk, eggs, dairy items and meat the in the freezer. These items are loaded into the specific location marked for them and the expiration dates are approximated based on common knowledge. Weight is tracked correctly.

From the above three options I would choose option #3. While it isn’t most expansive in terms of features it removes cognitive load from the user as it doesn’t doesn’t require the user to categorize the items or doesn’t require additional work such as partnering with grocery stores for getting the specific bags per item.

Market Size

US population – 330M

• Avg. Household size = 3
  • Households = 330 / 3 —> ~100M
• Households with children under 13 – 20% (Assume even distribution on age groups and 75 years as life expectancy. Therefore each year = 100/75 is ~1.33% and 13 years would have ~20% of the households )
  • Total households —> 20 million – This is your total addressable market for the first version.

Now before estimating the actual buyers we need to assess the price point for the product as that will shave this number accornidnly. If too high then only the affluent would buy it and if we are able to bring it low than it open up a much bigger market i..e we can go after the 20 million households.

However, given technology adoption and the fact that the price Is likely to be high as we haven’t yet hit economies of scale I would guess that we get around 2.5% penetration in year one. That would mean around 500K units sold in year 1.

Summary

To summarize I would build a smart refrigerator targeted at families with young children (<13 years). We believe this segment would benefit from the product and also have the buying power to spend on this. Based on estimates I believe we can get around 500K units sold in year 1 of the launch.

Me: First, let’s design our product and then we will calculate the estimated amount of units sold. Before we begin, is there anything specific about this refrigerator, is this a ‘smart refrigerator’?

Interviewer: Yes, you can consider this is a smart refrigerator but the overall design is up to you.

Me: Thank you, one more question, do we have a specific user in mind? Younger people? Retirees?

Interviewer: No, you can select the user base based on your prioritization, but theoretically it could be a refrigerator for everyone.

Me: Perfect, sounds good. So, the way I want to approach this problem is: first, I want to take a look in a category of users, second, after prioritizing a user base, I will talk about the pain points of current refrigerators and smart refrigerators. Then, after prioritizing the pain points we can talk about the product vision and its features and we will continue from there, is that ok?

Interviewer: Ok, sounds good.

(users)

Me: Ok, so I can think about four categories of users:

1. family households with 3+ people that are usually composed by a couple and at least a son
2. single people who share apartments
3. single people who live alone
4. retirees

Given that categories 2 and 3 are mostly composed of people with that are more tech-savvy and willing to spend money in novel products, I will focus on them, specifically category 3, who are usually composed of busy singles that would usually have higher disposable income.

(pain points)

For single people who live alone, I have a few pain points that I can describe:

5. single people usually don’t track well when they are running out of a specific product, and usually, only realize that once it’s almost over
6. given that they live alone, they usually don’t stock groceries and whatever is in the refrigerator it is their stock for the week or every two weeks depending on the frequency they go do groceries or order groceries online
7. singles often forget when they put the product in the refrigerator and when they first opened it, which can lead to food spoil
8. on the flip side of that, sometimes a single buy a too large of a product (like large milk) and it spoils because it’s not consumed in time

(vision)

While I do believe that all the pain points are valid and should be addressed by a smart refrigerator, I want to prioritize the first two. For that, let’s come up with a product vision: imagine a smart refrigerator that replenishes its products automatically by understanding the consumer habits of its user and the products he/she uses while recommending new ones to try. This product would have the following features:

(features)

9. it would identify the products when the user first puts them in the refrigerator and would be able to calculate how much of it was already consumed on a given time
10. through ML algorithms, it would learn the consumption habits of the user and based on that would automatically order replenishment from a local grocery store
11. by identifying the products, it would often suggest other products similar to those that are better/premium or that have a better cost/benefit based on settings defined by the user

While this product would be a success for busy professionals, there are some trade-offs we should think about:

(trade-offs)

12. the products would have to be identified and likely placed in a pre-defined area in the refrigerator to track its usage, which can be a bit time-consuming
13. we are assuming that the users do not change habits very frequently in order to keep replenishing and offering similar products, for example: if I got a milk chocolate today, am I going to keep ordering it every week or I am going to change it to orange juice? While a user’s habits could be learned through ML, initially one would need to change that in the refrigerator configs and this might cause a decline in ‘productivity gain’ that is the product’s value add.

(metrics)

Finally, I would try to evaluate a few metrics in the product to make sure that our value adds thesis hold true:

14. how often our product recommendations are working?
15. is there any churn rate on automatic replenishment after X number of weeks?

Now that we have a product, we can estimate its market size, is that ok?

Interviewer: Sure! But can you give me a summary before we move on to make sure we’re on the same page?

Me: Definitely. We just created a smart refrigerator targeting busy singles that is able to learn its users’ consumer habits and products in order to be able to 1) provide recommendations on similar products and 2) automatically order from local stores so that the user never runs out of the product.

Interviewer: Sounds good, now let’s move on to the estimation part.

Me: Ok, so to estimate our market let’s first calculate the population of busy singles in the US that could buy our smart refrigerator, then we calculate the ones that would effectively buy it at some point and finally we would select the ‘early birds’ who would be keen to buy it in the first year, how does that sound?

Interviewer: Sounds like a good plan.

Me: Cool, so let’s consider the population of the US as 320 Million and we have four segments with roughly the same amount of people: 0-20, 20-40, 40-60, 60-80. For the sake of this estimation, I am going to consider that only people on the 20-40 segment are going to buy our refrigerator, while I do believe some people on the other segments would buy it, I think that 20-40 would be the largest segment given our target audience (busy singles).

That said, we have 80 million people in that segment. I also believe that around 50% of that segment is still single, so that would make the number down to 40 million.

Next, let’s calculate who can effectively buy the refrigerator – given that this is going to be a refrigerator more on the higher-end, I believe that only middle class and upper-class people will buy it initially in different proportions, and based on common sense I would think that about 10% of those singles could be considered upper class while 60% could be considered middle class. So we would have:

Upper-class – 4 million singles

Middle-class – 24 million singles

Now, let’s calculate who of those are likely going to purchase a smart refrigerator over the next few years, and I believe that the proportion is higher on the upper-class segment – 80%, and a bit lower on the middle-class segment – 30%. We now have:

People who will buy smart refrigerators over the next years:

Upper-class – 3.2 million singles

Middle-class – 7.2 million singles

Finally, we need to calculate how many of those will buy our refrigerator when it’s launched in the market during the first year and I believe the proportion will be considerably higher on the upper-class segment (10%) compared to the middle-class segment (1%):

Upper-class – 320K

Middle-class – 72K

Total sold the first year: 392K

Though I don’t know about the refrigerator’s market on an annual basis, this number seems to be small. Now, let’s evaluate some of these numbers:

16. the first thing that comes to mind is that we are only considering the busy singles segment, and even though our product is targeted to them, very likely we are going to have a considerable number of people from other segments buying our refrigerator as well.
17. perhaps given the novelty and leveraging the Google brand, likely that more people are going to be purchasing the product, so the proportion of early birds that I assumed is probably in the lower-end.
18. finally, we are assuming that only a part of the population are going to be purchasing smart refrigerators over the next few years. Very likely that those are going to become the norm like smart TVs and comprise the large majority of the market if not all of it.