I would clarify what the product is in the interviewers eyes to make sure my assumptions of Amazon Go from what is publicly available is aligned. Assumption to clarify – expedited grocery store with smart tech identifying who is shopping and what they’ve selected while shopping to charge their Amazon account rather than making them spend time in line at checkout.
Now what is the goal and purpose of Amazon Go? In my clarification with the interviewer it was hinted that it’s the viability of a grocery store without checkout. I’d explicitly clarify that goal with the interviewer.
Say we align on this goal…”Prove Amazon Go concept is viable”
Now we need to determine what success for that goal looks like. I’d discuss options with the interviewer to align how we measure success.
A few options include:
1) number of successful shopping transactions?
2) number of repeat customers above x%
3) number of items in checkout
4) average $ value of shopping cart at checkout
We would also need to have a standard grocery store with normal checkout to compare as an A/B test to insure the metrics above are at parity with or better than standard grocery stores. Reason I say this, if Amazon Go launches and gets 1M users right away it may not be a success if they spend less in shopping cart value, buy less profitable items, etc. In grocery business margins are key, quite often grocery stores will have a loss leader item that attracts users but shoppers have to spend more than the loss leader loses to make it worth it. Will discuss this further during pre-launch and launch.
-Goal: Prove viability of Amazon Go concept by proving that Amazon Go saves shoppers time shopping while causing no harm to required margins to stay profitable”
I would have two stores with the exact same suite of products including loss leaders to attract users, price points, etc. This will allow me to compare impact of the time savings without checkout on core metrics for grocery stores of margins, total $ volume, and # of users = profit.
Now we need to clarify who the target user is and location of launch along with the strategy to get target user in target location.
I would look at several possible target segments of grocery shoppers:
a) young (18-30) single adults
b) married but no children/dependents (likely 30-50 years old)
c) married with children/dependents
d) married and/or single but older retired with no children/dependents
I would focus on segment C) as that is the group that likely spends the most money on groceries. You could get more narrow by sub segmenting this group into high income, low income, etc. but for simplicity lets focus on the broader whole of segment C).
What’s the focus area for launching the product?
I’d target location C) here as that gives the highest opportunity to capture the most grocery store visitors and making some assumptions I’d propose that metropolitan life is busier and users in this area value time saved shopping versus the more laidback lifestyle in rural and suburban areas.
Also going to go with a metropolitan area where we have a presence to easily monitor and test the concept in with quick and frequent changes.
We need to focus on announcing and drawing target users into store once it’s launched. We’re in a metropolitan focusing on married couples with children/dependents. Making assumptions here but going with the fact that they read newspaper, watch the news, and are active amazon app users. Would run campaigns in all three mediums announcing the launch of the store.
I wouldn’t mix things up too much during launch as I don’t want to muddy the variables of the A/B test. If there are more than the above stated goals we could always run a multi-variate test with 3+ stores with each having a different mix to find out which one most greatly impacts revenue. I’m sticking with one control store and one test store for simplicity of first test and MVP.
This is where critical measurement of core metrics matter.
I would look at:
1) total number of customers from test store with Amazon Go compared to total number of customers at control
2) then compare number of successful transactions (users bought items) for both stores
3) I’d look at time from entry of store to exit for successful transactions of users at both stores
4) I’d look at total $ of purchases at both locations
5) I’d look at avg $ of each shopping cart at each location
6) I’d look at profits from each location
If the number of shoppers at one location was different than the other I’d take a random number of users from the store with more shoppers equaling the store with less shoppers to help compare exact numbers apples to apples on top of totals. For example, if Amazon Go had less shoppers but was more successful in terms of profits, $ volume spent, less time spent in store per successful transaction but had less total shoppers you would consider it successful and see if you can run another test to find ways to attract more shoppers.