Design a chat-based product recommendation engine for an online flower/gift store.

Question

Design a chat-based product recommendation engine for an online flower/gift store.

Asked at Shopify over a year ago

1.6k views

Product Design category

Asked at

1.6k views

How to answer Product Design questions

Interview Guide

Top Product Design interview questions

How would you design a bicycle renting app for tourists?62 answers | 82.5k views
Build a product to buy and sell antiques.54 answers | 66.8k views
How would you design a web search engine for children below 14 years old?36 answers | 42.9k views
See Product Design PM Interview Questions

Invite members
Invite by email

Add another

sairah814 · Answer 1 · 2021-10-13T15:51:51+0000

First I want to clarify some assumptions.

1. Flowers are fairly perishable. Does this store sell any other perishable items like food arrangements (ex:chocolate strawberries etc)?

2. What would our goal be here? My assumption is that we want to increase revenue by offering product recommendations that (1) increase likelihood to convert (2) increase basket size and average order value.

3. Are we doing this for mobile app or a web based solution? I would choose mobile app just given the growth of online shopping in this channel

4. I also want to focus on the US market

5. How large is my customer base? Ex. specialty shop that's newly launched vs. Amazon. I will assume it's a smaller, newly launched online store that is only focused on flowers/gifts

Next, I'd want to understand who we're building this for. I think there are several different types of shoppers/gift givers.

1. Regular flower purchasers - enjoy having flowers in their home and purchase flowers on a weekly basis. Care about seasonal trends

2. Occasion gift giver - Wants to buy a gift for their signficant other, family member or friend for holidays and birthdays. Doesn't have a lot of time to think through the options and willing to pay for something reasonable that looks thoughtful without thinking too much about it

3. Sentimental gift giver - Wants to customize the gift for their friend/family/partner and feel involved in the process of selecting the best gift or flowers.

I think the largest target group is the occasion gift giver, so I will select that one.

Next, I'd want to understand some of the user needs of this group.

Want to find a gift that is relevant to the holiday (ex: roses for Valentine's day)
Want the gift to look thoughtful (ex: card, additional item that makes it feel like they went the extra mile)
Want the item to arrive on time (for occasion) especially when I order something last minute
Want flexibility to pick up or deliver the item
Want the item to look fresh

Next, I'll suggest some solutions to address these needs.

Solution	Impact	Effort
User chats specific holiday and details on gift receiver (ex: spouse, female, allergic to chocolate) and chat bot provides a suggestion	High - helps user achieve immediate need without too much back/forth	Low
Based on items in the current basket and purchase history, user asks what else should they add on and chat bot provides relevant suggestions (may need use ML to achieve this at scale)	Medium - as a newer store, I probably don't have as much data on purchase history and can't make as good recommendations on current basket	Medium
Use machine learning to recommend items that are trending in the market that match the holiday (ex: digital frames as a mother's day gift)	High - would provide great, unique recommendations	High
User types in preferences and chat bot provides multiple gift options that the user can say yes vs. no to and the chat provides additional recommendations based on their answers	Medium - may get frustrating for the user if the options don't feel relevant to them	Medium
User is purchasing gifts and chat bot asks whether user has thought of gifts for upcoming holidays. user can answer questions about who they want to buy for and schedule future orders	Medium - focused more on user needs related to resolving future asks rather than the current need. Most users are shopping for the holiday closest to them	Medium-Low

Based on the above, I would make the recommendation to go with option 1. For a newly launched store, this would be lowest effort to provide the most relevant recommendations.

Arkapravo Chakraborty · Answer 2 · 2020-04-14T21:08:52+0000

Design a chat-based recommendation for online flower/Gift store

Clarify the questions

Who is it for – The visitor of the flower/gift shop?

What is this – an embedded chat wizard that enables users to find the desired gift quickly

Why a chat-based recommendation – Easier Product discovery, personalization, Quick conversion

User Behavior

Wants to gift to friends/family for a special occasion. The receiver could be in the same city/different city
Could order an online gift to deliver at home before attending a function
Home décor
Search ideas what can be the gift based on knowledge about the receiver
Takes a leap of faith that the gift will make a positive experience for the receiver

Use case –

Quick discovery - As a user I should be able to quickly discover the product based on my preferences

Personalized recommendation - Improve receiver experience - Create a good experience or improve the experience of the receiver

Prioritize

Quick discovery (User impact – High, Revenue – High, Cost - Low)
Improve receiver experience (User impact – High, Revenue – Low, Cost – High)

Solution

Use case – Quick discovery

The system should ask questions-

Receiver address – to narrow down the options
Receiver demographics - gender, Age
Relationship with the receiver – Mom/Dad/ girlfriend/Boyfriend
Purpose of the gift – Birthday, Marriage anniversary
Type of gift – Flower? Gift hamper?
Budget?

Show items recommendation

Show what is hot.
Show how many people considered this.
Show items left.

Package recommendation

Special notes

Compare multiple selections

Add to cart/Direct checkout

Use case - Improve receiver experience

Solution

Users give feedback about the receiver experience of the previous purchase
The recommendation system takes that into account and recommends people who like this also like this kind of recommendation

MVP consideration/Summary

The system should ask questions-

Receiver address – to narrow down the options
Receiver demographics - gender, Age
Relationship with the receiver – Mom/Dad/ girlfriend/Boyfriend
Purpose of the gift – Birthday, Marriage anniversary
Type of gift – Flower? Gift hamper?
Budget?

Show recommendation

Show what is hot?
Show how many people considered this?
Show items left?

Package recommendation

Special notes

Add to cart

Success Criteria

Acquisition/Activation

#of users using the feature
%of New users using the feature
%of repeat user using the feature

Engagement-

Average time spent on chat
Average No of step completion
Conversion rate

Retention

%repeat usage in 7 days, 30 days, 6 months
Churn rate
Exit rate while using the feature

Monetization

average revenue per user
GMV sold through chat

KN · Answer 3 · 2020-04-11T10:35:23+0000

So, the recommendatio engine should take few data points from the end user's data and then ask few questions about the particular need and then match the two and recommend flowers or flower shops as the case maybe.

What are the things we would want to know about the user about himself?

First data point we will ask the user is the city where the flowers are to be sent.

Second data point we will ask him is his Name and Gender.

Third data point we will ask the user is the occassion of flower gifting.

Fourth we will ask for the relationship of the user with the person whom he is gifting flowers.

Fifth, we will ask if the person for whom the flowers are meant for has any preference for a particular colour of flowers.

We have enough data.

We will use the city of the person to show the local flowers or the flowers which can be shipped there.

We will use the gender to draw on our DB to see what particular type of flowers girls and boys want.

We will use the third data, the occasion to know the type of flowers to recommend. Anniversay/valentine will direct us towards Red Roses etc.

This will be further strengthened by the relationship between the sender and receiver. If i am sending flower on the birthday of my brother vs on the birthday of my girl friend

Finally, when we have more than one type of flowers, we will ask the color preference of the receiver if any.

I have intentionally not asked the price here. As people will quote a lower price always and the recommendation might be skewed towards business goals of revenue.

After we match all this data, we will some to a subset of the flowers we have and which can be shipped to the person and which meet the occasion and relationship constraints.

We will recommend the flowers and give options to sort by Price, Relevance, Shipping time and filter option by Type of flower.

We will also give customisation options of a gift card with personalised message once the user has selected the flowers.

We will also recommend what other users with similar data bought using collaborative filtering after the payment has happened or order has been placed.