Walmart Chile
Supermercado Lider
Lider (Walmart Chile) is the cheapest and the biggest Supermarket in Chile.
It's value proposition is to make customers save time and money with their slogan "Everyday low prices."
Data value exploration (2021 - Current)
Making experiments with data models.
The mission of this area is to discover opportunities or needs to solve with data
intelligence through data models, machine learning, and advance analytics.
I've been working on understanding how to spread the knowledge of experimentation and need-finding to product teams,
to give them tools for creating hypotheses of things they need to try in an experiment as CRO or data intelligence.
I feel exiting about the challenges of evolving the user experience with data, learning how to work closely with data scientists or data areas.
I created this logo!
#1 Experiment: Own Brand recommendation at checkout
Save money shopping within Own Brands.
My participation: I brought the idea to the table and prototyped it.
Strengthening Own Brands turns into savings for the business and generates more sales margins. And for our customers, buying Own Brands is cheaper.
Hypothesis: By offering lower prices within our Own Brands at the supermarket e-commerce checkout, customers will consider the offer a better option, switching the actual products they already have at the shopping cart to the cheaper ones.
KPIs: We considered three metrics to validate the hypothesis.
- Own brands sales share
- Product replacement at the checkout
- Average customer savings
How did we test it?
1. Prove of concept
To prove the concept, We built an experiment using Google Optimize, where we offered a replacement for only 20 products with its patterns. With this first approach, we wanted to check the desirability of this feature.
We ran a test with a low audience; 30% of desktop users, where 15% of users were part of the treatment group, and the other 15% were part of the control group.
The results of this first experiment weren't statistically significant, but we proved that customers were clicking the banner for replacing X products for Y "Own Brand" ones.
2. Grow the test
Own Brands are part of the company strategy, so our stakeholders were excited about growing the experiment.
The second iteration was about growing the audience to 50% (25% treatment/25% control) on desktop and mobile and the number of products offered at the first test to all the products that have a one-on-one pattern for replacement.
3. Testing Data models
We finished iterating this experiment at the end of 2021.
The final test had two treatments and one control group:
- 33% Multipattern recommendation. For E.g. rice branded "X" or "Y", could be replaced by Own Brand rice.
- 33% Multipattern and products recommended by a "Recommender model". For E.g. rice branded "X" or "Y", could be replaced by Own Brand rice or other product recommended by the model.
- 33% Control group
In terms of results and KPIs, in 50 days:
- Own brand sales share: The control group had 15,16%, Multi-patterns group had 15,10% (-0,4%) and Multi-patterns + recommender model had 15,19% (+0,2%).
- Product replacement: Multi-patterns had 1.260 replacements and Multi-pattern + recommender model had 1.421 replacements (+12,8%).
- Average saving and total saving: Multi-patterns had an average saving per shop of $498 and a total saving of $628.527. And the Multi-pattern + recommender model had an average saving per shop of $677 CLP and a total saving of +58,6%.
(The currency we used is Chilean pesos).
#2 Experiment: Promotions recommender
This advanced analytics model named Cheapy recommends whether or not a product has to be part of a promotion listing.
My participation: User journey, user testing.
Hypothesis: By creating a promotions recommender model, the products and categories will perform better on each promotional listing, increasing sales.KPIs: Incremental sales %.
How did we test it?
1. The model
The Data scientist team created this model, considering several inputs that affect the results of the outputs. Those depend on the business, the context, past listings, Suppliers, etc.
The output tool is a spreadsheet that suggests the products on every list.
The solution was well-received, but still, there was a big gap to understand the entire promotional process in terms of the people who work on it.
2. Understanding and empathizing
I started helping in this experiment when it was in the middle of the road. I found that one of the issues was the understanding of the promotional workflows of each user.
I interviewed users to check their understanding of the tool and learn how they use it. I surprisingly found the information brought in the spreadsheet was okay, but they weren't receiving the file at the correct timing.
To solve that, I worked together with other UXers who work at products that intersect in the promotional experience, and we facilitated a journey map session together with users.
3. Next steps
After discovering the opportunities for improvement, I delivered the findings to the Data scientist and Commercial teams to improve the product.
The big next step is to deliver the model to the team who will adopt it into their product.
I'm working together with my UX coworker in this process to prototype and test the best way to integrate Cheapy.
In terms of results and KPIs, after a year:
Cheapy is the more successful model and experiment of 2021. The percentage of incremental sales it brought to the business is significant, and that's why we are working on delivering and implementing it as a product or as part of a product.
#3 Experiment: Improving the experience of stock availability on e-commerce.
Save money shopping within Own Brands.
This experiment includes the stock prediction model and the substitution suggestion model to check the stock of products at e-commerce, predict the "out of stock" possibility, and offer a substitution to customers before the checkout.
My participation: Prototyping
Hypothesis: By letting our customers know if a product would be potentially out of stock at the picking moment and giving them the possibility of switching it for another product that the model knows it's available, we will deliver a better customer experience.
KPIs:
- NPS
- CSAT
- Order completion
How are we testing it?
1. Prove of experimental design
Using Google Optimize, the team set up the experiment for 10% of the users, where 5% is the treatment group and the other 5% is the control group. It's very recent, so we haven't concluded anything yet.
2. Next steps
The team will set up the experiment for 100% of users, where 50% will be the control group.
3. Prototypes
The first prototype shows the message for products that the model predicts as possibly out of stock.
This second prototype shows another way to test the model at the checkout, showing the alert in products in the shopping cart, and allowing users to choose a replacement in case the one they bought is out of stock when pickers are putting together the order.
Growth hacking Walmart Chile (2020)
Supermercado Lider
In the supermarket area (grocery shopping), we created the growth hacking team.
Our purpose was to find exponential growth ideas to increase the KPI we established based on AAARRR metrics on each experiment.
We organized the team, adapting the agile methodology into growth. Every two weeks, we invited coworkers from other teams to brainstorming sessions, where we proposed a topic to work on ideas.
These are some of the experiments we did:
#1 Experiment: Referral
Share my shopping-cart
We did this experiment to measure the desirability of having the share feature.
KPI: CTR
Hypothesis: Having a "share my cart" button, people would share their cart with others.
We built an A/B/C test with two variants: share my cart and share my cart on What's App.
Results:
Conclusion: The ~2% of the customers were likely to share the cart, and the share on What's App button was more atractive than the other (+47%).
#2 Experiment: Awareness
Tik Tok Christmas challege.
It's the first Tik Tok challenge for Lider. KPI: Number of participants.
We asked Itay Vargas, a Tik Tok influencer, to invite his audience to participate challenge, showing Lider's products.
Hypothesis: Creating a Tik Tok challenge through an Influencer would attract more than 1.000 people to participate.
Here is the challenge: https://vm.tiktok.com/ZMJWutBFJ/
Results:
Views: 1.9M
Likes: 700.000
Engagement: 4.9% ([(Likes + Number of comments)/Number of followers]*100)
Challenge: 1.500 participants
Cost per view: $2.5 clp
Conclusion: Generating engagement in social networks gets the brand positioned into younger generations. Some studies say it also reduces churn.
Walmart Kiosk
UX/UI at product
As a senior UX designer, I worked with product and tech teams, leading the digital experience and the in-store service of the Kiosk. I was part of the general merchandise pick-up service discovery and creation.
Case study:
General merchandise Pick-up service.
The context:
In Chile, on October the 18th, we had riots because of political issues about the cost of life and non-equality. People were looting and setting fire to retail chain stores, drugstores, and our supermarket stores. To make it easier for customers to shop General merchandising stuff without paying the delivery fee, we had to implement a Store Pickup service for general merchandising. (By that time, we only had a pickup service for grocery shopping).
How did we start?
We needed to understand how the service workflow would be, so I facilitated a Roleplaying session at the store with all the roles at the play to discover all the touchpoints we were missing to cover the service well.
Roleplay sessions were:
1. When the customer goes to pick up at Customer Service area
2. When the customer goes to pick up at Electro area
3. When receiving the package from the distribution center to the store. (How, when, and where would be storage before the customer pick it up).
At every stage, we had a whiteboard to write every insight we discovered from the session.
The results:
Every area involved took the insights and issues they had to solve, and we mapped them in an event storming session and followed up with them every week until we launched. After launch, the company created a dedicated team to continue and improve the service.
Take aways:
When creating a new product or service, it's a must to plan it with all the actors in play from the beginning to check all the details of each area involved.
