Insights/Action/Outcome: With Amplitude Analytics and the use of alerts, Mercado Libre noticed users opened and closed the application on screens where this was not the expected behavior. After an investigation, the team noticed certain versions of the app did not allow users to move freely when registering a credit card. Fixing this issue reduced abandonment at a crucial step in the affected app versions from approximately 20% to 11%. During this time, overall checkout conversion rates improved 12%, with new user conversions up 11%. The resulting cumulative effect is expected to translate into a 15% increase in monthly revenue.
I always say that we know our users by their experiences and that first impressions always count. For example, if a purchase generates a lot of revenue, but the user encounters too many friction points to fulfill that purchase, their bad experience will likely mean the user will not return again. In other words, that individual can give me revenue today but will not be a returning user. We always want our users to return and we create the experiences to make that happen.
For example, I can focus on improving the conversions of a payment selection screen—which will lead to increased revenue—but it is always important to go further. Driving conversions is about, among other things, improving user retention and knowing all the steps you have to go through in order to meet your goals, because a positive user experience guarantees that users continue to choose us and, ultimately, that’s what we want.
And that is why improving micro conversions and elevating the user experience is vital for Mercado Libre because in the end it will improve conversions.
The first impression always counts. A purchase conversion can give me revenue today, but if the flow is not completely clear and the user encounters too many friction points, regardless of whether they have been able to make the purchase or not, they may not return and we lose the customer forever.
Mercado Libre is a household name in Latin America, but those outside of the region might not know that we are an ecommerce platform with more than 200 million registered users across 19 countries. We are like the eBay or Amazon of Latin America—on our site or platform, different buyers and sellers of services and products interact.
An additional part of our ecosystem is Mercado Shops, which allows sellers to create and customize their own virtual store while receiving logistical help and support from Mercado Libre. These support systems include Mercado Pago, a fintech company handling all payments and transactions as well as a virtual wallet, and Mercado Envíos, our shipping and logistics arm. Everything related to buying, selling, payment, and shipping of products happens in-house, which makes it an interesting place to work, since there’s always some new project to work on.
I’ve been at Mercado Libre for nine years, working with data in behavioral analytics and user experience. The data team is 300 to 400 people working in five data verticals across the organization, and I’m the technical lead in the data analytics vertical. Our focus is always on improving conversions by understanding user behavior and improving the user experience. I lead the implementation of the CRO (conversion rate optimization) methodology at Mercado Libre and throughout Latin America, where CRO as a methodology is still underdeveloped.
Conversion rate optimization is less about the end goal than it is a constant search for insights to improve a user’s experience as they move toward that goal.
CRO is less about the end goal of conversion than it is a constant search for insights to improve a user’s experience as they move toward that goal. We, as analysts, must put ourselves in the position of the user, since we cannot improve something that we do not know about. It is necessary to put yourself in the user’s shoes and go through the same thing, to be able to understand the process and what its critical points may be, and improve them.
The CRO methodology is a lot about analyzing conversion flows, assessing where user friction may exist, posing hypotheses, and running experiments based on those hypotheses. For example, we go to the shopping cart flow, we analyze the flow and we notice that the greatest abandonment occurs in the “shipping method selection” step. A hypothesis we propose is that for certain products, the shipping price exceeds the price of the package to be purchased. In that case, we raise the hypothesis and set up an experiment where for certain products and packages to be sent the shipping price cannot be higher.
After the implementation of the experiment, we analyze it. If favorable, we generate an effort/impact matrix to establish how much profit we might obtain if we put the experiment into production. Once it has been put into production, it is analyzed again, especially to ascertain whether the favorable change is due to seasonality or not. Then the process starts again, since CRO is cyclical.
An incomplete, unreliable data picture comes at a cost
Until recently, Mercado Libre didn’t have the necessary tools to perform this kind of multi-step conversion analysis. For example, Mercado Libre has around 15 steps in a typical purchase flow, but using an amalgamation of tools meant we never had a complete, real-time picture of the user’s movement through that process.
Google Analytics gave us some data, which we sent to BigQuery, but it was not enough since we did not have real-time data or we could not complement it with transactional data. Faced with the problem of not having a tool that allows analyzing a flow of more than 7 steps, we internally developed Sweden.
But even so, we could not know how much income we lost approximately due to user abandonment in any specific step; this is very important for the CRO methodology as we need to be able to estimate the impact that user abandonment entails in each step. But the data in Sweden wasn’t real time, and I couldn’t make comparisons between different platforms, countries, and conversion funnels. To do that, I had to run queries in BigQuery, and then extract that information into a visualization tool like Tableau or Looker. With the data spread across these different tools, I never had a complete picture of the situation, plus it was very expensive.
With app analytics spread between different tools, you always lack an end-to-end understanding of the user experience.
Carrying on in this way was financially expensive. BigQuery charges for every query, and the value we gained was limited. But it was especially expensive in terms of my time. In order to validate the data returned by a query, it is necessary to perform several checks. Later, if I needed to know the approximate income in dollars that we lost due to the abandonment of a user in a specific step, for example, I had to run a series of long queries and complement them with several transactional tables, which took hours and, in some cases, days.
And if it has taken me a long time, with my experience, then how could this be expected from someone else at Mercado Libre who does not have the technical knowledge or experience? Access to the data has to be simple and we should not all need to have a technical profile to know how to do it. The same goes for the CRO methodology: It is important that any user profile can implement it without the process being very tedious and time-consuming. Everyone in an organization must have the tools to easily access and work with data.
Amplitude brought everything together
For all these reasons, we began to think about moving away from Google Analytics and exploring another tool that would offer what we needed. That’s how we found Amplitude Analytics. In Analytics, we saw a tool that would provide us with a holistic picture of users, bringing together real-time behavioral and transactional data, easily visualized in a single dashboard.
Having run some proofs of concept to test various solutions, Amplitude gave us the most complete view of our data. Features like user segmentation and behavioral cohorts would also give us the depth of analysis we needed, allowing us to identify issues from thousands of variables, such as the user’s browser or operating system.
We began using Amplitude in early 2023, and the impact has been remarkable. We uncovered two unexpected insights that helped us enhance the usability and performance of the checkout process at Mercado Libre.
The first of these had to do with our credit card selector at checkout.
It began with an Analytics alert that told us a number of users closed the app just to reopen it moments later. Looking at the operating system and app version of these user segments, we discovered that users of a particular app version couldn’t input their credit card number on the payment page. The arrow button to go back to the previous page didn’t work either, so users would exit the app and then reopen it to try again.
The issue understandably caused a lot of frustration for users, not to mention lost revenue for Mercado Libre. By identifying and fixing this one issue, our cart abandonment rate at this step dropped from 20% to 11%.
Amplitude focuses our attention on the user experience
Using Amplitude has saved me days of work not double-checking queries. We have all the data we need in one place, in real time, and I have confidence in that data. Features like user segmentation and Pathfinder provide us with unexpected insights that we can act upon.
The proof is in the numbers.Since implementing Amplitude, we have been able to formulate hypotheses that generated an overall checkout conversion rate of 12% and new user conversions of 11%. We were also able to generate another hypothesis that, if implemented, would improve the cart abandonment rate by 20%.
The cumulative effect is expected to result in a 15 % monthly revenue increase.
The proof is in the numbers. Since implementing Amplitude [...] the cumulative effect is expected to result in a 15 % monthly revenue increase.
I’ll be honest: Fostering a data-driven culture takes some change management. Transitioning from Google Analytics to Amplitude is more than just a swap in tools—it’s a paradigm shift in how teams think about and apply analytics to their own work, with a new focus on user behavior. To facilitate this shift, we organized talks and workshops educating teams about digital analytics principles and best practices, helped by teams from Amplitude, and backed this up with a comprehensive wiki for using Amplitude.
Comprehensive education and training are crucial, but the right tool makes all the difference to adoption. Amplitude’s ease of use for many different user profiles–and not just data analysts or technically proficient users—puts data in the hands of everyone at Mercado Libre. That’s how we at Mercado Libre continue to create exceptional experiences.
