Becoming a CROcker

What it is and how we apply Conversion Rate Optimization at Mercado Libre.

Daniela Medina
Published in
8 min readApr 28, 2021


At Mercado Libre, we are in constant search for excellence and improvement of the experience and conversion of our users. For a rockstar, this would equate to hitting the big stages.

How many times have we found ourselves in situations in which we’ve had too much information, a myriad of tools and teams spread across different areas that had the same objective but it was hard to find the synergy to achieve it.

Thus, after much trial-and-error, the Data Analytics Team implemented the CRO (Conversion Rate Optimization) methodology to become rock stars of data, analysis and conversion.

Base image retrieved from

What is CRO?

CRO (Conversion Rate Optimization) is a methodology that allows us to find improvements in user experience and conversion of certain flows. CRO is based on taking qualitative and quantitative data, analyzing and complementing them in order to find areas for improvement.

“The more and the better we measure the initiatives, the better the user experience will be.”

In our Data Analytics Team, this methodology has become a key tool to delve into the main KPIs of each project, such as improving user experience in the purchase flow so that users complete the operation and we don’t lose them along the way.

“We analyze data, hypothesize, discuss improvements and measure results.”

Good Practices

Correct and Complete Tracks

The key is not to “collect data just because”, but to have qualitative and consistent data so as to ensure that what we are measuring answers the questions raised and makes sense with the flow analyzed.

In order to ensure the quality of the tracks and of the qualitative data collected and sent to each analysis tool, our Data Analytics Team works on an initiative that ensures consistency, quality and good tracking practices for each tool used.

Correct choice of the Best Tool for Analysis

Our Data Analytics Team works with several tools related to digital analytics that provide us with qualitative and quantitative data, such as Google Analytics, Looker, BigQuery and HotJar.

In addition, we have tools developed in-house that allow us to further open the spectrum of analysis. It is important to know what answers each one gives us, what for, when and how we can use them.

“We call conversion to the fulfillment of an objective.”

Setting up Funnels

Base image retrieved from

It is important to define the step-by-step process that leads us to the objective we set out to achieve. In our experience, the first thing to do is define the objective of the flow to be analyzed (e.g. successful purchases) and know and recommend the sub-flows that help us and that participate directly in its fulfillment, i.e. search and choose the product, view it, select the shipping and payment methods and review the purchase.

We present this as a composition; for example, to reach the purchase objective, we have to go past those sub-flows mentioned above since they are direct participants.

From each flow that we build step by step, in the Data Analytics Team we analyze each platform and country separately to detect behavior patterns.

For analyzing the flows, defining step-by-step processes and identifying where users come from and go to, we use a tool developed in-house called Sweden*. It was born out of the need to be able to analyze flows of several steps — in Mercado Libre there may be flows of over 15 steps. This is how — given the non-existence in the market of a tool that provides us with this data and is flexible enough to be used by any user profile — Sweden was born. Among its many virtues, it can relate not only behavioral but also transactional data sources, which allows us to know the estimated GMV (Gross Merchandize Value) that we lose to the abandonment, for example, of the purchasing process by the user. It also indicates, say, if a user does not register and how much of the estimated revenue we would lose with this potential buyer.

*In the coming post “Sweden: Joyride Making Funnels”, we will be commenting on what it is, what it is for and what the need was at the business level that led us to develop an in-house tool for flow analysis.

Agile dynamics in A/B Testing

It is paramount to count on a good A/B Testing tool and analyze the data correctly in order to make timely decisions.

At Mercado Libre, we have a tool developed in-house (Melidata), but we also use other open source tools such as Google Optimize to perform A/B Testings with little effort. It is a very intuitive tool that enables users with little experience or those who are not developers to generate really good insights thanks to the proposed A/B Testings.

Meeting of the Minds

We set up CRO squads with different profiles such as Digital Analysts, Developers, Product Owners and UX Designers, among others.

We generate work tables for all of them to contribute with their viewpoints and add value. In these meetings, we brainstorm with each squad and reach agreements; we prioritize tasks, raise hypotheses after analyzing different scenarios, come to conclusions and make the necessary follow-ups to achieve our results.

More Than Words

As important as the communication between the different referents is the way in which we transmit to the sponsors and leaders of each initiative the results and impact generated. For the interlocutor, it is essential to know how to transmit the context of what we are going to analyze, what we observe, what we will do and subsequently the results obtained.

When we talk about “communicating the results”, we have found ourselves in situations where the way in which we tell the story can change the priority of one team or one idea over another. For instance, when we referred only to percentages or absolute values, many times we did not have the same impact as when we supported ourselves and complemented the analyzes with values ​​on a monetary scale. It is not the same to communicate that the change made generated 30% more users who buy on the site, than to say “the change we implemented has generated a monthly increase in gross income of $ 300,000, which translates into a 10% increase in GMV.”

“Knowing what to transmit in storytelling is a fundamental part of the methodology.”

The following example showcases shopping cart abandonment rate in the successive stages of the flow. As we can see, the graph and the explanations tell a story in a simple way, enabling each squad member to relate to it easily. This graph allows us to know the end-to-end of the scenario suggested so each interlocutor can make decisions and come up with hypotheses.

Bonus Track: Debit and Credit Card Registration Improvements

What do we observe?

As a first step, we defined the objective “To make a successful purchase”. Once we had all the tracks complete and well-defined, we created in Sweden the different flows that would participate in the purchase process. Thanks to the alerts received from Sweden, we detected that there was a significant drop in conversion between the payment step and the review step (review of the purchase).

To be even more specific, we defined the flow subflows that generated the greatest dropout, uploaded the funnels in Sweden and again, thanks to its alerts, it was easy for us to detect that the card registration flow (credit and debit) was the one that was generating the greatest friction and bad user experience.

What did we do?

  • We implemented HotJar to be able to get feedback from users on this particular screen.
  • We segmented the traffic and compared the different screen resolutions and their conversion to be able to identify if it was possible to find insights to improve the flow in a specific screen resolution.
  • We segmented users into new versus recurring and by applying these segments to the same card registration flow, we found behavior patterns that were in line with each user’s specific features. This meant improving their experience, especially for new users.
  • Many errors were detected that did not allow users to continue with the loading of the cards and that generated friction; hence, they would leave the flow midway.
  • Different proposals were designed for the same card registration flow, which were analyzed using the A/B Testing tool developed in-house.
  • The double flow for debit and credit card registration was eliminated and a single registration form was developed for both types of cards, since when users entered a flow and the system detected that it was the wrong type of card, they were not allowed to continue loading.

We put together a matrix of priorities to classify the improvements. This type of matrix is ​​a very useful tool that not only optimizes time management but also serves as a decisional matrix; that is, in our case it has helped us make decisions since it allowed us to prioritize the points found. With this matrix, we were able to classify problems, their impact, and the effort it would take to resolve them.

The priority matrix that we share below is based on two axes: Ease of improvement implementation and Impact:

What did we achieve?

  • Solution of errors in the card registration forms, which allowed users to continue the flow without problems, minimizing form errors.
  • Narrowing of the gap that existed in the conversion between different screen resolutions.
  • Improvements in UX with the card registration form, making the flow more user-friendly, especially for newcomers.


As the song by Poison “Every Rose Has its Thorn’’ goes, CRO experience at Mercado Libre is not and will not be the exception. Many times we were not in a position to implement CRO in our projects because we did not have well-defined tracks or appropriate tools. There were also cases in which CRO was implemented but the results were not conclusive to be able to decide for one proposal over another. It is here where we highlight the importance of the bases (tracks) and the resilience of the squad to avoid getting frustrated.

There is no perfect CRO school or a place of great inspiration where everything works as described in Strawberry Fields Forever (The Beatles), but you can start working with this methodology, quality data, the best storytelling and complement them with different tools, teamwork and out of the box thinking.

Once you have all these points you can grab the opportunity … for a Conversion Rate Optimization and Let It Be.

If you liked this story please clap 👏 and start following us for more tech posts from MELI!




Daniela Medina

Technical Leader, Data & Analytics @ Mercado Libre