<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Martín Di Luzio on Medium]]></title>
        <description><![CDATA[Stories by Martín Di Luzio on Medium]]></description>
        <link>https://medium.com/@martin.diluzio?source=rss-af0cdc09ea4c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*2_ouzyp43XO9VLG3_-VZZg.png</url>
            <title>Stories by Martín Di Luzio on Medium</title>
            <link>https://medium.com/@martin.diluzio?source=rss-af0cdc09ea4c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 18:45:56 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@martin.diluzio/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[How we are using AI in Mercado Libre’s accessibility team]]></title>
            <link>https://medium.com/mercadolibre-tech/how-we-are-using-ai-in-mercado-libres-accessibility-team-e960b83283a9?source=rss-af0cdc09ea4c------2</link>
            <guid isPermaLink="false">https://medium.com/p/e960b83283a9</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[accessibility]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Martín Di Luzio]]></dc:creator>
            <pubDate>Wed, 08 Oct 2025 12:58:21 GMT</pubDate>
            <atom:updated>2025-10-13T13:35:22.802Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AajxRG-d0zEsO18YYvts0A.png" /></figure><p>AI isn’t here to replace accessibility work — it’s here to make it stronger.</p><p>Today, our accessibility team faces a challenge: supporting hundreds of designers and developers with questions, reviews, and continuous improvements. To scale this impact, we’re exploring how artificial intelligence (AI) can help us automate tasks, provide immediate answers, and foster collective learning.</p><p>That’s why we’ve launched several initiatives to speed up work, increase independence, and strengthen execution capacity.</p><p>Note: In many of these initiatives, we rely on automatization tools available in our <a href="https://medium.com/mercadolibre-tech/subpage/fury">internal development ecosystem, Fury</a>.</p><h3>The A11Y assistant for everyday work</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iVq3f4CGAPiO1iNUrFyv1A.jpeg" /><figcaption><em>Image 1: Automation flow that integrates different nodes, from the user’s question to obtaining relevant information, processing it, and providing an answer.</em></figcaption></figure><h3>How does it work?</h3><p>This assistant:</p><ul><li>Activates when mentioned in the support channel.</li><li>Processes both messages and screen images.</li><li>Consults internal documentation, training materials, historical queries, previously reported accessibility tickets, and our design system.</li><li>Uses a large language model (LLM) with <strong>Retrieval-Augmented Generation (RAG)</strong> to deliver reliable answers.</li></ul><p>The result: contextualized responses grounded in internal resources and standards. From the query initiation to gathering additional context, and finally to delivering a response backed by verified information, the entire pipeline is crucial. This approach helps avoid “hallucinations” (invented or incorrect answers) and ensures the A11Y assistant stays focused on trusted resources. Thus, it provides solutions and resources that are not only accurate but also directly applicable and aligned with Mercado Libre’s internal accessibility standards and tools.</p><h3>Understanding problems better to deliver better solutions</h3><p>One of our biggest learnings is that teams need information that’s clear and easy to understand — not just technical reports.</p><p><strong>In manual accessibility audits</strong>, we provide technical details such as assistive technologies used, reproduction steps, and affected WCAG (Web Content Accessibility Guidelines) criteria. While useful, this level of detail can be overwhelming for those not deeply familiar with accessibility.</p><p>That’s where AI comes in. For each generated ticket, AI automatically adds a contextual note covering three key aspects:</p><ol><li><strong>Technical impact explained in user terms</strong>: A short, clear description of the problem — not just “this is wrong,” but why it’s wrong from a user-experience perspective.</li><li><strong>Suggestions or recommendations</strong>: Concrete guidance, such as code examples or references, always framed within our knowledge base and design patterns.</li><li><strong>How to verify the fix</strong>: Accessibility, like any bug, must be tested. To flatten the learning curve, AI provides the tools and exact steps to validate the correction.</li></ol><p><strong>Real example</strong>: In a search ticket, the “Skip to results” link was missing. AI explained the impact on keyboard users, suggested the correct code based on our guidelines, and listed the steps needed to verify the change.</p><p><strong>For automated tools like Axe</strong>, which detect accessibility errors via rules, we enrich the detected issues with AI-generated context. This adds practical details that help developers better understand the problem in the HTML where it appears.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/831/1*kh5aYUJLGRx7L6J7Vm1c3w.jpeg" /><figcaption><em>Image 2: </em>Screen showing an explanation of the axe aria-progressbar-name rule error and suggesting possible solutions with additional context for the developer.</figcaption></figure><p>Each additional piece of context we provide significantly scales understanding without requiring constant involvement from the accessibility team (though we’re always available when needed). At the same time, this empowers development teams to become more autonomous in solving accessibility issues.</p><h3>The accessibility notes assistant for handoffs</h3><p>To bring accessibility into the UI design phase, we created an assistant that helps UX teams identify and add accessibility specifications in their handoffs.</p><p>Starting from a screen image (desktop web or native mobile), the assistant analyzes the visual context and platform type. It then generates a <strong>descriptive visual map</strong>, detailing the elements on the screen and recommending annotations based on our accessibility notes for Figma.</p><p>The assistant’s prompt focuses on explaining, in simple and clear terms, how to annotate components like headers, links, buttons, groups, or dynamic areas. The emphasis is on semantic specifications rather than technical implementation.</p><p>This tool resolves common doubts and helps teams incorporate accessibility independently, ensuring more complete and inclusive design deliveries. For now, responses are text-only since they’re far more accurate than image-based annotations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kU2Yk1GH4WR1IrDdIqh8RA.jpeg" /><figcaption><em>Image 3: The GPT assistant indicates which tags to use to identify relevant semantic elements on the attached screen: headings, buttons, and components from our library.</em></figcaption></figure><h3>Assisted reviews of resolved issues</h3><p>Our team conducts weekly reviews of resolved accessibility tickets to validate fixes and learn from applied solutions. To scale this process and reduce manual review time, we developed an automation that integrates <strong>Jira, GitHub, and an AI agent</strong>.</p><h3>What does this flow do?</h3><ul><li>Pulls all accessibility tickets resolved that day.</li><li>Breaks down each ticket into key information.</li><li>Instructs the AI agent to:</li><li>Analyze ticket comments and solution evidence.</li><li>Identify linked GitHub pull requests (PRs) and review their technical content.</li><li>Assess the clarity, relevance, and documentation of the solution.</li><li>Classifies the ticket with a traffic-light emoji (🟢 green, 🟡 yellow, 🔴 red) based on fix quality.</li><li>Stores the results in a shared spreadsheet.</li></ul><h3>Why is this useful?</h3><p>This workflow filters which tickets need additional manual review, allowing us to focus our team’s time. With an agent aware of both ticket context and code details, we can:</p><ul><li>Automatically flag tickets lacking sufficient technical evidence.</li><li>Objectively classify fix reliability.</li><li>Request targeted information from teams when needed.</li><li>Centralize the analysis in a shared dashboard.</li></ul><h3>Next steps: shifting accessibility left</h3><p>Our goal is to bring accessibility further <strong>left</strong> into the early stages of the product development lifecycle. This means preventing issues before they arise, integrating smoothly into design and development workflows, and leveraging emerging paradigms like <strong>Vibe Coding</strong>.</p><p>Here, <strong>Fury</strong>, our internal development platform, plays a key role in scaling, securing, and centralizing these AI agents.</p><h3>Conclusion</h3><p>These initiatives show that AI doesn’t replace accessibility experts — it multiplies them. By automating responses, enriching tickets, and supporting teams, we gain time to focus on the next strategic challenges.</p><p>As accessibility professionals, we must actively engage with AI initiatives by testing, experimenting, discovering, and encouraging its responsible use so it benefits everyone.</p><p>A special shout-out to <strong>Rocío Ruiz, Julieta Natali Navazo, Manuela Sánchez Marino, and Guilherme Zangarini</strong>, among others, who are leading the way in AI and accessibility explorations at Mercado Libre.</p><p>If you work in tech, we invite you to experiment with AI and accessibility in your processes. Share your learnings, and let’s build a more inclusive community together.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e960b83283a9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/mercadolibre-tech/how-we-are-using-ai-in-mercado-libres-accessibility-team-e960b83283a9">How we are using AI in Mercado Libre’s accessibility team</a> was originally published in <a href="https://medium.com/mercadolibre-tech">Tecnología de Mercado Libre</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Lessons learned when developing accessibility in prices at Mercado Libre]]></title>
            <link>https://medium.com/mercadolibre-tech/lessons-learned-when-developing-accessibility-in-prices-at-mercado-libre-59fd62fe85cb?source=rss-af0cdc09ea4c------2</link>
            <guid isPermaLink="false">https://medium.com/p/59fd62fe85cb</guid>
            <category><![CDATA[accessibility]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[accesibilidad]]></category>
            <category><![CDATA[mercadolibre]]></category>
            <category><![CDATA[a11y]]></category>
            <dc:creator><![CDATA[Martín Di Luzio]]></dc:creator>
            <pubDate>Thu, 19 May 2022 18:58:11 GMT</pubDate>
            <atom:updated>2022-11-16T19:10:08.796Z</atom:updated>
            <content:encoded><![CDATA[<p>Today, May 19, it is “Global Accessibility Awareness Day” (GAAD), so we think it is a good moment to share the lessons we have learned to improve user accessibility to the products we develop at Mercado Libre.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YYXHmsLcnls6rLiLXFHMKg.jpeg" /></figure><p>Product prices can be reasonable, that is to say affordable, but they must be well communicated. Immersed in a “user centricity” culture, we seek to improve the browsing experience of our site’s users. That is why we have taken on the challenge of making prices “accessible” to screen readers, assistive technology that represents text and image content like voice output or Braille. <a href="https://medium.com/mercadolibre-tech/what-we-have-learned-from-working-on-digital-accessibility-954f275fdff3">At Mercado Libre we started a process to improve accessibility</a> to our products, understanding that many of the obstacles and barriers come from the way in which software developers think, design and build products and services.</p><p>A cross-cutting issue in Mercado Libre ‘s user experiences that has remained current over time is Price. <strong>We’ve noted that </strong>in its varied available versions and implementations,<strong> it was being displayed to screen reader users in a poorly understandable way and with inaccurate values.</strong></p><p>In order to generate interfaces that can be understandable, perceivable, operable and robust, we’ve started implementing accessibility improvements in our Andes Design System. In this article, we put the focus on a specific component for the web: the Price component.</p><h3>What were our first steps?</h3><p>The main idea was to develop a Price component, AKA MoneyAmount, that teams could adopt, with its different variants springing from a unique and consistent approach in Mercado Libre. This component added great learning about assistive technologies, browsers, implementations and concrete experiences.</p><h3>The challenges we faced</h3><p>This was the scenario when we started our journey:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FHy4jNfgOyDs%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DHy4jNfgOyDs&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FHy4jNfgOyDs%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c70d500007e2abc2feae1fea1a66fc3d/href">https://medium.com/media/c70d500007e2abc2feae1fea1a66fc3d/href</a></iframe><p>Verbalization through screen readers depends on several factors:</p><ul><li>the browser (user agent) and its way of displaying <a href="https://web.dev/the-accessibility-tree/">the accessibility tree</a>,</li><li>the operating system,</li><li>the screen reader and its settings,</li><li>the voice used (synthesizer). The same screen reader, on the same system, operating with the same browser but with different voices, can read the same text in very different ways,</li><li>the syntax and formatting of the code.</li></ul><p>We’ve specifically found the following challenges:</p><h4>Focus jumps in screen reader browse mode</h4><p>When the user scrolls through the price region, there are different jumps in the sequential navigation that display currency, values and characters separately, which that makes it a little unintuitive and inconsistent.</p><h4>Currency “$”</h4><p>The currency symbol “$” is verbalized/read as “dollars” by the screen reader, regardless of the language or region. This generates inconsistency with the country. The screen reader’s pronunciation dictionary could replace this behavior but we think the solution should not be left to the user.</p><h4>Integer and decimal separator character</h4><p>Each country or region formats figures in different ways using separator characters for integers (with a thousand separator) and decimals to identify the relationship between digits of the same price. For example, in Mexico, the thousands are separated with a comma, while in Argentina they are separated with a period.</p><p>Reference: <a href="https://en.wikipedia.org/wiki/Decimal_separator">Decimal separator — Wikipedia</a></p><h4>Previous prices and discounts without context</h4><p>Generating context for the user to recognize the price variation is essential in decision making and this was not being communicated at all.</p><figure><img alt="Price crossed out to identify a previous value of the price." src="https://cdn-images-1.medium.com/max/596/1*qYHCYvfeTKUCtdlEkGEACA.png" /><figcaption>Information should not be conveyed only visually; a crossed-out value does not convey its purpose to screen readers.</figcaption></figure><h3>So what did we do?</h3><p>After an exhaustive analysis of the situation, we started with a validation stage based on 3 pillars: specifications and standards, validation of references in the field, and tests with screen reader users.</p><h4>Specifications</h4><p>In the official HTML and accessibility documentation provided by WHATWG and W3C WAI-ARIA, we obtained information and authoring practices for different components. Even if we knew that these recipes are not perfect and do not always offer the best experiences to all users, we capitalized on them as a starting point for the analysis.</p><p>They are necessary reference sources that allow us to validate the quality of our code and, although we know that the specification support is not always stable, we wanted to be accurate in our implementation.</p><ul><li><a href="https://html.spec.whatwg.org/">HTML Specifications</a></li><li><a href="https://www.w3.org/TR/wai-aria/">WAI ARIA</a></li><li><a href="https://www.w3.org/WAI/ARIA/apg/">ARIA authoring practices</a></li><li><a href="https://www.w3.org/TR/speech-synthesis11/">TTS: Speech Synthesis Markup Language (SSML)</a></li></ul><h4>Technical validations</h4><p>We wanted to achieve a robust solution in verbalization, although we knew there was no general solution. Yet, it was necessary to avoid expressing the wrong price to screen reader users. Thus, we first ran tests with separator characters and observed unstable results depending on combinations of speech engines (TTS) or punctuation settings in screen readers.</p><figure><img alt="Table with some tests carried out combining screen readers (JAWS, NVDA , Voice Over and Talkback) and separators (space, comma, period and no separator), indicating the support they provide in their pronunciation. It is observed that Talkback is the one that has full support for any combination of separators, and if a separator is not used, we get the most robust result (more consistent pronunciation between the screen readers used)." src="https://cdn-images-1.medium.com/max/1024/1*9EFViw3gwYxjtRqQ5aBg1Q.png" /><figcaption><em>Table of tests performed with different screen readers to validate some cases.</em></figcaption></figure><p><strong>References</strong>:</p><ul><li>* verbalizes only digits to the left of the cent separator, ignoring the remaining digits</li><li>“-” verbalizes digit by digit</li><li>“ok” verbalizes hundreds, thousands, millions</li><li>“W/S” without separator</li></ul><p>Then, we analyzed documentation from different accessibility referents and documentation of solutions applied by teams outside Mercado Libre, but we did not find specific implementations for our case.</p><p>Documents from accessibility teams</p><ul><li><a href="https://a11y-guidelines.orange.com/en/">Orange</a></li><li><a href="https://www.bbc.co.uk/accessibility/forproducts/">BBC</a></li><li><a href="https://design-system.service.gov.uk/accessibility/">UK Gov</a></li></ul><p>Some references:</p><ul><li><a href="https://www.scottohara.me/">Scott O’Hara</a></li><li><a href="https://adrianroselli.com/">Adrian Roselli</a></li><li><a href="https://moiety.me/">Zoë Bijl</a></li><li><a href="https://sarahmhigley.com/">Sarah Higley</a></li></ul><p><strong>It is important to note that browsers do not always provide the expected support for screen readers, even after the implementation of the accessibility specifications generated by </strong><a href="https://www.w3.org/WAI/standards-guidelines/aria/"><strong>WAI-ARIA</strong></a><strong>. For this reason, it is important to validate with </strong><a href="https://a11ysupport.io/"><strong>Accessibility Support</strong></a><strong> to find a widely supported solution.</strong></p><h4>User experience</h4><p>Last but not least, the implementations we generated needed to be validated by users in order to identify if the experience was being effective. To this end, we relied on organizations such as <a href="https://nahualit.com/">Nahual IT</a> and <a href="https://www.itgrarte.org/">Fundación itgrarte</a> to help us recognize key points.</p><h3>How did we implement it?</h3><p>The first thing we did to solve the focus jumps in browser mode, currency “$”, and integer and decimal separator characters was to remove the visual price representation from the browsing for screen reader users. In order to do so, we used the aria-hidden attribute on the container element. This ARIA attribute allows us to hide an element from the accessibility tree used by assistive technologies so that they ignore it.</p><p>Observations: The price is made up of different classes that allow you to apply style to your visual presentation.</p><pre>&lt;div class=&quot;&quot; <strong>aria-hidden=&quot;true&quot;</strong>&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>It is important to clarify that this attribute should not be used in keyboard focusable elements or with tabindex=”0&quot; since it generates inconsistency between the focus with TAB key and the scanning mode of the screen readers that will ignore the element.</p><p>After analyzing the challenges we had in finding a robust solution as well as the tests performed, we decided to verbalize the price in Spanish with the following format in pesos and cents:</p><p><strong>{integer} pesos with {decimals} cents</strong></p><p>Where {integer} is the number without thousands. For example: 10000 and {decimals} is the number of cents.</p><p>To achieve the specific verbalization to the screen reader, we use a CSS Screen Reader Only class that allows us to visually hide the expression, but that is achievable in the navigation of these technologies.</p><pre>&lt;style&gt;<br>  .sr-only {<br>  border: 0;<br>  clip: rect(0 0 0 0);<br>  clip-path: inset(50%);<br>  height: 1;<br>  margin: 0 -1px -1px 0;<br>  overflow: hidden;<br>  padding: 0;<br>  position: absolute;<br>  width: 1;<br>}<br>&lt;/style&gt;</pre><p>Our code block will finally look like this:</p><pre>&lt;div <strong>class=&quot;sr-only&quot;</strong>&gt;<strong>22999 pesos con 08 centavos.</strong>&lt;/div&gt;<br>&lt;div class=&quot;&quot; aria-hidden=&quot;true&quot;&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p><strong>Disclaimer</strong>: Screen readers do not generally require or need the addition of verbosity to textual content, but they do for our implementation.</p><p>In some of the tests evaluated, the aria-label attribute was intended to be applied directly to assign a programmatic name to the price container:</p><pre>&lt;div class=&quot;&quot; <strong>aria-label=&quot;22999 pesos con 08 centavos.&quot;</strong>&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>This is also read only by screen readers, but the ARIA specification indicated that it should only be applied to elements with implicit or explicit roles, thus disregarding this implementation.</p><p>As regards the challenge of “Previous prices and discounts without context”, in order to solve the statement of a current price and a previous one, we applied the same approach to provide context in price differentiation by adding the text “Before” and then the text “Now” to the current price.</p><p>Since the native HTML &lt;s&gt; tag (visually strikethrough text) does not announce any semantics to screen readers, at least in the tests performed and in the validated documentation, we generated this code block:</p><pre>&lt;div <strong>class=&quot;sr-only&quot;&gt;Antes: </strong>24999<strong> </strong>pesos.<strong>&lt;/div&gt;<br></strong>&lt;s class=”” aria-hidden=”true”&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;24.999&lt;/span&gt;<br>&lt;/s&gt;</pre><pre>&lt;div <strong>class=&quot;sr-only&quot;</strong>&gt;<strong>Ahora:</strong> 22999 pesos con 08 centavos.&lt;/div&gt;<br>&lt;div class=&quot;&quot; aria-hidden=&quot;true&quot;&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>This is the final result of all this process:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FwE2sMOGIJb4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DwE2sMOGIJb4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FwE2sMOGIJb4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/05928fd5fe88fa56887dab882e184f60/href">https://medium.com/media/05928fd5fe88fa56887dab882e184f60/href</a></iframe><p>With this implementation we have achieved <strong>accurate price pronunciation, prioritizing the sensitivity of the data</strong> and not a solution for a particular screen reader.</p><h3>Learnings</h3><p>Context is essential to perceive information and to be able to generate coherent experiences for all users. We understand that it is important to prevent programmatic information from differing from visual information and in order to do so, we’ve validated expected behaviors with the experience of users of assistive technologies.</p><p>We’ve also discovered that specifications differ between browsers and screen readers, as well as that support and adoption of new or improved standards do not always go hand in hand, and that browsers compensate for poor implementations by forcing unsound decisions, that is, with solutions that work well for some but not for others.</p><p>Further, authoring practices, approaches or recipes do not always generate good experiences and it is important not only to evaluate consistency within our products, but also in the most widely used products on the market.</p><p>Throughout this process, we have generated knowledge and experiences that we want to share with the community. We have also been the recipients of good practices generated around other products. In this way, we were able to work together and in synergy so as to put an end to this paradox: high or low prices, but accessible.</p><p>Lastly, I want you to also learn <a href="https://medium.com/mercadolibre-tech/accessibility-in-prices-our-learning-from-the-android-perspective-7032e81e11d7">everything we learned on the Android platform while we improve the accessibility of prices</a> from the hand of <a href="https://medium.com/@juan.unzurrunzaga">Juan Ignacio Unzurrunzaga</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59fd62fe85cb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/mercadolibre-tech/lessons-learned-when-developing-accessibility-in-prices-at-mercado-libre-59fd62fe85cb">Lessons learned when developing accessibility in prices at Mercado Libre</a> was originally published in <a href="https://medium.com/mercadolibre-tech">Tecnología de Mercado Libre</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Aprendizajes al desarrollar precios accesibles en Mercado Libre]]></title>
            <link>https://medium.com/mercadolibre-tech/aprendizajes-al-desarrollar-precios-accesibles-en-mercado-libre-c74ad7c62b0d?source=rss-af0cdc09ea4c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c74ad7c62b0d</guid>
            <category><![CDATA[a11y]]></category>
            <category><![CDATA[accesibilidad]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[accessibility]]></category>
            <category><![CDATA[mercadolibre]]></category>
            <dc:creator><![CDATA[Martín Di Luzio]]></dc:creator>
            <pubDate>Thu, 19 May 2022 18:57:34 GMT</pubDate>
            <atom:updated>2023-02-28T13:41:32.336Z</atom:updated>
            <content:encoded><![CDATA[<p>El día de hoy, 19 de mayo, es el “Global Accessibility Awareness Day” (GAAD) por eso nos parece oportuno compartir los aprendizajes que tomamos para mejorar la accesibilidad en los productos que desarrollamos en Mercado Libre.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YYXHmsLcnls6rLiLXFHMKg.jpeg" /></figure><p>Los precios de los productos pueden ser accesibles, es decir asequibles, pero tienen que transmitirse correctamente. Inmersos en una cultura “user centricity”, buscamos mejorar la experiencia de navegación de las personas usuarias de nuestro sitio. Es por ello que nos embarcamos en el desafío de hacer los precios “accesibles” a lectores de pantalla, tecnología asistiva que representa el contenido de texto e imagen como salida de voz o braille. <a href="https://medium.com/mercadolibre-tech/qu%C3%A9-aprendimos-al-trabajar-en-accesibilidad-digital-d2c9ecf8af37">En Mercado Libre iniciamos un recorrido para mejorar la accesibilidad</a> de los usuarios en nuestros productos, entendiendo que muchos de los obstáculos y barreras surgen a partir de la manera en que los desarrolladores de software pensamos, diseñamos y construimos los productos y servicios.</p><p>Un elemento transversal en las experiencias del usuario de Mercado Libre que mantiene su vigencia a lo largo del tiempo es el Precio<strong>,</strong> y en sus variadas versiones disponibles e implementaciones,<strong> detectamos que se anunciaba de manera poco comprensible y con valores inexactos a personas usuarias de lectores de pantalla.</strong></p><p>Para generar interfaces que puedan ser comprensibles, perceptibles, operables y robustas, comenzamos a implementar mejoras de accesibilidad en nuestro Design System Andes, centrándonos en esta nota en un componente específico para web.</p><h3>¿Cuáles fueron nuestros primeros pasos?</h3><p>La idea principal fue entonces el desarrollo de un componente Precio, AKA MoneyAmount, que los equipos puedan adoptar, con sus diferentes variantes desde un enfoque único y consistente en Mercado Libre, que sumó grandes aprendizajes sobre tecnologías asistivas, navegadores, implementaciones y experiencias concretas.</p><h3>Desafíos encontrados</h3><p>Con este escenario iniciamos nuestro recorrido:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FHy4jNfgOyDs%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DHy4jNfgOyDs&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FHy4jNfgOyDs%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c70d500007e2abc2feae1fea1a66fc3d/href">https://medium.com/media/c70d500007e2abc2feae1fea1a66fc3d/href</a></iframe><p>La verbalización través de los lectores de pantalla depende de varios factores:</p><ul><li>el navegador (agente de usuario) y su forma de exponer el <a href="https://developers.google.com/web/fundamentals/accessibility/semantics-builtin/the-accessibility-tree?hl=es">árbol de accesibilidad</a>,</li><li>el sistema operativo,</li><li>el lector de pantalla y su configuración,</li><li>la voz utilizada (sintetizador). El mismo lector de pantalla, en el mismo sistema, operativo con el mismo navegador pero con diferentes voces puede leer el mismo texto de manera muy diferente,</li><li>la sintaxis y formateo del código.</li></ul><p>Encontramos concretamente los siguientes desafíos:</p><h4>Saltos del foco en modo exploración de lectores de pantalla</h4><p>Cuando el usuario recorre la región de precios, se producen diferentes saltos en la navegación secuencial, anunciando currency, valores y caracteres por separado, lo que resulta poco intuitivo e inconsistente.</p><h4>Currency “$”</h4><p>El símbolo currency “$” es verbalizado como dólares por el lector de pantalla, independientemente del idioma/región, generando inconsistencia con el país. El diccionario de pronunciación del lector de pantalla podría sustituir este comportamiento, pero no creemos que la solución deba depender del usuario.</p><h4>Caracter separador enteros y decimales</h4><p>Cada país o región da formato a las cifras de diferentes maneras con uso de caracteres separadores para enteros (con separador de millares) y decimales para así identificar la relación entre los dígitos de un mismo precio, por ejemplo, en México se separan los millares con coma, mientras que en Argentina se separan con punto.</p><p>Referencia: <a href="https://es.wikipedia.org/wiki/Separador_de_millares">Separador de millares — Wikipedia, la enciclopedia libre</a></p><h4>Precios anteriores y descuentos sin contexto</h4><p>Generar contexto al usuario para reconocer la variación de un precio es fundamental en la toma de decisiones y esto no se anunciaba en absoluto.</p><figure><img alt="Precio tachado para identificar un valor anterior del precio." src="https://cdn-images-1.medium.com/max/596/1*VrSC57kayZb9Kg1P_oiK8w.png" /><figcaption>La información no debe transmitirse solamente de forma visual, y un valor tachado no transmite su propósito a los lectores de pantalla.</figcaption></figure><h3>¿Qué hicimos entonces?</h3><p>Luego de hacer un análisis exhaustivo de la situación, comenzamos con una etapa de validación, para ello nos basamos en 3 pilares: especificaciones y estándares, validaciones de referentes en la temática y comprobaciones con personas usuarias de lectores de pantalla.</p><h4>Especificaciones</h4><p>En la documentación oficial de HTML y accesibilidad brindada por WHATWG y W3C WAI-ARIA obtuvimos información y prácticas de autores para diferentes componentes. Incluso sabiendo que estas recetas no son perfectas y no dan siempre las mejores experiencias a todas las personas usuarias, las capitalizamos como puntapié inicial del análisis.</p><p>Son fuentes de referencia necesaria que permiten validar la calidad de nuestro código y aunque sabiendo que el soporte de la especificación no es siempre estable, buscábamos ser precisos en nuestra implementación.</p><ul><li><a href="https://html.spec.whatwg.org/">Especificación HTML</a></li><li><a href="https://www.w3.org/TR/wai-aria/">WAI ARIA</a></li><li><a href="https://www.w3.org/WAI/ARIA/apg/">Prácticas de autor ARIA</a></li><li><a href="https://www.w3.org/TR/speech-synthesis11/">TTS: Speech Synthesis Markup Language (SSML)</a></li></ul><h4>Validaciones técnicas</h4><p>Queríamos lograr una solución robusta en la verbalización sabiendo que no había una solución general y era necesario no expresar un precio equivocado a personas usuarias de lectores de pantalla.</p><p>Primero hicimos pruebas con caracteres separadores y observamos resultados inestables según combinaciones de motores de voz (TTS) o configuración de puntuación en lectores de pantalla.</p><figure><img alt="Tabla con algunas pruebas realizadas combinando lectores de pantalla (JAWS, NVDA , Voice Over y Talkback) y separadores (espacio, coma, punto y sin separador), indicando el soporte que brindan en su pronunciación. Se observa que Talback es el que tiene soporte completo para cualquier combinación de separador y si no se utiliza separador tenemos el resultado más robusto (pronunciación más consistente entre los lectores de pantalla utilizados)." src="https://cdn-images-1.medium.com/max/1024/0*v54npaO8yh1ffBc5" /><figcaption><em>Tabla con algunas pruebas realizadas combinando lectores de pantalla y separadores.</em></figcaption></figure><p><strong>Referencias</strong>:</p><ul><li>* verbaliza sólo cifras a la izquierda del separador de centavos, ignorando las restantes</li><li>“-” verbaliza cifra por cifra</li><li>“ok” verbaliza cientos, miles, millones</li><li>“S/s” sin separador</li></ul><p>Luego, analizamos documentación de diferentes referentes en accesibilidad y documentación de soluciones aplicadas por equipos fuera de Mercado Libre, pero no encontramos implementaciones específicas para nuestro caso.</p><p>Documentos de equipos de accesibilidad:</p><ul><li><a href="https://a11y-guidelines.orange.com/en/">Orange</a></li><li><a href="https://www.bbc.co.uk/accessibility/forproducts/">BBC</a></li><li><a href="https://design-system.service.gov.uk/accessibility/">UK Gov</a></li></ul><p>Personas referentes:</p><ul><li><a href="https://www.scottohara.me/">Scott O’Hara</a></li><li><a href="https://adrianroselli.com/">Adrian Roselli</a></li><li><a href="https://moiety.me/">Zoë Bijl</a></li><li><a href="https://sarahmhigley.com/">Sarah Higley</a></li></ul><p><strong>Cabe destacar que los navegadores no dan siempre el soporte esperado a lectores de pantalla, aún siendo implementadas las especificaciones de accesibilidad generadas por </strong><a href="https://www.w3.org/WAI/standards-guidelines/aria/"><strong>WAI-ARIA</strong></a><strong>, por eso es importante validar con </strong><a href="https://a11ysupport.io/"><strong>Accessibility Support</strong></a><strong> para encontrar una solución soportada en gran medida.</strong></p><h4>Experiencia de personas usuarias</h4><p>Por último y no menos importante, las implementaciones que generamos necesitaban ser validadas por personas usuarias para poder identificar si la experiencia estaba siendo eficiente. Para ello nos apoyamos en organizaciones como <a href="https://nahualit.com/">Nahual IT</a> e <a href="https://www.itgrarte.org/">itgrarte fundación</a> que nos ayudaron a reconocer puntos clave.</p><h3>¿Cómo lo implementamos?</h3><p>Para dar solución a los saltos del foco en modo exploración, currency “$” y caracter separador enteros y decimales lo primero que hicimos fue quitar de la navegación para usuarios de lectores de pantalla la representación del precio visual. Para ello utilizamos el atributo aria-hidden en el elemento contenedor. Este atributo ARIA nos permite ocultar un elemento del árbol de accesibilidad utilizado por las tecnologías asistivas para que lo ignoren.</p><p>Observaciones: El precio está compuesto por diferentes clases que permiten aplicar estilo a su presentación visual.</p><pre>&lt;div class=&quot;&quot; <strong>aria-hidden=&quot;true&quot;</strong>&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>Es importante aclarar que este atributo no debe ser utilizado en elementos enfocables por teclado o con tabindex=”0” dado que genera inconsistencia entre el foco con tecla TAB y el modo exploración de los lectores de pantalla que ignoraran al elemento.</p><p>Analizando los desafíos que tuvimos a la hora de encontrar una solución robusta y las pruebas realizadas, decidimos verbalizar el precio con el siguiente formato:</p><p><strong>{entero} pesos con {decimales} centavos.</strong></p><p>Siendo {entero} el número sin separador de millares, ejemplo: 10000 y {decimales} la cantidad de centavos.</p><p>Para lograr la verbalización específica al lector de pantalla, utilizamos una clase CSS Screen Reader Only que nos permite ocultar visualmente la expresión pero que es alcanzable en la navegación de estas tecnologías.</p><pre>&lt;style&gt;<br>  .sr-only {<br>  border: 0;<br>  clip: rect(0 0 0 0);<br>  clip-path: inset(50%);<br>  height: 1;<br>  margin: 0 -1px -1px 0;<br>  overflow: hidden;<br>  padding: 0;<br>  position: absolute;<br>  width: 1;<br>}<br>&lt;/style&gt;</pre><p>Quedando nuestro bloque de código finalmente así:</p><pre>&lt;div <strong>class=&quot;sr-only&quot;</strong>&gt;<strong>22999 pesos con 08 centavos.</strong>&lt;/div&gt;<br>&lt;div class=&quot;&quot; aria-hidden=&quot;true&quot;&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p><strong>Disclaimer</strong>: Los lectores de pantalla no requieren ni necesitan, generalmente, el agregado de verbosidad al contenido textual, pero sí para nuestra implementación.</p><p>En algunas pruebas evaluadas se pensaba aplicar directamente el atributo aria-label para asignar un nombre programático al contenedor del precio:</p><pre>&lt;div class=&quot;&quot; <strong>aria-label=&quot;22999 pesos con 08 centavos.&quot;</strong>&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>Lo cual también es leído únicamente por lectores de pantalla pero la especificación ARIA nos indicaba que sólo se debe aplicar a elementos con roles implícitos o explícitos, desestimando esa implementación.</p><p>Con respecto al desafío de “Precios anteriores y descuentos sin contexto”, a fin de dar solución al enunciado de un precio actual y uno anterior, aplicamos el mismo enfoque para dar contexto en la diferenciación de precios agregando el texto <em>“Antes”</em> y a continuación el texto <em>“Ahora”</em> al precio actual.</p><p>Dado que la etiqueta nativa HTML &lt;s&gt; (texto tachado visualmente) no anuncia ninguna semántica a los lectores de pantalla, al menos en las pruebas realizadas y en la documentación validada, generamos este bloque de código:</p><pre>&lt;div <strong>class=&quot;sr-only&quot;&gt;Antes: </strong>24999<strong> </strong>pesos.<strong>&lt;/div&gt;<br></strong>&lt;s class=”” aria-hidden=”true”&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;24.999&lt;/span&gt;<br>&lt;/s&gt;</pre><pre>&lt;div <strong>class=&quot;sr-only&quot;</strong>&gt;<strong>Ahora:</strong> 22999 pesos con 08 centavos.&lt;/div&gt;<br>&lt;div class=&quot;&quot; aria-hidden=&quot;true&quot;&gt;<br>  &lt;span class=&quot;currency&quot;&gt;$&lt;/span&gt;<br>  &lt;span class=&quot;fraction&quot;&gt;22.999&lt;/span&gt;<br>  &lt;span class=&quot;separator&quot;&gt;,&lt;/span&gt;<br>  &lt;span class=&quot;cents&quot;&gt;08&lt;/span&gt;<br>&lt;/div&gt;</pre><p>Éste es el resultado final de todo este recorrido:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FwE2sMOGIJb4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DwE2sMOGIJb4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FwE2sMOGIJb4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/05928fd5fe88fa56887dab882e184f60/href">https://medium.com/media/05928fd5fe88fa56887dab882e184f60/href</a></iframe><p>Con esta implementación logramos una <strong>precisa pronunciación de precios, priorizando la sensibilidad del dato </strong>y no una solución para determinado lector de pantalla.</p><h3>Aprendizajes</h3><p>El contexto es fundamental para percibir la información y poder generar experiencias coherentes para todo el conjunto de usuarios. Entendemos que es importante prevenir que la información programática difiera de la información visual y para ello validamos comportamientos esperados con la experiencia de personas usuarias de tecnologías asistivas.</p><p>Descubrimos que las especificaciones difieren entre navegadores y lectores de pantalla, como también que el soporte y la adopción a nuevos o mejorados estándares no siempre van de la mano, además de que los navegadores compensan las malas implementaciones obligando a tomar decisiones poco robustas, es decir con soluciones que funcionan bien para algunos pero mal para otros.</p><p>Las prácticas de autor, enfoques o recetas no generan siempre buenas experiencias y es importante evaluar la consistencia dentro de nuestros productos pero también en los productos más utilizados del mercado.</p><p>A partir de todo este recorrido generamos conocimiento y experiencias que queremos compartir con la comunidad, como así también nosotros fuimos receptores de buenas prácticas generadas por otros productos. <br>De esta manera logramos complementarnos y actuar en sinergia en pos de ponerle fin a esta paradoja: Precios altos o bajos, ¡pero accesibles!</p><p>Conoce también las <a href="https://medium.com/mercadolibre-tech/precios-accesibles-nuestro-aprendizaje-desde-la-perspectiva-android-2b2d1d1a7330">experiencias y aprendizajes que tuvimos con esta iniciativa para Android</a> en el artículo escrito por <a href="https://medium.com/@juan.unzurrunzaga">Juan Ignacio Unzurrunzaga</a> y la <a href="https://medium.com/mercadolibre-tech/precios-accesibles-nuestro-aprendizaje-desde-la-perspectiva-ios-785e7f971983">historia desde la perspectiva iOS</a> desarrollada por <a href="https://medium.com/u/83012da8aaa5?source=post_page-----785e7f971983--------------------------------">Ana Calderon</a> y <a href="https://medium.com/u/9224ed540e30?source=post_page-----785e7f971983--------------------------------">Laura Sarmiento</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c74ad7c62b0d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/mercadolibre-tech/aprendizajes-al-desarrollar-precios-accesibles-en-mercado-libre-c74ad7c62b0d">Aprendizajes al desarrollar precios accesibles en Mercado Libre</a> was originally published in <a href="https://medium.com/mercadolibre-tech">Tecnología de Mercado Libre</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>