Javier Ruiz Hi, thanks for joining us for a different animal.gov and design for a dead horse. My name is Javier Ruiz, Senior web publisher for Miami Dade County. I was supposed to be joined by my boss, Jamie Shycko. We've lost her to an endless string of emergency management meetings as the county works its way to the present crisis. Before she was consumed. She left us with this presentation. Bear with me as I fly solo. We're going to talk about some of the challenges faced on big institutional projects and how we hopefully overcame them. Javier Ruiz So what happens when someone wants to report a dead horse? Yes, it's happened. And well, it depends. The Animal Services Department might be responsible for picking up a deceased animal, but only if the animals owner is unknown. And if it's in the public right of way, but not a railway. Because railway isn't a right of way. It's private property and public schools. They pick up deceased animals at public schools and courthouse parking lots emphasis on parking lots. So how do we make sense of information like that? Javier Ruiz So let's begin by defining the problem with a few quick stats. Miami Dade County has 2.8 million residents. There's 34 municipalities within our county. There's a growing list of more than 600 services ranging from pet adoptions to potholes, more than 25,000 employees with lots of employee communications and newsletters. There's three main languages in our county, English, Spanish and Creole and we distribute information all three of those languages. There's more than 10,000 pages of content, countless PDFs. We have a 311 contact center that fields resident calls. And they answered more than 8000 calls a day, probably more now with the COVID crisis. And there's 70 county agencies complicating those matters. That content was scattered throughout various departmental silos. It was often out of date and even incomprehensible. Javier Ruiz So this project really was from the ground up. It started with the upgrade of our content management system and the implementation of a new search engine. And it even touched on the database that our call center uses to answer questions. It rolled out in three phases. We had the system upgrade, an animal services pilot project and a full rollout. we tackled it using customer data, multidisciplinary teams, agile and iterative processes, realistic milestones setting and demonstration of progress and user testing. We'll talk about some of these processes and how we applied them to get the site done as we go through the presentation. Javier Ruiz Minimum Viable Product, nothing succeeds like success. When handling large scale projects stalled momentum can be a project Small consistent successes keep your project on track. Visible progress keeps out at bay. The trick is to eat the elephant one bite at a time, define the minimum product that would meet user needs. If the vast majority of users only use services provided by a particular by particular agency, start there, then start with the most use services within that agency, maybe pick six of them. By working on a small test case, you'll finish faster and learn lessons quicker. Don't try to solve for everything at once. at this scale, it's just too much. Javier Ruiz In our case, we started our beta project by working within Animal Services to get our MVP. This allowed us to work out the kinks in our content models, user interface elements and taxonomies at a more manageable scale. Once we demonstrated success, other departments wanted in. So why Animal Services? Well, first off, it's small, single domain manageable has a clearly defined taxonomy. Javier Ruiz It contains a variety of content types that we were looking to regulate using schema.org as a guide, had various use cases for presentation, including marketing campaigns, service requests, calendars and so on. Going into the redesign, we knew we wanted to develop a search based architecture. To do so the central node of the model would need to be the content type, services, news events, organizations, etc. The goal was to get to beta start testing and iterate. To do so the team would need to develop templates to store and regulate the content with a user friendly and ADA compliant presentation. formalize a taxonomy and make it available via tagging through the templates. Develop components that would allow publishers to pick content for display on a page created Basic pages for departments, marketing, etc, will be called profile pages that would display a variety of content types and were visually appealing, and we had to feed the search engine. This would allow us to build a beta Miami dade.gov page and an animal services page with five primary links in the global navigation services, use events, agencies and contact each item in the navigation link to our index pages or directories for more information. Javier Ruiz And really, the goal was to try to get as much content out of silos as possible. To this architectural model we successfully dismantled the traditional departmental websites and moved into a seamless Miami dade.gov web portal. This allowed us to assemble pages by selecting content types from various departments maintaining a consistent global you Our interface using the hurricane guide as an example, and we get a lot of hurricanes, you'll see there are various services provided not only by the Office of Emergency Management by but by animal services and other departments. On the service page, each department has an agency card that links users back to the parent organization. And the use of breadcrumbs allows users to work their way back through our hierarchy. Javier Ruiz So here's an example of a Department website. We didn't get rid of them altogether, we just sort of took the content apart, and use our search based architecture to sort of reassemble it. In this case, we have a homepage with a global navigation breadcrumbs and page navigation. content was laid out by content type using the page navigation to drop down to each section that showcased high priority content to visually appealing components. A few more button passes parameters to the search engine and pulls up more content. From there users could access more content available in those parameters, clear them Start a new search or use the search filters to explore other content types, categories, audiences, or information provided by an organization. This means one record per talk topic. And we're not having to maintain static lists. It's all being assembled dynamically through the search engine. Javier Ruiz So before you get started on projects of this scale, you really have to take stock of what you've got. Before we got started, we had to do some homework. So where do you start, you have to define how big your messes you need a content inventory, which is a list of everything. It sounds daunting, and it can be if you have to check every link by hand. But we're all about working smarter, not harder. These methods give you a handy excel sheet. For one you can start with Google Analytics gives you a list of every page that's been visited. And that's great because it'll catch stuff that maybe is bookmarked, but not linked. It's not comprehensive, and if something hasn't been visited, it's not going to show up on your inventory. So you might need to do a little Bit more double checking. In this case, I kind of like to use Screaming Frog. weird name serious software, the frog can crawl a site and give you a list of every page that's linked on that site. It's free for small sites. We're not talking about small sites here. So you might have to pay a few bucks to get it. But it's worth the price. It's a reasonable price tag, and it's a really powerful tool. Javier Ruiz Content inventories and audits can be used in many ways. A content strategist, for example, might make subjective judgments about the quality of content. But for our architects, they're more useful for defining content types, building taxonomies, setting priorities and figuring out what you need to architect around us. For content audits. Most existing content on a website will fit three categories, current, necessary, outdated, unnecessary, or just unnecessary. Sometimes people call this rot or Ouch. There's different methods for architecture business. versus we want to keep the first two, which is content that's current and necessary and outdated but necessary, and we want to throw out the unnecessary and irrelevant content and pages. If applicable, a client or subject matter experts should be included in this process to help make these determinations since they know their content better than maybe we would. What's left in that keeper pile the stuff that's current unnecessary and outdated but necessary. That's the stuff that we need to architect around. that forms the basis of a card sort in our situation. That gives us a better idea of where to start with the taxonomy. More on that later. True intense studies often show up beside intercept surveys, and they're a great way to learn about actual users. With a few short questions, You can learn who they are and pinpoint their pain points. Some good questions to start with. First, describe yourself its demographic and nature is the user resident, a resident or contractor, maybe they're looking to adopt the pet. If you have a fairly good idea of the user types for a project, multiple choice options are fine. If you're not, you can add an other field. But be careful because this makes it harder to dice your data later. How often do they visit? power users are often really adept at solving problems simply because they've already figured things out. Good interfaces shouldn't require users to figure them out. So a high completion percentage among power users can be pretty misleading. It might actually hide serious UX issues, described today's task, what is the user hoping to accomplish today? It's the most basic question of any website. Were you able to accomplish your task today? task completion can be a good thing, but not always. So it's important to follow up by asking how difficult was it to accomplish that task today? Because just because they completed the task doesn't mean it was easy. A few other things To consider a Form Builder or survey tool we use Qualtrics at the county can make it easier to set up a study. Consider using required fields if you can. Generally blank responses and incomplete surveys just aren't as useful. Beware of open text fields. They're a great way of gathering subjective data that doesn't fit into a multiple choice format, but it's going to make it a lot harder to dice your data later. Short, easy surveys are more likely to be completed. Definitely under 10 questions if you can possibly manage it, less than five would be better. Aim for 300 plus responses. This can be a balancing act between how long you want to leave a survey up and how much information you want to gather. But 300 is a pretty good starting point. With this simple set of questions, you can use Excel pivot tables to carve up your data by user segment and pin down opportunities for improvement. You can even use this data to build personas by identifying large user groups and their associated tasks and needs. Might even inform user journeys later on down the road. Javier Ruiz So you might be feeling like you've got a pretty good handle on what you need to do. Unfortunately, in large projects with many stakeholders, it can be difficult to secure buying. Thankfully, real data is hard to argue with the preliminary report is your chance to make that argument. Yes, this means PowerPoint, you've got lots of great data now, carve it up, find opportunities for improvement and justify your decisions using your data points to make your case. For example, when we recently redesigned our clerk of courts website, our true intent study allowed us to identify our largest user groups and showed that across every subset, users were struggling to obtain court records and case information. We also noted that occasional users such as county residents struggled more than power users in the legal segment. Based on our Google Analytics data, we also knew that items in the NAB court In category courts and records drew more site traffic than any other sites section. This was not a trivial problem. This was a high demand function that was performing poorly. And that adds up to a problem that needed fixing the client bought in and we got to work. Javier Ruiz So we've analyzed our content we've learned about our users, and we've secured by and we're working on the project. Now, on new projects you might be starting with no content, or after auditing, you might find that none of the existing content is usable. When we found ourselves in a similar situation, migrating a giant but messy call taker database of 71,000 lines in Excel into our existing public facing services repository. We decided to start fresh and work from the top down through knowledge extraction workshops, we engaged our 311 call takers the primary touch point with many of our residents to list all the common questions they receive We then asked them to sort those questions into categories. These collaborative workshop sessions helped us create both the preliminary taxonomy, and it helped us list all the content that needed addressing. Javier Ruiz When redesigning large scale websites, we found that it's helpful to create micro teams focused on specific aspects of the content development process and placing people in roles where they excel. This is essentially helpful in highly decentralized environments. It's like a virtual assembly line. The cam team has an in depth knowledge of county services and subject matter experts within departments or at our 311 call center. The content development team is responsible for ensuring that the content that the knowledge management team assembled into a single record is written well, and it's web ready. The content publishing team as the team will enter that finalized content into the new templates. So in my early days as a content strategist, I stumbled on the idea that the way that pages or chunks sort of forms a narrative. I thought that was a pretty cool realization. And then it turned out that Smashing Magazine had already done an article about it a month before. The the writer in that article called it a content priority guide. It's a high level way of organizing the importance of content on a page. Think of it as a wireframe. But for words, to build the priority guide, you start with the actual page content that you've gathered. Next, you sort those bits of content into blocks in order of priority. Don't worry about layout or design this step. It's more about making sure that you have all the content that a user needs, and that you're arranging this content in a way that is helpful to users provide the most important information first, and the least important information last, give them useful information before presenting them with an ask. Priority guides are all about putting down the broad strokes of a content type or template. This process will help you decide what information you need, how it will be blocked, and where to live in the structure of the page. Javier Ruiz So content priority guides are our primary tool for organizing content than the taxonomies. And classifications models are really how we organize and classify all the products and information we delivered through our websites. Most often, these inform the makeup of our site navigation, but they also affect meta data use this filters and facets. While we could as eyes attempt to classify these things, according to our own understanding. From a UX perspective, this might be problematic. After all, the objective is to serve the needs of the users. Javier Ruiz Fortunately, there are a few exercises we can perform to help us understand how users view the relationships between the things we classify open card sorts in an open card sort. Participants sort through the cards representing all the content pieces we collect. During our content audit, they sort them into piles. What makes it open is that in these, these piles are unnamed, allowing the participants to name the categories as best fits their mental model. Closed card sorts are similar to open card sorts. However, in a closed sort, the category labels are already defined, and users sort the cards into pre named piles. This is useful as a quick way of testing how well an existing or potential taxonomy might work. And then there's tree testing. It's a sophisticated way of testing existing and potential potential taxonomies as users navigate through a fully formed taxonomy, and indicate where they'd go to solve a specific task, so in this case, you give users a task and you just let them navigate through a potential taxonomy and then sort of gauge how well they do. While tree test can be performed manually. It's easiest to reform these using a service like tree jack by optimal workshop. Javier Ruiz But there's more to organization than just taxonomies and categories, structuring and modeling your content and pages makes it easier to ensure the accuracy of your content and maintain a consistent user experience that helps users quickly form mental models. When we model content, we look at the information we need deliver and come up with content types that help us deliver that content. At a high level. These are basically temp templates like news releases, service pages and licenses. For example, look for common patterns in your content. The more structure you can provide, the easier will be to maintain large complex content repositories. And the more flexibility you'll have in displaying that content in different contexts, like wearable devices, apps and more. Javier Ruiz And structuring your content has several benefits. It allows the centralized storage of critical chunks of data. It's easier to maintain and ensure an accurate single source of truth that gets expressed in different presentations. It allows us easier reuse of that content in different contexts like a headless CMS. It makes it easier to assemble pages using those content types. Javier Ruiz So you have a working taxonomy, you know your content types, and after forming some priority guides, you know which blocks you need to populate a page. Next, you need to gather that content and structure it so that you can wrangle it and quickly build pages from pre assembled blocks of structured content. Here's how we assemble structured bits of data to help rein in our content. We create content types, blocks of information that house related bits of data into, say, a government service. When we populate that service, we include information meishan such as the name and the description of the service, and then we select from a picklist, a department or agency providing that service. The information is displayed in an agency card, a UI element that displays the organization data on every service page, allowing users to easily find contact information. If there's a new sheriff in town for a particular agency, for example, we simply change the director field in the content type, and every page connected with that data type. All of the service pages automatically pick up the change, ensuring that we can easily maintain the accuracy of large groups of content from one place. Here's an example of some of the content types that we might employ into a single service record. An organization might be Animal Services, the service itself might be Schedule A spay or neuter surgery. Location might be the pet adoption or protection center. The license might be a dog license news might be getting rid of rabies in Miami Dade County, for example. And an event might be a 24 hour mega match a THON, where we connect potential adopters with pets. Javier Ruiz Another advantage of structuring our content is it allows us to easily implement micro data like schema.org, enabling us to better provide information to search engines and address future channels of interaction like Google's Assistant and Amazon's Alexa. Javier Ruiz So coming back to our original question is a dead horse an animal services issue or a police matter? Solid Waste? Because circumstances departmental organization and mental models vary. This content likely has to live in more than one place to serve the needs of our users. The dead horse is an example of a poly hierarchy. piece of content that is a child node of two or more parent nodes in our taxonomy. Javier Ruiz When dealing with thousands of pieces of content and large, diverse audiences, poly hierarchies are an effective way to make content more findable. Really, they're almost inevitable. To provide users with a stronger set of navigation elements that they can use to weed through these poly hierarchies. We employed content types and our taxonomies as facets placed in the left hand side of our indexes services for example, these facets allow users to refine results and allow us to tag a single piece of content with any number of relevant facets that are all pulled in dynamically into the navigation by our search engine. Javier Ruiz By starting at the bottom and building our way up using content types as a central node of our architecture, we develop a search based navigational structure so users would not need to know Which organization they would need to contact to report say, a dead horse. Weather users navigate by audience segment, department or keyword. Our dynamic navigation can express the poly hierarchies inherent to government services. We started the project by getting feedback about our previous site to figure out a baseline for what we needed to fix. But testing really is a full time exercise. Keep an eye on your user feedback and use small set moderated testing to verify that your assumptions are working for users adjust and iterate because the job is never done. Thank you Transcribed by https://otter.ai