Content modelling beyond the tech: User needs first

Omnichannel Podcast Episode 28

Watch now

In this episode, Noz Urbina and Michael Andrews dive into the world of structured content and content modelling. They discuss the importance of bridging technical and business requirements, balancing the needs of omnichannel delivery with content creators’ usability, and why not everything needs heavy modelling.

“People are rushing into content modeling as if it’s just a technical necessity, missing the opportunity to thoughtfully integrate business and user requirements. You don’t have to model everything; sometimes you can leverage existing databases, allowing you to focus on structuring what truly needs to be unique and managed within your content system.” – Michael Andrews

Listen now

 

What you’ll learn


  • The common challenges faced in creating and implementing a content model
  • How to strike a balance between technical requirements and content creator usability
  • Real-world examples of what to model and what to reference externally
  • The evolving nature of content models and their role in digital transformation.

Also listen and download on…


Speaker(s):

Noz Urbina
Noz Urbina
Urbina Consulting

Full session transcript

THIS IS AN AUTOMATED TRANSCRIPT

 

Noz Urbina  

Welcome back to the omnichannel X podcast. This is your host Noz Urbina, and I’m here today with Michael Andrews, who is a well known name and in the little community of content modeling and structure content folks, who we’ve been talking to a lot on this series. About content modeling, and designing content for omni channel and personalization. This is part of that library of content. So if you haven’t checked that out already, check out the session series under events on our on our website, and you can sign up for all the sessions that came with that you’ll have the whole on demand library that you can use to educate yourself on all things content models. I’ve had we’ve had myself and Carrie Hane and Chris Saunders and Jeff Eaton and Juncker even off and Michael Andrews and and and and so if you are into content modelling, that’s a treasure trove of material. So really recommend you check that out and sign up for that whole package. It’s all free of charge just for just for registrants. Today with Michael, we’re gonna be talking about the fact that people are really rushing into content modelling kind of like a tech thing, like I bought a database and I’ve got it set up my database and that’s, that’s, that’s the way they’re looking at the whole content management and content modelling exercise when it’s really much more about understanding your user requirements, your brand goals, your channels. What kind of experience are trying to support and we wish that people would take a lot more time preparing before they actually went out and bought the systems and then started pressuring themselves to check their content in there. Also the fact that you don’t have to model everything you know, if you are a content modeler, you can think about where you can, as much as we want to reuse and talk about single sources of truth. Where are there times where you can just point out to an existing database and say, I don’t have to model that somebody else has already done a database that handles that for me like Product Information Management or like a movie database or if you’ve got product, a product data, database, or math or financial information, lots of these things. You don’t have to model yourself if if they’re already available in some structured format that you can integrate with. You don’t have to everything you present to your users doesn’t have to be managed in the one CMS. It just needs to be interconnected in a way in your concert hall. And then how do you decide what to structure which is, I think a great question which I’m going to try to bring up on our last. We’re having a panel of experts we’ve done two already. With with Carrie and Jeff and myself crews and Megan and recording stopped recording in progress and we’ve also talked about how you decide what you want to structure which I think is a really good question which we I’m going to bring up on our next panel discussion because we’ve had these panels with Megan Casey and crew Saunders and myself and Eden and Carrie hain. But we didn’t actually know we’re talking about, you know, assuming you want to model things, how do you decide what goes in your model? What doesn’t need modeling? Or what do you make an actual enter attribute or property or field as opposed to just leaving us text or whatever. And then, of course, AI Where does AI come into modeling in the future of structured content? Michael turned me on to this acronym rag which I’d never heard before retrieval augmented generation. I apologize. These acronyms are going to keep coming fast and furious as this new technology develops. But that basically that’s the retrieval word is the big one. The fact that you rather than kind of AI’s making stuff up when what based on whatever happens to be in their training set. You can use the AI to handle the conversational part but the facts are actually coming from a knowledge graph or a database where it’s can look up the real facts and make sure that it’s properly taking that into account when it constructs a response. So that sounds like a really interesting thing. I don’t know if this particular rag thing has to be the thing everybody goes with but that concept of fact based AI that doesn’t make stuff up and doesn’t propose non facts in its responses sounds really, really important. So without any further ado, enjoy the episode.

Michael Andrews  

Um, hi, everyone. I’m Michael Andrews. And I’m really happy to be here with Noz today.

Noz Urbina  

Awesome. Right. So Michael and I have known each other for Wow years. I don’t know We’ll come on a decade. Yeah. As a as people in the omnichannel unstructured content world we’ve kind of been in each other’s orbits. And as we’re doing this Omni, sorry, a structured content, designing content for omni channel and personalization session series. I thought it was great to have Michael on the podcast this year to add to that library of content. So I’m sure this is going to be evergreen because you know that we are talking a lot about a lot of the same things in the content modeling world as we have been for 20 plus years. It’s just that we’re taking them to new people. And talking about how they impact new channels, you know, but the core stuff is all it’s all been, you know, been there and the fundamental discipline is there. So hopefully these sessions are being educational for you. So hey, can you do this? I can tell you that these skills will last you a lifetime. So Michael, do you want to tell tell the people a little bit about yourself, your background, what kind of work you do?

Michael Andrews  

Sure. Thanks for having me. Um, I work as a content strategist for maybe the past 15 years and prior to that had a background in user experience. So I really think a lot about the intersection content and user experience. I’d like a lot of people have gotten very interested in content architecture. And obviously, structure is a huge part of content architecture. My roles have varied over the years, I’ve worked as a senior manager and a major global consultancy, doing CMS implementations and website replatforming for them. Most recently, I was working as a content strategy evangelists for one of the leading headless content management system vendors. So I’ve seen this from a different perspectives. From the agency perspective. I’ve also worked in house for large organizations like Verizon, and then I’ve worked with vendors. So I sort of see how important it is for implementations to be done right. And they really set yourself up for future success in the structured content is a huge part of that.

Noz Urbina  

Fantastic. And your main role in the projects is content strategy, content modeler that kind of thing.

Michael Andrews  

Yeah, it is. So a lot of it has to do with eliciting the requirements that the people when they are looking to essentially implement new systems modernize what they are working with, and starting to connect often disparate systems that need to coordinate with one another. And so there’s obviously a lot of technical requirements around that. But there’s a lot of business requirements and thinking about the user requirements, both from the employees perspective to provide the right content for customers as well as the ultimately the customer. End users perspective of everything that these systems are producing. So a lot of what I do is being sort of a Sherpa, I guess you might say, helping to bridge all these different types of functional responsibilities. I, you know, get stakeholders involved. Try to make sure that the right people are showing up to get the right requirements so that people know what it is they need to build. Because we’re at a stage now you can’t just expect there to be an off the shelf solution that’s going to do exactly what you need it to do for you as an organization. You have to design it yourself.

Noz Urbina  

Yeah, yeah, absolutely. I think well, that you’re, of course speaking the language of omnichannel you’re at omnichannel X. The I’ve done several diagrams about the Venn diagram like the brand goals, the user goals, and then the technical reality and in the middle, you kind of have content modeling and content strategy. So in your experience, what are the kind of the common challenges in creating and creating and then rolling out a content model?

Michael Andrews  

I think there’s several one is just getting people engaged. It doesn’t always strike people as being a an important thing to do. To set up a content model. They see it as kind of just a technical necessity. They want to get it over and done with as soon as possible. I understand their desire to, you know, get something launched as quick as quickly as possible, but there’s often just kind of an impatience around. This and people may misunderstand this as being entirely a technical requirement and not understand that business stakeholders really need to participate in this as well as the UX team to really get the right perspectives and and think about all the different dimensions that you want to consider when you’re developing the content model. You know, what are the author’s requirements? How is this going to be delivered in different channels? What kind of interactions are customers going to have with this and how is that going to affect how we how we structure that the pieces of the content, things like the customer journey, something I know you talk a lot about with a lot of passion that’s hugely important is to be thinking about that customer journey as you’re developing your content model. So these are all really important factors. And sometimes they get short shrift. I mean, I’ve worked you know, certainly, from the perspective of a content management system vendor, I’ve seen a lot of customers are like, Great, I’m ready to go. And I want to just do this as quickly as possible. And let’s just take what we have currently, and ported over to the new system and get this launched as quickly as possible. And they’re missing this opportunity to really restructure what they have. So that could do a lot more. So that’s first barrier. The second barrier really is around education that a lot is still pretty new to a lot of people they don’t necessarily know how content models are developed, what goes into them what they need to do. Again, it can be treated as entirely as a technical exercise and you may get someone who has expertise in something like data modeling, which is very, very valuable, but you’re not getting all the elements that isn’t are necessary in the content model. You’re not thinking about things like guidelines necessarily, you’re not thinking about all the validations that are necessary to produce quality content, you know, so these are, you know, content has some, you know, all kinds of things that need to be there to support the quality of that content. And those things can be built into the model. So you can automate these things and make it less arduous for people to constantly be working to try to produce the right content. So that we can have systems help them with that.

Noz Urbina  

Yeah, absolutely. I know. I am definitely passionate about making sure these things map on the customer journey, and that the and it resonates with me as a recurring theme. This is just you know, this is just the database stuff. You know, we need to we need to have a Sunday at data model to throw our stuff into our database. You know, in an ideal world, people will be working on their content and their content strategy and model for months, if not a year, maybe a year or so for some big organizations before they actually purchased some of these new headless CMS which are which are so hot. I am seeing the kind of teenager clean your room kind of transitions to headless where you know, we bought a headless CMS, we want to decommission all these other CMS is that are running around. So I’m just going to shove everything in the closet and then we’re gonna figure out how we’re going to structure it later. So the room looks clean because we’re in the new CMS but literally doing nothing to take advantage of this new options that we’ve just

Michael Andrews  

paid for. Yeah, I’ve seen that too. Yeah. Yep.

Noz Urbina  

So you mentioned the differing requirements. I have this question that’s dear to my heart is I’m interested to hear how you balance the technical requirements of the content model like what it has to do for omnichannel what has to do for the for the technology to do what it’s supposed to do with which usually adds more metadata, more complexity, with learnability and usability for the content creators, because I think a lot of our audience is probably, if they are just trying to get to grips with content modeling, they’re going how do I make this not an unwieldy beast, that no one can figure out or understand?

Michael Andrews  

Yep. That’s a great question. And I think we have to look at the maturity of the organization as it relates to this whole journey. You know, this is this is part of the digital transformation. And so you know, organizations are at different stages in digital transformation. And, you know, that applies to working with structured content as much as it applies to, you know, automation in general for business processes and other sorts of things like that. So you got to look where the organization is and typically organizations are at different stages of maturity within a larger organization. I mean, you know, there are pockets that are are doing quite well are charging ahead and other parts that are aren’t getting the love they don’t have the budget things like that, too. You need to be realistic about what part of the organization you’re working with

Noz Urbina  

at a given time. Because even within a single department,

Michael Andrews  

yeah, yeah. And where they are, and kind of develop a roadmap and you know, the beautiful thing about a content model as it can evolve over time now, it’s best to try to get everything right from the start. I mean, I definitely agree with you about trying to get all of your you know, as much figured out upfront before you start implementing anything, but there isn’t an opportunity to evolve this thing over time. So if you have a roadmap and you at least have the foreknowledge of where you want to be going, you can make it a little bit more sophisticated over time. Now, I think your to your point about the usability of this, it’s really, really critical. I mean, it’s easy to overwhelm people with content models, and they can be overly complex. And, you know, I’ve seen people tried to introduce, implement a headless system, particularly people coming from, say, a data modeling perspective, and they want to break everything down to the nth degree. And, you know, for them, it’s like, oh, I just can avoid making a certain call to a database. And this is going to be the most efficient way to do it for someone who is developer but for the person who’s trying to develop content, you can find you have this what turns out to be this incredible nesting sequence where people have broken things down to so many levels of detail, that they get lost. They don’t even know where it is they need to go to add a detail or revise something and that doesn’t work really well. So I think the authoring experience always needs to be upfront in the minds of people when they’re developing the content model. That’s why it’s really important to bring in the you know, the content strategists and even authors who would be creating some of this key content and involve them in the process and like, ask these questions like, how do you think about the themes that you’re going to be providing to customers? How often do you need to update certain kinds of details, and you start prioritizing how important it is to break apart different pieces. By listening to these people, you start realizing that, oh, this is something that gets updated a lot. It makes sense to break this apart. You know, I can live with a bigger chunk of content, it all tends to stay together. And you know, it’s only looked at at once this whole big chunk. But when you start having pieces that need to recombine with other pieces, you want to just update one place and things like that it makes sense to break it out. So that’s the kind of, you know, dialogue. I don’t want to call it a negotiation because I mean, it sounds like a, you know, a treaty you’re trying to sign here but a dialogue that happens between the non technical folks and the technical folks as well. So you get that dialogue going and then you try to figure out what what makes sense for the organization given where they are. Yeah,

Noz Urbina  

well, something we talked about some of the other episodes about how you’re not trying to you’re not trying to release the iPhone 12 on on day one. You cannot and should not think that the your first iteration the content model will be the best iteration that will ever be because you waited too long. And that’s just doesn’t make sense. Like business sense does make technical sense to try to just keep engineering this thing until it’s perfect before you do anything with it. Simultaneously. I think we, I was thinking to try to get a little less abstract to try to think of examples like what are what are examples of content that you do need to break out and what are things that you can leave together and there’s also one thing is, where are there if you can think of any examples that you’ve touched in your career and I have a couple of my head of where you go, Okay, I’m not going to model this in the content model per se. In any in any real detail. This is just going to be a pointer to a database that has its own data model. And we’re gonna kind of bring that in and not try to manage it as content per se in the Content Management System. It just has to be a reference to the external database.

Michael Andrews  

Yeah, that latter point that’s a really good good way to look at this you know, there are things like you know, product information management systems and things like that. That’s all structured data. You don’t need to be managing that within a content management system because there’s a whole nother system that can be the system of record for that. The source of truth, we like to say, you know, so yeah, and there’s lots of neat ways now to combine your content model with other data sources, you know, there’s, you know, the whole world of API’s. You know, since content now is just being defined through API’s, you’re able to connect things any way you want. So, yeah, think about different sources of truth. And, you know, where it makes sense to define what what is the source of, of truth or that back to your question, though, about the what things to model what things really are not the things to model. I mean, I think, you know, the really big opportunities to give you, you know, some concrete examples, so we’re finding more and more folks who work in marketing, for example, are getting interested in structured content. Well, an obvious place for them to start are things like landing pages, so landing pages have different pieces to them. And you want to say be creating, creating different variations of these different pieces so you can optimize that landing page. You can personalize it or customize it for different audiences or different scenarios and things like that. So that’s a really useful thing to be able to model. There are things that you don’t require a lot of heavy modeling your blog posts won’t require heavy modeling at all. I mean, it’s a pretty simple thing. I mean, you unless you’re planning on reusing pieces of that, which, you know, oftentimes you’re developing a blog posts, you don’t really have the expectation that you’re going to reuse pieces of it. You’re not going to have that natural structure there so much so there’s that kind of continuum of things that are, you know, you’re going to be reworking in different ways, either to create variations or customizations or to reuse pieces of it or, or things like that, and then things that you’re not having those same demands on you for the content.

Noz Urbina  

I think that I think that the blog post example is an interesting one because it makes me think back to one of the first examples I used to refer to for structured content, which was actually a blog about movies. So it was movie reviews, reviews rather than a blog, but it’s like they’re very structurally similar. It’s just to kind of write it’s a point of view in a kind of narrative structure. But what was cool about it and this is an interesting thing to try to make this more concrete for people who are getting used to it. The Movie Database is not part of the content model. But a movie reference is so let me break that down. So this, I think, is to make this concrete like you can have an article a blog post or let’s say a movie review, and the movie database has got the movies, the dates, the years, the directors, the cover shot, it’s it’s got all of the structure data about the movies, right? So but in your model of an article or review, the only thing that you’re modeling into that is saying here I have a pointer and it is a movie pointer. I’m pointing to a movie, I’m actually making a hyperlink. This is not a link from point A to point B and you’re not supposed to click it and go somewhere else. It’s it’s a signal to the to the processor, when you’re serving up my my review or my blog, post, my article, whatever it is, go to a movie database and pulled whatever we’ve defined, we want to show about a movie that I’m talking about and show it in a way that it is appropriate. So there you don’t have to do all the modeling of what a movie is. Just as you want to. You want to include movies in a structured way into your content model. So that takes a little stress out of a model anywhere where do we already have structured data that is in a good state, we just need to define a meaningful pointer saying this is a product reference. This is a movie reference. This is a future reference and so on.

Michael Andrews  

That’s a really good example. So yeah, with movies we have something like IMDB, you know, that has all this structured data around the movies. And then we have different movie reviews. And so we may have movie critics and different newspapers, for example, and things like that. And you might be able to say, all right, that critic is the author of this review. Maybe there’s a tiny bit of structure there. They give a certain number of stars potentially to the movie. And the review itself me have a minor amount of structure I mean, there there may be some sort of summary headline review in a paragraph and then it goes into, you know, something a little bit more and you may be able to pull out that that summary introduction section. Right. Exactly. But, yeah, and that’s really a good good example. So you know, someone who is accessing a system that was imagined or out of a theater and at a kiosk or something and are trying to book that ticket there. I mean, they may even be pulling out a piece of that movie review just to sort of see what what one critic said, you know, in the headline if you’re trying to make a choice around something, but the rest of that the details of the movie would would be coming from a database.

Noz Urbina  

Yeah. So then I think the guideline is like, are you going to update it separately? Are you going to show it separately? And are you going to have very particular guidelines about it? So the way that you write a headline the way that you write short description or meta description, those are very specific and that’s where you have kind of your brand voice so that deserves to be broken out as its own thing. So you can give those particular guidelines like this paragraph at the top is not just a paragraph, it’s the you know, it is the the teaser or the short description or the tagline or whatever. And so you want to give elements to the things that are going to have lots of rules about them, even they’re not going to get reused that much. But if you want editorial consistency, you know, every time we want, every one time someone hits one of these these kinds of content. We always want this for sentence to be like this first paragraph would be like that. And then I want to see the bullets that do that etc. Those are things that you want to watch for to pull out. We’re giving guidelines to the human modelers, where do you see AI contributing in the world of structured content and omnichannel? I think this is where and how

Michael Andrews  

Yeah, I think this is a really exciting area. So um, first of all, I think that structured content is even more important now. In the age of AI. We know that AI by itself, or at least generative AI lets you know because there are many flavors of AI out there but generative AI by itself has a number of limitations that does some some amazing things but it has a lot of limitations. And I think structured content really helps to overcome some of those limitations. But conversely, I think that generative AI can aid a lot of things as it relates to structured content. So one of the areas really is around helping to connect the helping to develop the right content for very specific targeted things. So one of the one of the things that have been a challenge for organizations is developing all the specific content that you start to need to have as the online channels become our primary way of communicating with customers, and they expect everything they could possibly ever need to be available online. And it’s like, whoa, who’s created all this content? Well, we have an opportunity with generative AI to help fill in some of those gaps provided we use it in a very, very targeted way. So we give us some very structured inputs. We tell it we need some sort of description about something that doesn’t exist currently. It can generate that description and then we can slot it into our content model so that that well defined piece of content then can be delivered, where it’s needed when it’s needed. And it’s it’s gone through some sort of process review. So we’re not just getting some randomly generated thing and we’re not sure how good it is. So I think that’s really a big opportunity for generative AI. I think they’re, you know, other areas. We were just talking a moment ago about databases and connecting it to content. So you know, if you were well aware of there’s this whole area of semantic AI or you know, knowledge graphs and we’re seeing a lot of opportunities now to even connect generative AI with semantic AI and their opportunities. To make those things work together. There’s an approach and generative AI called rag, which is retrieval augmented generation and basically points to a database and you say, I need you to tell me the concrete facts about something before I can generate something that makes sense to people. And you’re able to use this rag approach with with knowledge graphs, starting people starting to publish how to do this and starting to develop tools on this. So it’s still pretty early stages. But this is exciting stuff. And I think there’ll be a lot of more of this happening. And it’s really going to be something that any of us who are interested in this area are going to look into and hopefully be able to take advantage of.

Noz Urbina  

Yeah, awesome. Well, it’s great because you kind of ping two of the two of my favorite things, which are one is I think so many people, and I just just did a webinar on this. So many people are kind of treating AI like, like a search engine. Yeah, in fact, they talk about querying it they give the rather than or like a prompt of language that we use. I think it’s kind of broken around AI, rather than rather than thinking about the AI something I’m speaking about generative AI, you can actually talk to church at BT Gemini Bard, club that kind of thing. We kind of like it’s it’s like we’ve swapped the word prompt the word query. I think in the general lexicon. People are just thinking about how do I construct the perfect query? How do I ask the right question? Yeah, it’s this sort of one shot thing is where as realizing that we can have a conversation with this thing, like you don’t have to just engineer your perfect paragraph and then send it and then get a response. You can talk back and forth iteratively and improve these things. Yeah. So but if if you’re doing that, I’m doing that in a way that matches with your constant level, which is what you just raised, which is one of my favorite things. And the other thing I don’t think people are talking enough about, which is that you can explain to an AI When you reply, I want you to reply. In this structure. Yes. You know, give me give me a summary of this thing, or give me expand out this summary, et cetera, et cetera. So you can tell the AI this is this is how my content works. And you give it the names and this is called the short description. This is called the headline and etc. And there’s a missed opportunity often to they just want to kind of get the AI to block out the result which then they rework, but you can explain AI structure and they can get it and you can plug it into your content. Model and say, This is how you be short description responses. This is how you do feature scription responses. This is how you do you know, chat responses. And then it’s very capable of actually doing that of understanding your content model and giving appropriate responses based on those guidelines you have on a element and element by element basis. And then this reg thing, making sure that it’s actually got facts behind it. Yeah, that’s, you know, I think our first podcast of this whole content modeling library that we’re putting together, I was talking with Yonko even off And at the time, that was the problem of hallucinations was top of everybody’s mind the fact that you know, these these things, were just making stuff up. They didn’t have a good answer. and they stopped they still they still do it. You’re just using chat GPT and hammering away at it. But I think that we need this popularization of something like, so what extent organ

Michael Andrews  

retrieval augmented generation

Noz Urbina  

retrieval augmented generation. Yeah, I like that yeah, so. otherwise, you know, get the facts. Yeah. But then wrap those facts in a conversational way so that they work for content context or conversational context. Wow. Okay, so we burned through that time. That was that we need to get you back on the podcast, I think because that that flew by. So I want to do one last question. How can you do give any tips or feelings on how you measure the quality or effectiveness of your model? So you’re done. You’ve designed it, you’ve got your stakeholders on board. How do you know that? Your model

Michael Andrews  39:38

is a good one? I think you have to get different stakeholders to buy into it. So you kind of go back to my earlier theme is, you know, one, one person can’t hoard the ownership of this, of this model. And it’s got to represent a lot of different points of view. So this is kind of basic user acceptance is what we’re what we’re talking about here. You know, I don’t know of any sort of, you know, technical way to evaluate this. I mean, you raise the very important issue earlier about what is the right level of complexity, and that’s so going to be so situational. And so yeah, I mean, it’s something that is a judgment call really, I hopefully this is over time going to get to a maturity you know, where we’ll practitioners like you mean, you know, the other folks on who aren’t been on your program, we’ll develop some protocols about how we can sort of make some judgments about these that are going to be a little less individual. But yeah, I think that’s something we want to work towards, but I don’t think

Noz Urbina 

we’re quite there yet. Yeah. Okay. All right. So yeah, I won’t, I won’t I won’t keep you any longer. I know that people said they wanted shorter episodes, but I’m dying to talk to you again. So we’ll have to find that time in place to make that happen. What’s the best place people can get

Michael Andrews  

in touch? Or your LinkedIn guy? I’m a LinkedIn guy. So I’m pretty much pretty much on LinkedIn these days. So yep, please, please connect with me there if you’re not already, and I’ll be happy to

Noz Urbina  

get to know you there. Fantastic. Alright, so that’s Michael Andrews. Thank you so much for your time, Michael and we look forward to chatting with again sometime.

Michael Andrews  

Alright. Thanks for having us.