Using Structured Content for a Personalized Customer Experience
When a customer does a search on your site, can she quickly narrow down the answers or will she have to sort through hundreds of documents to find the answer she is looking for? The challenge is too much information.
Wouldn’t it be better if you could serve custom content to every individual based on their interest and need? Structured content, or thinking of your content as data, lets you do that.
How do you turn content into data? By using XML or Extensible Markup Language.
In this episode, I interview Jonathan Price about using XML, structured content and customer forms to deliver a high personalized experience to your visitors.
With examples from Zappos and Amazon, Jonathan explains:
What is XML? And how to think of content as data.
How to use a database and XML to create custom documents for your visitors.
How to create more of an immediate conversation on your site through an FAQ.
Building individual personas based on interests expressed on the web site.
This episode starts abruptly. I was asking Jonathan some questions in advance about some life science websites he had been reviewing. I was already recording. We got well into the conversation so I decided to just keep going.
Our Sponsor: ACP-LS
About My Guest
Jonathan Reeve Price is a writer and information architect. He has worked with an A to Z of big name clients and regularly teaches online classes in information architecture, content management, XML, technical and web writing. You can find him on LinkedIn.
This transcript was lightly edited for clarity.
Chris: This episode starts abruptly. I was asking Jonathan Reeve Price some questions in advance about some life science websites he had been reviewing. I was already recording. We got well into the conversation so I decided to just keep going.
Jonathan: I wanted to see what exactly is life sciences and what kind of challenges do they face in marketing. So I took a look, and one of the things that really struck me was we’ve been talking about the need for starting with what customers want to know, the kind of questions they have, their needs at different points in the life cycle of the sale. And obviously, it’s great to start out by having questions and answers.
So one of the things that I look for is a good FAQ. And I think a lot of customers think, “okay, it’s more like a conversation”. So I’m more willing to go to a FAQ. The problem that I ran into when I’ve looked at some of the life sciences sites is that there are an awful lot of questions and an awful lot of answers. So it’s hard to find the particular question that I have in the midst of all of these many, many, many good answers.
Let me give you an example. On one site, there’s a search, and the whole idea is really good, because clearly scientists like search. They prefer search to browse. They feel like they’re going to be able to pinpoint what they want. And so you need some form of advanced search. And what this group was doing was saying, “Okay. First, tell me what kind of resource you want, like an FAQ, a protocol, selection tools, troubleshooting guide, anything. So you pick one.” I picked FAQ. And now, you can also refine your search by putting in a project name or a number, an application such as cellular analysis or cloning, DNA amplification. You select that or you add a category, such as cellular analysis or glycobiology.
Okay, so now you hit Submit. And for say, Application: Cellular Analysis, 101 results. Cloning, 1,160 results. DNA Amplification, 465 results. And there they all are, this big, big, big, big list. And if you’re really lucky and patient, you can find maybe your question. Maybe you won’t.
So the problem is that they don’t let you dig any deeper. And so you’re stuck at a level where you’re not going to likely find your question or get the answer you want. Now, that is a real problem is addressing what customers want. They come to your site. They do a preliminary search. And they get too many answers.
So we’re used to saying, “Okay, one solution to that is a more advanced search where you offer more categories to refine the search.” That’s okay, but you can’t predict when you have…like in some companies, there are say a dozen different categories, a dozen different application areas. It’s hard to predict which one of those someone might care about.
So the challenge is you got too much information. How are you going to make it easier to find? An analogy is buying shoes online. Have you ever been to Zappos?
Chris: Yes.
Jonathan: Okay. So they have at least 10 different menu systems. They all take you down to that one perfect shoe that’s exactly the one you want. But there are many different menus, which means that they reflect the different ways that people think about buying a shoe. Are you going to search first by size or by gender? It’s for men, for women, for kids. Are you going to do by brand? So they offer you an alphabet of brands. Or you choose “men,” and then you have clothing and shoes. And within shoes, then there are even more categories.
But once you get down to the subcategories, they do something that’s very nice. They tell you how many hits you’re going to get at a subcategory. You’ve gone to shoes, and now it says, “Running.” And it tells you, “120 different running shoes you can buy at Zappos.” So you know in advance, how bad is it? How many hits am I likely going to get? And that’s called faceted analysis. It means you’re not forcing someone to actually go in and look at all of them and count them yourself.
It anticipates, “Okay. There are all these subcategories, and here are how many records we found.” So this is database thinking. It says, “Okay, there are 120.” Now, when you go down, you say, “Okay. Let’s go in there.” I click that, and it opens up and it says, “Do you want me to sort by gender, performance, color, brand?” And for each of those it says, “Color: there are 16 colors. There are 15 brands, their performances.”
So I am filtering down my search in a browse-like way. I’m browsing down the set of possible results in there. This is called faceted analysis. For a particular shoe, there’s a record, a database record. And it has a bunch of fields, such as specialty, gender, performance, color, brand. And each one of those has been filled out. And therefore, the system can say, “Tell me how many shoes with a Nike brand that are running shoes are available,” before I even click and say, “Show me. Show me these things.” That would be a big help in these sites where there are many answers and many products, many applications, many categories.
Is there a way to think database? Some of these searches almost are document-based. They’re thinking, “Okay. We have a single question and answer that’s a document. We’re going to give it to you. And here’s a list of all the documents. Good luck.”
Chris: Right.
Jonathan: But it is disappointing to the user. So how would you go about solving that problem? One way is by hand. You go back to every record you have, and you say, “These are the categories that we think people are interested in.” Better would be if you go to customers and say, “Okay, you’re in the cloning business. What actions do you take? What tasks do you take? What actions do you perform? Okay. And what are the terms for those? Now, let’s see where those terms show up in all the information that we’ve got so that we can find them for you later.” That means you’re building a persona, a picture of each type of customer, usually a scientist of a certain level and interest. And what are they interested in? What are their tasks? And you take the names of those tasks and build a list. And we’re going to sound scientific and call it taxonomy. It’s really a hierarchical list of all the concepts that your customers care about.
And again, the analogy with Zappos would be, okay, people think in terms of shoes being for men, women, or kids. They think of them for some sort of purpose like sports. They think of them from a brand. Okay, all of those are going to categories in your taxonomy. And you’re going to apply those. And somehow by hand or by software, you’re going to apply those to all of the records in your giant database. And these companies live and die by databases. So it’s not something new. They already have the data. And it’s already what we call structured data in that it’s well-organized to a point. It’s just they need to go a little farther because they’ve got so much of it.
Chris: Okay. So I want to stop you there. So I’m going to include that in the podcast even though when I first asked you that, I was just trying to understand where you were going to go. But now that you’ve said that…And I will figure out how to make this work in the audio. But now you’ve introduced that topic… And I think it was a brilliant example of you go to a site and you search. And now you’re going to find out that you’re going to get some unimaginable number of records and hopefully but unlikely that the search mechanism on the site is as brilliant as Google is at figuring out what you actually meant. And you’re going to be overwhelmed.
And you’ve given an example of Zappos where you get to see how many records you’re going to have to dig through to find what you want. And so, we’re talking about structured content as data. So that’s the topic for today really is thinking about your content as data, and how you organize it.
And I think you mentioned XML. That was the original reason I connected with you. I read an article about create one to publish everywhere on the ACP-LS blog. So now, let’s talk about that, content as data and then describe XML, because I think people have likely heard of it. They may have even seen it. It looks a little like HTML, but it does something somewhat differently.
So if you would explain that whole idea of content is data and what is XML and how it works.
Jonathan: Yeah. So XML is the Extensible Markup Language. So Markup we’re familiar with. Markup is things like you circle something and you tell the writer this should be in italics or you [inaudible 00:12:06]. In HTML, we put tags around content. In XML, the tags are just a little more meaningful than in HTML.
The tag says emphasis. The style sheet says it should be bolded. We’re categorizing the information in XML. XML is just a machine for developing tags. It lets you say, “Okay, this is a biography, and this is a book about the person where this is the package insert. And in the package insert, this is the result of the FDA analysis.” So that later, we, as humans, we can find it quickly by looking at those tags. But also software can come back later and say, “Oh, let’s excerpt this FDA analysis, and we’ll use it over here in our brochure.” So the XML tag lets software find the particular piece of content.
Compare it with HTML where all it says is, “H1 Heading, Paragraph.” It doesn’t say what the paragraph is about. What XML does is it lets you create tags that give meaningful labels to your content, such as “The topic here is cloning. And in this section, we’re doing the introduction to it and here is an application.” And each of those would be a tag.
Chris: That you get to make, right? So that’s one difference is you define the tags.
Jonathan: Exactly, for your particular business. Now, these days, because XML’s been around for quite a while, each industry has its own set of tags that you can start with and that, generally, if you’re in a medium to large company, the IT folks know where to get this list of tags. And they know how to create the tags.
You already have a source of tags in your database. The thing about databases these days is they can export anything into XML. And they can take anything that’s written in XML and bring it into the database very easily. Hit a button and in it goes. Why is that? It’s because the tags appear in a hierarchy. Your database record has a hierarchy too. There’s the record, and then inside it, there are fields.
In an XML document, there’s the major topic of the document and, under that, a series of other topics that are going to be covered that those labels, those tags correspond to field names in the database, one for one, letter for letter. So the database looks at the XML document. It says,” Oh, I recognize all of these tags. And I will just grab whatever is between the tags.” You know how in HTML there’s an opening tag and a closing tag. The database says, “Oh, I know where I am. I’ll take the content that’s between those two tags, and I’ll say that’s the value to put in my field.”
And so it can zip through a whole XML document and rip it all off and put it into a database very easily. And in turn, you can reconstruct or you could have the database output from a set of records into an XML document, which is basically all text. It’s text only. It’s not formatted. It is not a Word document. It’s not yet a PDF. It’s just text.
That means that any software can then manipulate that text document to make it look pretty. You can use a style sheet in your web server to deliver it, Wellpage if you want. Or you can use another piece of software to turn it into a PDF with formatting.
So the value of XML is that it’s neutral as far as what it looks like. It has no information about formatting. So it’s not like a word document that has thousands of Microsoft tags hidden behind those scenes. When you press return in a Word document, if you ever get a chance to look at behind the scenes, there is like 40 different tags that apply to that paragraph, and they’re all about format, that the margin, there’s 1 inch that you’re using, Gill Sans, 14 point, it’s bold, it’s red, all of that information that’s hidden in that little paragraph return at the end of a paragraph in word. All of that is junk as far as the database goes. It is of no interest to the database.
All it cares about is: what is the value of the text that I have to put in a field?
Chris: Okay.
Jonathan: And that’s where XML helps.
Chris: So now we have an idea of what XML does, how it works. Now, let’s talk a little bit about how we can use it. I know you’ve talked about using this in technical manuals to assemble information, and maybe different customers of a company want different versions of a manual. But of course, you want the same technical content. But let’s talk about it from the point of view of marketing content. How do you see it being used there if people could do something similar?
Jonathan: There’s not a sharp distinction in the scientific world between technical information and marketing.
Chris: Sure.
Jonathan: A scientist buys based on what they like to think of as fact. They need a lot of the information, a lot of data, and they’re comparison shoppers. If they don’t get the information from you, they’re going to go next door and look. And if they get information there, they’ll feel more confident.
Now, as we just talked about it’s hard to find that information. And they would like to be able to pinpoint just the key fact that they think for themselves are the most important. Now, there are two stages. First, they have to find them, and second of all, they have to assemble them in some form that they can share with their colleagues, go to their boss and say, “I wanna buy this, and here is why.” And here’s the long printout of all the material that makes the case for this one person buying this one product from you.
That means, in effect, they need to be able to, first, find the right records in the database and then export them into XML and have that XML assembled by you into a nice PDF document that looks pretty that they can print out and take to their boss.
So the first problem is finding the information. The second problem is formatting it, which is relatively simple if you have tags. You say things like, “Oh, it says that the name of document is such and such. That seems to be a title, so we’ll make that 24 point bold. This section called ‘Additional Details’ seems really less important, so we’re going to just make it regular text and 12 point. And it’s going to appear less important because the formatting is less important.”
In that way, if we can let people find the information that they personally want, we should be able to offer them the chance to assemble that in a bundle, call it a document, an XML document is what it would first appear as, and then say, “Format it nicely as a PDF.” And software can do that. That is not a complex task.
Whatever we see these days on a website is being formatted by a style sheet. The same thing would be true here. The style sheet basically says, “Oh, additional details, formatted 12 point, Gill Sans,” and that’s it.
So what we’re looking at, the biggest challenge is to figure out how to help people find the information.
Chris: So I love that customer-centric approach. I was thinking of this more from a marketing point of view. So from my point of view and helping content marketers create content more efficiently, I think there’s an opportunity there, right? It’s to put your content as data into XML and then use that to assemble documents based on what you think a customer needs.
And I don’t even know if we need to talk about that. I think if they understand this concept, that could certainly be done. I’m curious to hear more about how a customer comes to the site, and would they simply ask a number of questions that would end up in a document being assembled for them? What does that look like?
Jonathan: Yeah. Let’s start off by helping someone develop a profile, some picture of them. And that means we’re going to ask them a number of questions in order to get a better sense of who they are. At the very least, we want to know things like, what kind of application are you doing? And check one. Is it epigenetics, glycobiology, sequencing? What is it? Just check one.
Okay, so now we know what application you’re in, tell us a little more about the tasks that you personally perform, and, hopefully, we have a list of those for each one, say, “Here are the 10 major things you do in cloning. Please click whichever ones you do.”
Okay, so now we’re saying, “Look, this is going to help you find the information that you want.” And we’re secretly under our voice we’re saying, “And it helps us know how to target you.”
But it’s like a profile in any site, e-commerce site. The more that we know about the customer, the more the customer finds it convenient because we can get past a lot of irrelevant stuff that we don’t care about. “I’m into cloning. I don’t want to hear about protein analysis. And please forget the protein expression. I know it’s exciting. You’ve got some new stuff. I don’t care.”
So a profile is critical for personalizing the information. It helps us know what you would want.
Chris: Yeah.
Jonathan: So again, what we’re looking at here is developing a sense of for each of the application areas that we already market to, what are the tasks that they do, the major actions that somebody performs in their language so that we can speak in their terms but also that we can say, “Okay, you’ve mentioned two things. Now, which of these would you like to know about?” And now, we run our search and find only documents that relate to that, only information that relates to that.
I think it’s helpful for marketing folks to always think of Amazon.
Chris: Yes.
Jonathan: Like “other people who bought this book, bought this”. That’s a data-driven site. It markets the heck out of everything. But it does it because it offers you a sense that they know you personally. They know what you’ve bought. They know what you’ve looked at. They know what you’ve reviewed. And based on that, they’re saying, “Here are some suggestions.” And often, it’s the suggestion that you really want, not the item that you searched for. And so serendipity counts here. And in marketing, we want to be able to answer their questions directly, but we also want to say, “By the way, you might also be interested in these things.” And again, the profile helps us sort through our own data before we display it to you over the web.
Chris: Right. I love that. This is way more of a new way of thinking about content and XML and structured data than I had even envisioned. So I had this idea that you and I discussed before. When I work with my clients, I try to get them to think of all the questions a customer might ask about their product and then to create an answer to each one of those questions by itself and then figure out how those could be assembled. So we’re really talking about very similar things here. And then create their content based on that and also what types of content people might be looking for, in other words, webinars or application notes or so on. You’re taking that a step further and just letting customers build a profile and ask questions and serve custom content to every single individual based on their interest and need.
Jonathan: Yeah, that’s the goal. Obviously, it’s not something to snap your fingers into right away. The good news is most of these companies are built around database, in fact, several databases. So because they have information carved up into tiny little chunks that each one of which sits in a record with a bunch of fields, it is mostly a change in a way we think about the marketing material that we produce. Instead of us lecturing the user on features and benefits, it’s the user asking us what we can do for him or her given our product line. It’s odd. It’s all this data, but it’s more of a conversation where the customer gets to talk to us and say what they want much more than the old way where we’d be in a marketing session and we’ve done our personas and we’ve done our demographics. We think we know what they want. And so we produce materials that appeal to people who’ve got those problems, those needs. And we stress those features. Well, okay, sometimes that works. And sometimes it doesn’t. In different companies, people have different roles. They don’t all do the same thing at company A, B, and C.
And if we think as marketeers, we’re marketing to a person who is a cellular analyst, our picture of that person is limited. It’s fixed. Whereas, the actual human being who’s coming in, who does cellular analysis, is also really interested in something else, and we can’t anticipate what they do, their tasks, their actions. We also can’t anticipate which concepts, that is, terms like epigenetics, that this particular person would care about.
So in effect, we’re saying, “Great, we have a lot of stuff for you, and just tell us a little about yourself so we can help you get the information you need, because we know you’ve got a meeting in 20 minutes. The printer’s down. The network’s flailing. And you’ve got to present your ideas to the boss in 20 minutes. We don’t want to waste your time. We’ll just give you the stuff you need.”
Chris: I love that. I want to wrap this up talking about one more thing. So you mentioned in the article that I first read that when you assemble content in this way, it has a different feeling because it wasn’t written with transitions between sections and so on. And I’m curious where you see that going, how that’s going to improve. And when you and I spoke before, I mentioned…
Jonathan: It’s staccato.
Chris: Yeah. Now, I mentioned the example. Before, I used to listen to weather radio when I was doing a lot of sailing. And obviously, there was a machine reading text in a very mechanical way. Now, I ask a question to Siri on my phone, and she talks to me in the same way you would. Somehow, that software has learned how to speak. Can a software learn how to transition so that the content that we deliver in a written form sounds more like a story or more natural than it currently does? Is that a possibility?
Jonathan: Yes, I think so, and oddly it’s because if this data approach. It’s like we ask Siri something. Now, Siri asks as back like, “Which application are you particularly interested in?”
Chris: Right.
Jonathan: From phone menus, we know how awful this can be. But increasingly, we can do it that way better than the chat people. But we also have to write scripts for the chat people. All of this is what the IT folks would call a branching diagnostic. Like you come in and you say, “I’m looking for something in the area of cloning.” Now, we ask you this. You responded that. We say, “Fine, okay, this is the area that we’re going to go to.”
At that point, we give our rap about that particular product or about the background on it and we open ourselves up to natural language questions like “What does the FDA say about this?” or “I’ve got a particular situation here. What’s this?”
A lot of what we’re going to be providing to the scientists and researchers is facts. It’s numbers. It’s data. At that point, they just pull out the fact. They’re very suspicious of what they call marketing. They’re quite impatient with anything that has a lot of adjectives and that talks smarmily about, “Oh, it’s an enterprise-wide solution.” They could care less. But they do want answers to their questions.
So I think as long as we can chunk this material so that it does answer particular questions or directs them to a particular page in a site, we will be responding to them in a way that they want. Where are things seem staccato particularly is, okay, we write a whole brochure or we say we assemble one of these XML documents we’re talking about. The pieces were not written in sequence.
Chris: Yes.
Jonathan: So one thing comes after another and doesn’t appear to be flowing. That’s painful for us as writers and as marketeers. But for your customer, it’s not so painful. It’s the good stuff. And all of that fluff, the stuff that we worked on hours and hours, all that verbal stuff is eliminated. Or at the best, it goes at the end where we put in the standard boiler plate about how wonderful our company is and how we’re eager to serve. That goes at the end. But first, all of these factoids, these tables, all that stuff goes at the beginning.
And again, it’s serving. It’s not beating the customer with slogans. It’s saying, “You’ve expressed interest in this. Here are the facts. What else would you like to know?” It’s like marketing in nouns and verbs, not adjectives.
Chris: I love that. You’ve opened my mind to a new way of doing things. It’s more customer-oriented than I imagined. So I really appreciate this conversation. I’ve learned a lot. It didn’t go quite where I thought it was going to go. In general, it did. But I definitely learned a few things that I did not expect from this conversation.
So Jonathan Price, I really want to thank you very much for your time today, and I’m looking forward to putting this one up on the website.
Jonathan: Okay, thanks a lot. It was good to talk to you.
