Delivering Product Value Means Delivering Insights

Charles Caldwell at UXDX EMEA. Video: https://youtu.be/r5kN5UhL-ZA

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

Applications that deliver insights

[00:00:00] Hello, everyone. I'm Charles Caldwell and I lead product at Logi Analytics. In my career, I've focused on enabling humans to make more effective decisions by delivering data-driven insights.

[00:00:14] I think we can take for granted that at this point, software has in fact eaten the world and that applications are a major part of our daily lives, both professionally and personally. One of the key lessons I've taken away, especially from the early days of internet commerce, was that the most successful applications were those that didn't simply facilitate transactions but focused on delivering insights. Insights that helped me make decisions faster, helped me be aware of options that were available to me, and helped me make decisions that were more effective for what I was trying to accomplish.

[00:00:48] The reason applications are so effective at this is because they take generalized technology like analytics and AI, ML, blockchain, geospatial, and they apply those to solving specific problems. An application exists in the context of the problem that it solves. It's got the ability to provide insights in context, where those insights have the most meaning and immediate impact on a problem that your end users are trying to solve. Applications are also purpose built, and they take on the structure of the decisions to be made and the actions to be taken. And applications, especially with mobile phones today, are often very close to where the action takes place. So they take those technical capabilities and they're able to translate them into real world value by translating them directly into action.

[00:01:48] Now, especially for those of us building B2B applications, I know we're feeling a lot of pressure right now to make our applications deliver the same level of usability that we see in a lot of the consumer applications. We do a lot of surveys on this at Logi Analytics, and I looked at a lot of research in the market on this to keep myself informed. Frankly, every survey that I see throughout my career, where we ask professionals in an organization about analytics, generally comes back stressing the same thing: we need more and there are lots of gaps.

[00:02:30] We see that today analytics is increasingly essential to the daily lives of your end users, and the seismic shift to a hundred percent remote work that we've all experienced during the pandemic has really served to highlight that this need is true across all types of work. It's not just the knowledge workers in the ivory tower. It's really everyone. And the gaps are still very large, with most workers reporting that they spend most of their time just looking for the information, trying to find the information that they need, and very often, once they do find it, it's not presented in a user-friendly format that they can take advantage of.

Three ways to deliver insights

[00:03:14] Now, delivering insights in your application, frankly, there are a lot of different ways to do it. Today I want to focus on three specific ways that you can take away and think about that'll allow you to add value to your application by delivering analytic capabilities within the user experience of your app.

[00:03:36] I'm not going to talk a lot about traditional reporting today, providing summary and detailed listing of data. That is a valuable use case, and I think we're probably all pretty familiar with it. So I'm going to omit it, but don't necessarily forget about it.

[00:03:52] The three that I really want to focus on today, that you may not be aware of, are, first, the idea of delivering an insight right at the point in your application where a user is going to take an action. The second one I'll talk about is enabling your users to explore so they can discover options that they may not have previously been aware of. And the third one I'll talk about is providing recommendations that will enable your end users to take a better or more effective action than they might have otherwise had you not given them that recommendation.

Insights at the point of action

[00:04:30] Let's take a look at the first example, and it's one we're all super familiar with from Amazon. The idea here is, if you want your end users to make decisions faster, more efficiently, you can provide an insight that increases their confidence right at the point of action. That's going to help them decide more quickly. Amazon, frankly, has done an amazing job of this with review summaries that show the summary rating and how many reviews, and you hover over and it shows you the distribution of the review scores. That summary is often enough, frankly, to move me from "this looks like an interesting product" to hitting the Buy Now button.

[00:05:14] In the context of your application, you want to ask a few questions. Where are users making a decision to take an action? Where are they buying a product or approving a purchase or prescribing a medical treatment, or selecting a vendor or identifying a risk? At each one of those points, ask yourself: is there an insight that I can provide right in line within this user experience that will help them know if this is a good decision or not, and if they're taking the right action? Then, utilizing very targeted data visualization within that user interface, in close proximity to where the user would take the action, you can deliver that insight.

[00:06:02] Now, these types of embedded analytics can be really straightforward. If I'm building a CRM system and I need to know if I should follow up with a contact based on a recent interaction, it's something as simple as whether their order is on time and how long is left to deliver that order, or if the contract is still unsigned and how long until the planned close date. These types of insights don't need to be complex in how you're presenting them. They really just need to matter to that end user in helping them decide: should I call this individual? Should I text this individual? Should I reach out to them? Or maybe everything's okay and I can leave them alone right now.

[00:06:49] You also get massively improved usability, because if I got to this contact screen and was wondering these things, I might have to go off elsewhere in the application to find the information and then navigate all the way back in order to take the action. So you can really get a lift on your usability with the right targeted insights embedded in the application itself.

Enabling exploration

[00:07:14] Now, there are times when you might not be able to define that perfect targeted insight, and you know that your end user has a more open-ended question, or they're going to get stuck and want to ask more exploratory questions. "Why" questions are good examples of these: why is something happening? In that case, your user is moving from a state of having a targeted decision, am I going to buy this product, is this a transactional risk that I need to flag, and they go more into a browsing mode. They're searching for options, or they're trying to explore relationships, or they're looking for explanations.

[00:07:58] In the example of Amazon, this happens when you don't like the product review. You look at the product review, it doesn't quite look right, and you start reading individual reviews. You start looking at comparison products, you may abandon the Amazon experience entirely, and you may go look at a buyer's guide somewhere to explore the category.

[00:08:18] Now, this type of decision support, this type of analytics, frankly, can be a lot more challenging than the targeted insight, because it does start to open up to feature sets and capabilities that are richer and more open-ended: self-service data discovery, guided analytics. But again, the place to start from a design perspective is to ask yourself: are there potential relationships in the data that aren't readily apparent from the standard user experience flow of the application? Are there things that the users, in the course of using the app, wouldn't happen across or be able to see or notice? Is there data that a user should compare that they just aren't normally going to see when working with the application?

[00:09:02] In those cases, you can start to present that information either as a guided flow or open discovery. Guided flows are much like a Consumer Reports type website. You arrange the information into related categories. In this case, we've got a wine CRM, and I've got key wines that I'm selling. I'm enabling my end user to pick wines and then explore the data. What is the income range of the wine buyers? What countries are we selling in? If we're going to run events, what are the weather patterns in the areas where we would run promotional events, so that we can schedule those?

[00:09:52] I can structure an exploratory experience that can allow an end user to start to ask about options, ask questions, get more information, and start to gain some of the benefit of making those comparisons. At the other extreme, you may just give the user a blank canvas with data and let them go crazy on slice and dice and filter and drill to explore all of the options on their own. It's really going to depend on your persona and how open-ended their questions are, and of course what their skills are, how analytically savvy that individual is.

[00:10:27] Now, again, the trick here is that the feature requests can start somewhat innocently. This can look like a pretty simple high-level dashboard with a few interactions, but very quickly, and especially as you get into the more open-ended discovery, you do start to get into some very complex feature sets: filtering, changing metrics, drill down, cohort analysis. There's a lot that can go on there, and it can cascade very quickly in terms of complexity. So you'll just need to balance that as you're thinking about implementing these types of discovery use cases.

Recommendations

[00:11:07] Now, the final concept that I want to talk about is recommendations. Very often both the inline insights and guided exploration can feel like a recommendation. In this category, I generally reserve it for instances in which, as a product team, I'm able to generate a recommendation for the end user based on one of three things. Either best practice: industry or subject matter expertise tells me in this situation you should do X or Y.

[00:11:44] Past practice: what has this user done in the past? And if I can, I actually want to say, what has the user done in the past that led to a good result? If I've got the data to do that, that's an even better recommendation on past practice. Where have we prescribed to a patient and the clinical outcomes were good, as an example? That's a good past practice.

[00:12:07] The third one is peer practice, and this is also known as benchmark. What are other people doing? What are other organizations doing? And for those organizations that are like me, are they getting better results, or am I getting better results? If you can help me understand why I'm getting better results or why they're getting better results, it'll help me change my own practices.

[00:12:30] Now, the presentation requirements here can be very simple. We've all experienced Netflix recommendations, and this is actually an example of past practice, and it's past practice with good result, because they know our viewing time. We can rate the movies as well. So they've actually got a way of saying: you've watched these kinds of movies in the past and liked them, here are some recommendations that we think you'll like. In this case, you can see I was spending time watching movies with my daughter, and they're queuing up more options for us to spend some quality time together, because they know what movies we liked watching together in the past. That's a good example of past practice.

[00:13:10] The displays, or the processing, can also be much more complicated than this. You can do alerts and notifications, and benchmark analysis can actually become quite a sophisticated set of analysis. So it really just depends on the level and the types of recommendations that you're trying to make. And frankly, this is where, even while the front-end presentation may not be complex, the backend processing on recommendations is where you may need to lean on a technical platform for things like AI/ML capabilities to generate those recommendations.

Build, buy or partner

[00:13:53] Now, having given you three high-level use cases to think about in terms of adding value for your end users using analytics, you definitely have to consider how you're going to get there. In full disclosure, I lead product at Logi Analytics, as I said, and we build development platforms that help application teams deliver embedded analytics quickly. So you can expect I'm going to be biased and tell you you should go buy a platform, but I'm actually not going to tell you that you need to go buy a platform. Not necessarily.

[00:14:28] As you're building out analytic capabilities, there are definitely times to take a build versus a buy or a partner strategy. When you have very targeted use cases, the simpler use cases, those inline insights, those absolutely in many cases can be built and maintained using some visual libraries, and your dev team cranks out the use case and you're good to go. Also, as you're thinking about an early POC, if you're trying to validate requirements, you clearly don't want to over-invest before getting some validation. Those are also great times to do some build on your own to validate those initial requirements and get to an MVP.

[00:15:16] Where I'm going to suggest you start to have some caution is as you're getting into the more complex and open-ended use cases. It's a combination of a couple of things. Sophistication of front-end requirements: how much interactivity, how specialized, do I need to define filters and share them, scheduling, collaboration, these types of very interactive, robust front-end features. And then it's also the backend processing: where do I need AI/ML, more sophisticated data access, search use cases, as well as graph database use cases.

[00:15:56] These types of complexities are where a platform is going to start to help you, especially as you're scaling to larger data volumes, larger numbers of users, as well as more product personas that will start to drive a higher diversity of requirements. Building these capabilities purely out of your development team, frankly, will start to swamp your roadmap. So if analytics starts to become a real key feature set for you, you're going to want to start looking for technology partners, either in a buy or partner type strategy, to help accelerate some of this and create some repeatability for you and your dev team.

Value for the end user

[00:16:42] The core idea that I want to leave you with today is that analytics can be complicated. There's a lot going on out there in terms of types of analytics; the algorithms and the technologies are always evolving. From a design perspective, it always comes back to value for the end user. The question that I always like to ask myself, to reground myself, is: can I help my end user know something that they didn't know before, that will help them do something different, that will create value for them? And can I do it right at the point of the action, in a very targeted way?

[00:17:27] I can help present them options and exploration when they're in a place where they're confused, they need relationships, they need more options. And when possible, I can give them recommendations to help them be more effective. If you use that framework, I think you'll find a ton of opportunity to get unmired [?] from the complexity of the technology and really find opportunities to create targeted value for your end users using embedded analytics.

[00:17:57] I appreciate you all spending time with me today. I'd love to hear from each of you about the applications you're building and how you think about creating more value in those applications with data-driven insights. You can find me on LinkedIn. Love to hear from you. Thank you so much.