Episode 56 Transcript

How to Leverage Data Analytics and AI to Improve PS Engagements w/ Tim Tutt

    Banoo Behboodi: Welcome to the Professional Services Pursuit, a podcast featuring expert advice and insights on the professional services industry. I'm Banoo, and I’m very excited that you've joined myself and my guest, Timothy Tutt today.

    Tim is the CEO and co-founder of Night Shift Development Inc., which focuses on helping organizations make sense of their data. Using his background in natural language processing, machine learning, and data analytics, he's created a productized solution called ClearQuery, which has a great tagline, analytics for humans. It's fantastic. I love that, Tim.

    Tim, welcome to the podcast. I'm very excited about this topic. I’ve been doing a lot of research on artificial intelligence and the context that it has in professional services and resource management for some of the other presentations that I have to do. I know how important data quality is and in anything that we do, whether it's as simple as predictive analytics or going as far as artificial intelligence. I’m excited to have you and excited to have this conversation. Welcome.

    Tim Tutt: Very excited to be here. Thank you very much for having me. Excited to chat today.

    Banoo Behboodi: Sounds great. Let's get started, Tim. If you can, please give us a little bit of background on how you've gotten to where you are today in the company that you've started, just a little bit more detail would be great.

    Tim Tutt: Absolutely. Well, I think you kicked off with a great introduction there. Like you noted, I've got a long background in large-scale data analytics and building search and discovery applications. A good bit of that was spent working inside of the federal government, helping them build large-scale data analytics solutions, drilling through and triaging the massive amounts of data to find those golden nuggets that really helped with making the right decisions for moving forward.

    One of my roles towards the end of my tenure working in the government primarily was to play middleman between my end users and their data. So I had access to a massive supercomputer. People would come to me and ask questions. I'd go run queries against the data store, give them an answer back. They'd have a follow-up question. I'd go run more queries; wash, rinse, repeat. That cycle just kept going over and over again.

    Myself and my co-founder, we both worked in the space, and we were part of a very small team that served hundreds of analysts and having to respond to a number of these things. We said, wouldn't it be nice if we didn't have to do this? Most developers will tell you if we have to do something more than once, we're going to automate it and never do it again. We had gotten to a point where we were doing a good job of automating, but there was still this, we were stuck in the middle, and we wanted to be working on more interesting problems.

    We took a big step back and started focusing on, how can we build a software solution to enable people to actually get that value on their own and then enable us to ask more complicated questions? At the time, we still had bills to pay and all that good stuff, so we kept working our day jobs but started working on product at night, which is why the name of the company is Night Shift Development. But our core product offering eventually became ClearQuery, which is designed to make it dirt simple for anyone to get value from their data, truly meant to be analytics for humans.

    Banoo Behboodi: I love that, analytics for humans. Let's talk about data quality. As I said, I've been doing a lot of research, and that's the critical component or foundation for analytics, AI, et cetera. If you don't have data quality, then garbage in, garbage out as they say. Tell us what are some of the challenges that are faced? How does one overcome those challenges around data quality to be able to get to the desired outcome in terms of analytics?

    Tim Tutt: Absolutely. You're exactly right. Data quality is the biggest issue that we encounter with most of our clients across the board. A lot of them have data that's coming in, and it's usually very messy. Things are not always spelled the same way. You may have, for instance, a column that is a set of states. I've had cases where we've loaded data up and ran analytics, and it's like, there are 85 unique states in the United States. Well, we know that's not right. It turns out it was just because someone said, VA uppercase and VA lowercase, or someone said Virginia instead of VA. You have the different types of data that's been fed in there from a variety of different sources.

    This cleaning the data becomes very important. We have so much messy data that if you wind up getting messy data, you get this garbage in, garbage out. That's usually the case. You're not going to get the right answers that you're looking for unless that data gets cleaned from the beginning; normalized, cleaned all of these different things. That's sometimes a very tedious process. One of the things that we work on is we try to help our customers clean that data as part of a data engineering exercise if they don't have already clean data. But you're absolutely right. Again, this data quality thing is the core beginning before you can even get to a good analytics solution.

    Banoo Behboodi: An extension of that aspect is the definition of quality data because there is the piece that you explained where you have 85 states where that's clearly inaccurate, and it's from misspelling, miscapturing the data. But also, when you go and extend that to artificial intelligence where it's continuously learning from data that's fed to it, it's directionally getting the wrong answer or wrong direction because of what the intelligence is using to learn from.

    Can you tell me a bit about that piece of the complexity? The example that you brought up, the state part of it, is easy to capture anomalies and clean up. What if the facts that are feeding the system, the predictive, are ones that are not necessarily going to get the answer that is the accurate answer or the right predictive answer?

    Tim Tutt: This is really boils into a bigger issue of understanding the domain and the problem set that you're working with. Whether you're tracking metrics for conversion rates or, for instance, price and sales. Sales is a very simple example. I'm trying to look at sales trending, where things are going over time. That's a pretty straightforward one.

    Now there's this question about cost and do we include taxes and what's our net profit? How do you actually calculate those metrics? If you're only feeding in gross revenue and not feeding in what your net revenue is, you're not going to be able to accurately predict your overall profit over time. If you're looking at conversion rates and you have different groups defining what a conversion rate is in different ways, that's going to be a challenge. If you have one group saying conversion rate is they filled out the form and another group saying, no, conversion rate is they actually bought our product, that's a very different thing. If you have all of this data feeding in to the same location, your machine learning models are going to pick up on all of that data and make the wrong assumptions because everyone's looking at this data in a very different way. Defining that domain becomes a very important aspect of it and making sure that everyone that is involved in the analytics process and understanding those business outcomes is on the same page.

    Banoo Behboodi: Do you start then when you take on assessing data quality with the output and the answer in mind or the results that the client is seeking as an end point, and then you work your way back to qualify whether the data has the quality needed?

    Tim Tutt: Yes, absolutely. It's always starting with the business outcomes. That's the number one thing when it comes to analytics and finding even the right questions to ask of that data is knowing what the business outcomes that you're looking for are.

    Banoo Behboodi: Yeah, that's perfect. Now, as you know, our audience are professional services professionals. Bringing it more in the context of professional services, whether it's embedded or pure consulting, what are some of the applications of data and being able to drive predictive analytics and artificial intelligence as we evolve?

    Tim Tutt: Yeah, absolutely. If you look at professional services organizations, a lot of the times, they are really driving at the end of the day, how do we find the right answers? How do we find the right business outcomes for an organization? When you're an organization and you're seeking help from a professional services organization, you're really looking for someone that can help you drive to those business outcomes faster, someone that knows the ways to ask the right questions of the data.

    There's a lot of different ways this applies, whether we're looking at, let's call it, cybersecurity data. I've got a group of folks that are looking at red teaming my systems. They need to be able to collect the right data points to understand the effectiveness of those campaigns against my systems. If I am looking at, again, let's go back to the sales and marketing examples, in each one of those cases, I need a good metric to determine what the actual ROI and the value is for those professional services organizations of those professional services organizations in terms of getting to that business outcome.

    For the person providing that professional service at the end of the day, using data analytics can really help you drive those outcomes and show those organizations this is where we've been effective. Here's what we've been able to drive for you, and here's where we can help accelerate in terms of your business and what you're really trying to get to from the outcomes perspective. Using things like predictive analytics to determine, this upcoming season, we know we had this certain product SKU sell more times than most others. We need to make sure we're stocked up. We were sold out. People were trying to add this to their cart, and they weren't able to do that because we didn't have anything in the inventory. Well now that's something that we can use that data to help you make the right decisions on what you need to stock up on.

    If you're looking at things from a broader perspective in terms of HR, for instance, helping to understand attrition rates and things like that, understanding the raw data points or what it is that's impacting attrition rates can help you drive the right decisions for what needs to change in that organization. Here are the things that you need to adapt to. Data really helps to drive the key decision making and the recommendations that a professional services organization might make to a business at the end of the day.

    Banoo Behboodi: I think if we look at the professional services, the underpinning resources or underpin a resource for a professional services delivery organization, resource is the asset and the revenue generation. I think it's important to reflect that with the right data, what the organizations can get to is more the predictive analytics of let's take something like skill sets and skill and development strategies for individuals, understanding what projects they've worked on, which projects they've had higher utilization on, using that information to then predict what is the best project therefore for these individuals to be optimal on, where are some skill set gaps based on what they have done and where they have not outperformed as they usually do, and therefore be able to have a development plan based on that.

    I think these are all the areas that are the power of using the data to help us guide and make sure that our workforce and our talent is positioned for success. Therefore, if they're positioned for success in a service delivery organization, that's going to inevitably help meet the customer outcomes. It's fantastic. I think it's critical. I don't know that every organization is there, but I think everyone is looking to get there.

    Tim Tutt: Absolutely. One thing I'd add on is even stepping just a little bit beyond data, and data analytics in particular, if you look at all of the capabilities that you have in terms of AI and machine learning that are out there, from a professional services standpoint, the other big benefit is it accelerates your time value. You move away from I have to do this very manual thing to I can accelerate the time it takes me to deliver an asset because I can leverage various pieces of technology to get me there, which also enables you to change maybe your pricing rates. Maybe it's instead of an hourly bill rate, I can give you a fixed price because I know predictably this is how long this is going to take me because I'm using the right tools and capabilities to drive these things.

    Banoo Behboodi: That's an interesting angle that I wasn't thinking about is further automation of this is more in terms of embedded service probably for software companies with service delivery where this technology can help automate and productize potentially some of the more manual labor-intensive components of the delivery, making it efficient and therefore, driving quicker time to value for a customer. That's a great point.

    Now you have a fascinating blog series that talks about the seven key data strategies. Can you walk us through those at a high level? What are those seven?

    Tim Tutt: Yeah, absolutely. This is actually a blog series that just kicked off. We've released the first one, and the second one is actually coming here later this week. But this blog series is based on a set of strategies that we've identified for chief data analytics officers, data leaders across the board.

    With the key seven, I'll run through the seven, and I'm happy to dive into each of them a little bit. There's this data discovery strategy. This one is really focused on how we find the data that an organization needs. How do all of the users and consumers of that data find where it even exists across their organization?

    The second one is this access control strategy, which is really looking at how do we secure the data to make sure that we're not sharing things with people that don't need to have access to it. Data life cycle, a lot of people tend to think data needs to live forever, and that's really not the case. Sometimes I get data coming in real time, and I don't need that forever. I just need it for a certain period of time. How do we age that off appropriately and keep the metrics that matter to us over time to save on costs for storage and things like that?

    The fourth one, we move into this real-time processing of data. Everyone looks at data in a lot of different ways, but sometimes knowing how to look at data as it's coming off the wire in real time to make decisions immediately is important. That includes things like alerting or automating events based on certain things coming into your system, whether that's coming off of a real time from my website who's putting things in their cart? How do we retarget and send ads to them? Or how do we make decisions from a cybersecurity perspective based on the events that are occurring on our system?

    The fifth one is probably one of my favorites is the state analytics and democratization strategy. How do we help make sure that we get data into the hands of more people? How do we make sure people are able to analyze and ask the right questions of that data? The sixth dives into a little bit of this data storytelling aspect. How do we communicate the value that we're getting from that data to the senior business leaders that are making the right business decisions overall?

    The last one is this AI and ML for data. One of the things I'll say is not every organization needs artificial intelligence and machine learning. Most aren't even ready for it. We talked about this at the beginning. Data quality is such a big issue. If you don't even have that solved, you shouldn't even be thinking about the AI for ML, AI and ML for data, but it is something that you should consider at some point once you've gotten to these things and how does this fit in our organization, what types of predictive things that we do want to do, so understanding how to actually leverage those.

    Those are the seven key strategies, and the blog series will dive a bit deeper into each of those and tools and recommendations for how to drive into those.

    Banoo Behboodi: A couple of the ones I want to dive in, I love the last one. I think it was the storytelling. That's the critical part. You have all this data. You have a question you're after as an analyst, and there is a set of outputs that you get as a result of the query. Now you have the outputs, and it's telling you something. How you convey that to something that's actionable at the executive level so that you can get sponsorship and get momentum around what the data is telling you is fascinating. What are some of the tricks on being effective on that piece?

    Tim Tutt: Sure. Absolutely. I'll talk a little bit about one of the ways that we do it with ClearQuery. Everyone builds dashboards, and everyone talks about dashboards being this key way to help tell the stories. The problem with dashboards is you sometimes wind up with lots of charts, lots of visualizations where there's too many metrics. No one knows what any individual KPI means, and there's a lack of context around all of these things. What do these things actually mean for me and the business outcomes?

    We actually have a capability that we call Insights Canvas, which allows you to build a presentation based off of the analytics and the analysis that you're doing inside of the application. That helps to really go and present to your senior leaders where you're saying, here's this metric, but here's the context around this metric and why this matters and what this impacts. Being able to tell that story around the data, how different certain values are from others, and why that's impactful, what are the key drivers for something actually happening in an organization, it's immensely important. I think this data storytelling strategy, it's not only just the tech behind it, but it's also the how do we communicate as analysts, as business leaders, as professional services organizations? How do we communicate to those decision makers? Here's what matters, and here's why you should care about it.

    Banoo Behboodi: And these are the actions you should take as a result, so the recommended actions. That's excellent. The other piece I wanted to dive into a little bit more is the security component of that. We've talked about the data quality aspect, but obviously, one of the things that everyone's raffling around with all the data that's available is that data security, data privacy, and cybersecurity, all the aspects that can actually impact things. Tell us a little bit more about that piece.

    Tim Tutt: Sure. When we talk about data security in particular, we're looking at a combination of things. There's the, how do we make sure all of our data is secured from outside parties? That's phase one. Do we have things locked down enough so that we're not going to get hacked? If we are hacked, we limit the amount of exposure of data that is potentially leaked. We’ve now hit this space in the world where if someone really wants to get to you, they can probably get to you. Now the question is, how do we minimize the risk to our organization?

    What that means is what data are you storing? What data do you actually need to keep? Do you need to keep credit cards around, or can you use a third-party payments processor to manage that aspect for you so that you don't have to worry about that for your organization? Do you need to keep PII, things like addresses or even social security numbers? If you do need to keep those things, how many people in your organization actually need access? How do we limit, even within our organization, the people that can get access to those data points, but also enable enough of the access to the data so that the people can get access to the data points that they need. What that means is you're actually not just looking at low-level security, you're also looking at field-level security because sometimes people need to do analytics on a data store, but they don't need access to every given field there, and you want to reserve those things.

    Again, the two big aspects are the external protections and then second is the internal protections, and building that strategy for the internal protections is one of the things that I think we find lacking in a lot of organizations. Sometimes we get to a point where people say, you know what? It's too hard, so we're just going to block off access to four people. That actually impacts your ability to democratize the analytics and impacts the ability for people that understand the business outcomes to get access to what they need to actually analyze and get to actionable insights on it. You just have to have a strategy that helps to encapsulate all of these things from a security perspective.

    Banoo Behboodi: When I think about a professional services organization and what they need to focus on in terms of data capture, to get quality data, you've got to have defined tools and processes to be able to capture that data quality, or you have to continuously roam through it and cleanse it, which is not efficient and effective either. The front-end processes and tools have to be built, designed with that end in mind so that you are collecting, over time, the right data.

    Now I would imagine the more disparate your solution is, so if we take a professional services organization, you have sales coming in through your CRM. You then have potentially your resource managing in an Excel, or you have something like Kantata, a PSA that's helping you with that process. What do you see and what are some of the challenges to that data quality on the whole technology infrastructure that a company may have and any suggestions you have on that piece?

    Tim Tutt: At the end of the day, it really does boil down to what are your end goals when building an application? Always start there. Anytime you're bringing in any technology, anytime you are building new technology, start with what the end goals and the business outcomes are. If you understand what those business outcomes are, you can walk backwards to this, okay, that means we're going to need these data points to make the right accurate decisions. If we need these data points, that means when we're building a new application, we need to ensure that there are certain dropdowns.

    I have a client, for instance, where they're using our capability for analytics on a lot of old data that was not clean. It was not dirty, so we started building out an application front end for them to collect that data in the first place. What that application actually looks like is a lot of dropdowns. We removed as much as we can a lot of the hand-typed manual entries so that we can actually get to the analytics that the organizations are really looking for. It allows that whole process to flow all the way through in a more effective way. We can alert on the right things. We can analyze the right decisions and help them ask the right questions of their data with that.

    It really starts with the, what are the business outcomes you're looking for? Walk backwards, and then every time you're coming in with a new piece of technology or you're starting a new technology product, understand what those outcomes are and start building out, focused on that end goal and what those business outcomes are. I think that'd be the biggest piece of advice that I think can help with a lot of those things to begin with.

    Banoo Behboodi: Yeah, design with that outcome in mind, but also, I would think design in a way that you can monitor adoption. If you built this overengineered solution that users can't use, the data is not going to be there irrespective of how you've designed it because you're not driving adoption.

    Tim Tutt: That's absolutely correct.

    Banoo Behboodi: Thank you. It's been very informative, definitely appreciate your time. But before we close out, I do ask, as you know, typically about a mentor or a favorite book, if you can tell us about a recommendation, a book because I think you have one that you use as a reference continuously throughout your career. That would be very helpful.

    Tim Tutt: Absolutely. That book is called The Hard Thing About Hard Things. It's by Ben Horowitz. It's a phenomenal read for any entrepreneur getting started, but I think any business leader getting started. Even if you're not a business leader, there's also aspects of it that help you understand decision-making processes and how things operate inside an organization. It's something that I use constantly as I'm thinking about how we build our business and grow the culture that we are building internally, the night shift.

    One of my favorite quotes from that book is, “Take care of the people, the product, and the profits in that order.” That is a guiding philosophy for us and how we build everything and how we operate as an organization. We take care of our people first, make sure everyone has the right culture and has all of the things that they need from a benefits perspective. We take care of the product. We make sure the product is what our clients need, and it's constantly designed for being as simple as possible for any user. The profits follow after that. That's the way we look at it.

    Banoo Behboodi: Perfect. Well, Tim, thank you for making the time, as I said, and it's been fun. It's been fun talking through some of the challenges around data and what you can do as a company to assist.

    As always, we'd love to hear from our listeners. If you have any questions for myself or for Tim, please do reach out to podcast@kantata.com, and thanks for joining us today.

    Tim Tutt: Thanks so much for having me. You have a great one.

    Banoo Behboodi: Yeah, you too.