Real-Time Student Data Orchestration: WGU's Databricks and Tealium Integration

Summary

Western Governors University (WGU) migrated from Excel-based reporting to an enterprise lakehouse on the Databricks Data and AI platform, handling petabytes of student, financial, and assessment data for 201,000 active students.
WGU's Tealium customer data platform integration enabled three high-ROI use cases: ad suppression saving 5 percent of outreach spend, personalized program recommendations increasing applications 22 percent, and a Learner 360 system tracking 54 hero moments for real-time student intervention.
The medallion architecture, combined with AI-driven transcript analysis and segmentation, enables WGU to detect at-risk students before they disengage and deliver personalized experiences at the moment they matter most.

Real-Time Student Data Orchestration: WGU's Databricks and Tealium Integration

Higher education institutions collect vast student data across siloed systems but rarely activate it in real-time. Western Governors University, working with Tealium and Databricks, transformed disparate student records into actionable insights delivered in the moment. Narendra Pandya, Distinguished Engineer at WGU, shares how real-time data orchestration powers personalized student experiences and measurable academic success.
See how WGU migrated from Excel-based reporting to an enterprise lakehouse running on petabytes of student, financial, and assessment data. Learn the three use cases that drove ROI: ad suppression saving 5 percent in outreach spend, personalized program recommendations increasing applications 22 percent, and learner 360 tracking 54 hero moments for real-time intervention. Explore the medallion architecture, Tealium's customer data platform integration, and strategies for detecting at-risk students before they disengage.
🤝

Chapters

FAQs

What is Western Governors University and how large is its student data platform?

Western Governors University (WGU) is a not-for-profit, competency-based online university founded in 1997 by 19 state governors, currently serving 201,000 active students pursuing bachelor's and master's degrees. The university operates a petabyte-scale enterprise lakehouse on the Databricks Data and AI platform, having migrated from Excel-based reporting to unified student, financial, and assessment data.

How does WGU use Tealium with Databricks?

WGU integrates Tealium's customer data platform with Databricks to activate student data in real time, enabling personalized experiences delivered at the right moment. Tealium ingests behavioral signals and routes them through the lakehouse, powering use cases such as ad suppression, personalized program recommendations, and real-time learner intervention.

What results did WGU achieve with personalized program recommendations?

By using the Databricks Data and AI platform and Tealium integration for personalized program recommendations, WGU achieved a 22 percent increase in application rates. Additionally, ad suppression saved approximately 5 percent of outreach spend by preventing targeted ads from reaching current or recently engaged students.

What is WGU's Learner 360 system?

Learner 360 is WGU's real-time student intervention framework that tracks 54 hero moments — key behavioral signals indicating when a student may need support or engagement. The system enables WGU staff to act proactively before a student disengages, combining the medallion architecture's data quality with Tealium's real-time activation capabilities.

Full transcript

[00:07] All right. Well, I think we'll go ahead and get get started. Thank you for for joining us. I know getting from the keynote to sessions right away is is kind of a it's kind of a challenge, but we're we're excited. We've got a lot of content. We'll have probably some time at the end for
[00:24] even some Q&A. But before we get started, we should probably do some introductions. Yes. So, I'll go Hello everyone. My name is Narendra Pandya and I'm a distinguished data engineer working for Western Governors University.
[00:39] If I were to sum up what I do for WGU, I'd like to show this. This is what my job is. Free the data. That's basically what I help do for our university. Zack. Uh and my name is Zack Wenthe. I'm
[00:55] customer data evangelist for Tealium. I'm not distinguished in anything. So, we're going to let we're going to let him do most of the talking, but you know, I'm actually pretty familiar with with WGU not actually because they're a customer my my wife actually got her MBA in health
[01:13] care from there, but for some of those in the room who may not be as familiar with WGU or you know, all of the the stats behind them. Why don't why don't you give us a quick rundown before we dive in about about WGU? Absolutely. Thank you, Zack. So, Western
[01:30] Governors University, we are not-for-profit competency-based online university. Back in 1997, 19 different governors from the western states of of the country came together, both side, Democrats, Republicans came
[01:48] together and they said that for our workforce, we need a university that is not elaborate on the infrastructure but focusing on competency-based education and that is when WGU was born.
[02:05] Our main mission is to provide or to change the lives for better by providing opportunities, pathways to opportunities. That's basically our key. Our core center is our student. Education should be affordable and
[02:22] accessible. That's one of the core thing that we work on. And if you are not familiar, these are the numbers that you can see on the top. We are one of the nation's biggest online university. We have currently 201,000 active students pursuing their masters,
[02:38] their bachelors and we have 502K of alumni. We just completed the half a million mark last month in the month of May. One of the largest, as I said, 100% competency-based. What does competency-based really mean would be
[02:54] education at your own pace. These are working adults who would want who have full-time job, full-time family and they also want to pursue their dream of pursuing a degree. So, how do we work with this is the affordability comes into play by
[03:09] term base. So, we have a all you can eat buffet in a 6-month term. Basically, you pay for a term and then you complete your masters or bachelors. So, if you want to pursue a MBA, you can join WGU and then you can say that okay, I'm going to pay for the first term and if
[03:24] you are fast, you have experience at your workplace, 10 years of experience managing team and stuff, you would say, "Oh, I know these things. Let me go, give the assessment, pass it, move on to the next subject." There are people who have There are students who have done their masters in 6 months and even in 1
[03:40] year. So, basically, it's making it fully affordable and available at the time. So, obviously with that many active students and and, you know, uh over half a million degrees earned, that's a lot of data. Yes. Um, and the and and being online. So,
[03:56] why don't you let's start with kind of the data journey and kind of the the history of WGU and how you kind of developed and then we'll get into some of the specific use cases as as we as we go through this. Absolutely. So, with this, um, one of the statement that our president, Scott
[04:11] Pulsipher, he keeps referring to is we want to be the most student-centric university in the entire world. And to tag along with that, I said, "If we want to be the most student-centric, our approach towards our data should be data-centric. Data should be our primary
[04:27] and permanent asset. We need to create products of data that everybody can consume and rely on making decisions. Being online, everything is collected into form of data. We have data flowing in from different systems." Now, looking at this journey, it's it's
[04:44] very close to my heart. It is the journey that I lived at WGU. I joined around 2012 time frame. And this is something that everybody can relate in this room. Everything starts with Excel. So, the first thing we did was we had 17
[04:59] different Excel files attached to an email sent on the 10th of the month. 10th is our census date. So, on the 10th of the month, it was sent as an email attachment to our president and on to our senior leadership. When I joined in, I was asked, "Okay,
[05:15] we want to get out of Excel." Because, you know how the Excel story goes? My Excel greatest. Right? People change numbers. This So, all right, how do we get everybody on the same page? So, the first journey, the first task that was assigned was to automate this 17
[05:30] different Excel files to be generated as a single PDF. And that's where the journey started to build our first enterprise data warehouse. So, as you see, our when when I first reported the data, it was about 38,000 students with about only 20,000 graduates. So, moving
[05:46] from that Excel-based solution to a enterprise data warehouse, we were tight we started with a proper database approach and everything was great, but the growth that we were seeing during that time kept us challenging. We did all sorts of expansion vertically,
[06:03] horizontally to to shrunk to do work everything on our data window. Midnight we would start the processing and we wanted everything to be done by 6:00 a.m. That window was getting shrunk day by day. We were having challenges running these
[06:19] systems. And that's where we thought that, "Okay, we've learned a lot in during that 4-5 year time period." We said, "Okay, vendor lock-in, we had enough of it. We want to go into a system where we should be owning our own data, we should be owning our own business logic, no proprietary things.
[06:34] And if you don't like the vendor, we can spun up and do our own stuff." And that's where first we came to the Spark Summit here. I came to for the Spark Summit in 2016 and we adopted using Spark based solution. All our data was in S3 and all our code was in Python. We
[06:50] moved migrated everything from the proprietary system into Databricks area or back in the day the Spark area. And that's where the first EDW 2.0 was launched. Eventually, we moved and migrated everything into Unity Catalog,
[07:06] but again, as we were going through that journey, we came across steps like Delta Lake was one thing that Databricks announced and we were like, "Okay, no, we are again going back into that proprietary loop and we might just get sucked into that cycle." So,
[07:23] lucky enough that Databricks every 6 months down the line they kept making things open source and now they do everything open source. So, we are not locked down into a vendor specific thing. We if we want we can spin up our own clusters, we can do our own stuff or open source data could be built up.
[07:40] Our current journey which we have scaled up to is the enterprise data lakehouse. The enterprise data lakehouse is currently hosting over a million records of our students, our prospects, and all sorts. So, this entire journey is I would call it not
[07:56] over, it's just the beginning. We We are ready now for helping and to do more use cases which are more student-oriented. So, yeah, having a scalable single version of truth is what I'm depicting over here. Sure. And obviously, before we get into those use cases, you know, everybody
[08:12] likes the architecture slide. So, let's talk a little bit quickly about what that architecture kind of looks like, and then we'll we'll we'll go from there. Absolutely. So, one of the key thing, as I said, is treating data as an asset, as a product. How do we really make data
[08:27] look like a product? At the end of the day, data when we talk about it in the simple layman terms is a table. So, how do we organize those tables? So, we relied on the medallion architecture. Even before we went into Databricks, we used this medallion architecture where we created our flow of bronze, silver,
[08:43] and gold tables. So, down I've named uh one example here, which is with a reference to performance assessment. Within WGU, we have two types of assessment, objective assessments, which are your objective-type questions, and performance assessment where you write papers and thesis. So,
[09:00] when we first started, we had Taskstream as one of our system of records that would provide us the performance assessment data. It was XML files, huge-size XML files sent to us. We processed those files, loaded it into bronze layer, make it into silver, and then finally made a curated table for
[09:18] all our assessment reporting. And that was our gold standard table. But, as I said, tools, technologies keep changing. We adopt new technologies, we decommission old technologies. So, over the period of time, we came across Emma
[09:33] and Aurora-based solution. Now, the challenge was by that time on the gold layer there were hundreds of reports already built. There was dependency, there was reliability, everything was established. So, how do we how do we continue to support the same thing? And
[09:48] that is where the concept comes up like how to treat this as a product. So, we were able to build that with the layering approach where we could ingest the two bronze layer table data into the silver layer and then continue to proceed with the same data set in the same format. Similarly, in 2024 we had
[10:07] said okay, again we had a homegrown solution that we built which was a database driven or Postgres based system. And again, the flow as you can see, it helped us to keep our downstream dependencies pretty stable. And this layering approach, I'm I'm giving an example of one such table. We have
[10:24] a bunch of tables uh we have a master table for student records which shows everything about their journey. And that is used by like 80% of our analysts, our data scientists, and anyone who wants to know about our student journey. So, this medallion architecture layering has has
[10:41] consistently helped us with our with our approach to treat data as a product. So, obviously you your your shirt says free the data. Yes. But we know data is not free. So, obviously the other part of your your role is then managing cost of all of this. Right. Before you even get into use cases and how do we
[10:57] actually turn turn value into this? It's it's managing the infrastructure. So, give us a kind of a breakdown of of what you guys are doing. Absolutely, Zach. So, one of the key thing that I've learned and in the process that we have built such a massive data lakehouse was cost.
[11:13] That was something that was constantly growing. And when we have new features, one of the thing that we our leadership always promote is fail fast. So, we we would say okay, try out. But then we realized that okay, this is costing us money. And one of the important thing as I said
[11:29] in the beginning, which was we are a for profit. So, for us, every single student dollar is sacred. We would want to use the best of that dollar for our students. And so, one of the thing that we identified is, since we are into education, making
[11:45] people literate helps. So, we started with making people aware and literate. Literacy actually empowers people to be more responsible, to be more vigilant about what they are doing. So, what we
[12:00] do is, with our literacy drive, we have meetings on a regular weekly cadence, where we observe all our data, all all our cost. And we have like the AWS cost, Databricks cost. We have dashboards created in ThoughtSpot, Tableau. A lot
[12:16] of dashboards now Databricks has also provided, which gives us cluster-level detailing. And And the tagging that we do, it helps us to track at the project level. In individual project level, what's the Databricks DBU cost? What is the AWS cost? We can combine it, because
[12:31] the same tags that we are using in Databricks are flown into the AWS side. So, all these monitoring helps us to keep track and help us to prioritize the data in real time. Got you. And then obviously keeping an eye on it, too. Yes. So, as I was saying, these are some of the dashboards. It gives you all by
[12:48] SKUs, by clusters, by by warehouses, by jobs. All that is possible by tagging. And then the next level that we are doing is budgeting also. So, we are applying budgeting constraints. We are applying project-wise allocation of data. The next thing that we are also doing is we
[13:05] are doing chargebacks. So, we are preparing all our data to be tracked and also made available to the end user. All in all, in the process to be make people literate, to make them aware that, okay, this is where, even if you write a simple select star from a table name,
[13:21] that is cost. And you should be knowing that is it 25 cents or it is 25 dollars. So, that's the level of recommendations we provide. And then we also have within our data engineering team different types of health checking and chargeback recommendations that we come across.
[13:37] So, obviously it's very very foundational. It's taking you some time to get here. And and and now we're going to kind of switch into the use cases. But before we do, you know, obviously let's talk a little bit about the challenges and why what drove this modernization. Yes, absolutely. And we have data. If I
[13:52] were to make a statement, data is in action is what is helpful. If data is not in action, it is just cost. It is just lying there no purpose. So, we had systems various systems student system. We have finance system. We have
[14:08] assessment system. We have survey data. All of that data was siloed. The main objective what we wanted to fulfill with building a data lakehouse was to bring all of that data together in one place where it can be tied and utilized. So, some of the challenges that we
[14:24] we were facing mainly were siloed systems. We had issued with like insights were delayed. And I'll give you one example when we when we deep dive into it. And then it doesn't work every time to have one one size fitting everything. So, there were challenges
[14:40] that we worked towards the process and then came to solutions. Right. Right. And that's obviously where the Tealium comes in, right? Because you you you know, we're looking at ways to kind of change the engagement model. So, for those of you who are not kind of familiar with with Tealium, so we are a
[14:56] customer data orchestration platform. We help WGU kind of sit you know, you know, as part of their stack to help both collect that that that customer that student data from all of their systems and and get that into their their kind of source of truth or
[15:12] that student 360 and then ultimately turn that into from the business teams you know, value. Whether that's through engagement and we'll walk through some of the ways that they did this, you know, ad suppression and and whatnot. But, typically, this is the question that always comes is like, "Well, where does that fit?" before we get into the
[15:28] use case. So, the obligatory architecture slide. But, let's talk about how WGU actually uses this. So, so Uh if you want to run down kind of some of the data and and the integrations that you guys have already built, and then we'll talk about then what what that what that provides.
[15:43] Absolutely. So, the very first use case where Tealium come came to our our attention and we adopted was the very first use case was the advertisement or the the outreach dollars that we were spending. Every time we have such a big student
[16:00] base, learner base, and employee base, every time the paid search would show up whenever somebody searching for WGU, it would show us WGU and we were paying that. So, the most important use case that we started was with suppression. As I said, every single dollar of student
[16:15] counts for us. So, the student suppression or or the not student suppression, but the mainly the advertisement suppression or outreach suppression was our primary use case. We were losing a lot of dollars for just sending advertisements or showing to
[16:30] people who already are part of WGU or know about WGU. And so, when we first went into that use case, we saw that, okay, there was about 5% of saving that we could gain by suppressing ads from our students or from our active alumni
[16:46] or from from the existing employees. Now, how that was possible is previously it was a manual approach. We would just take the data and send it to the vendor for the clickstream information, and it was done every 3 months. By the time the system would get the data, it would be
[17:02] stale, and it would not know about the multi-device approach that the student might adopt. And with Tealium, we were able to do that, change it into a real-time solution. So, we feed that data into Tealium, and then Tealium is is to identify ahead of time that okay yes,
[17:18] this is an existing employee or this is an alum or this is a basically our existing prospect. So, you don't need to show them branding examples. You want to focus more on the other other side of it. So, that was our primary use case which gave us a saving
[17:33] of about about 5% of our cost overall which showed us the return on investment for for the Tealium product. Right. Right. So, then you have see a real-time customer data or or student data flowing into native tables in Data Bricks. All your interaction data, your full
[17:49] historical view. Yeah. Uh and then you have, you know, as as you go now you can start to drive interventions and and you talked about that that first use case which is the the suppressing, you know, the paid ads from students and staff and and I think uh I want to reiterate what you said. So, that first use case paid the ROI of the
[18:07] platform. And so, everything after this now becomes an incremental, you know, incremental opportunity which is amazing. Yes. True. Um but you didn't stop. You didn't stop. That was the easy one. That was like, okay, we got that one out of the way. And then so, next, what was what was next? So, the next big challenge was as these
[18:23] numbers were growing, we could see that okay, our KPI metrics, even like drop rate for instance, our objective as a not-for-profit is we want people to be successful. We want them to come in and then finish the line and get out of it. So, because that's
[18:41] where the value comes in. The education value comes in when you finish the when you cross the finish line. So, so basically to adopt that particular thing, we want to optimize the messaging. Okay, most importantly we saw that a lot of people come a lot of
[18:56] students who come in or learners, they're not clear on what they want. They might be, oh, I might start with a cybersecurity bachelor's and then within a month or two they might realize, no, this is not for me. I want to probably just go for a data science thing. So, how can we optimize our messaging so
[19:13] that we can help with our students to make the right decision? Personalizing. And that is where our use case came up with regards to like personalizing the messaging, personalizing and outreach to our upper funnel, which is our lead information, and to our students. So,
[19:30] and timely because you it doesn't make sense if you if you reach out to them after a week or after 3-4 days because this is real time and since we are online people want to make decisions in real time mode. So, that's where if you see here we have we've started showing
[19:46] relevant messaging right in the portal itself. And this was all again possible with the help of Tealium because Tealium could get all our customer data combined together into action. And obviously with your student profile, you know, these are busy working adults, they have families. So, the the the time
[20:02] that they spend interacting with WGU is just one small part of their life, unlike a more uh we'll call traditional student, right? So, obviously doing that in real time and not after the fact or a week later or 3 months later like the the ad expression is is is is is critical. Absolutely. And it had a big impact. Did you have
[20:18] you drove some some significant numbers? Yes. So, basically as you can see, since we were able to focus and provide correct guidance to our learners, we could see that the application rate in the upper funnel increased by 22%. So,
[20:35] people were upright when they come in, they don't need to see 120 programs of WGU, but when they come in, okay, your focus, your intention is technology. Let us show you technology related data. No need to show you something with regards to our
[20:51] our business programs or teachers programs or nursing programs. So, that's where you scope down, narrow down to okay, here are the 40 programs that are related to technology. And then on top of it you are into data or you are into management. So, we'll show you on those.
[21:06] So, that kind of optimization has helped us to provide more proper one-on-one guidance to our students, right? And not just on the the the application, but then on success, too. Yes. Yes, overall, if you make the right decision at the upfront where you have
[21:22] chosen the right set of program that you want to be in, the drop rate decreases. Drop rate is one of the major concern that we had as a as a team as a as WGU is because people start, but they don't finish. How can we help them?
[21:37] As the number grow, we could see that okay, our percentage of drop is still 3% or 5%, but when it was 50,000 students, 3% was a smaller number, but that same 3% when it comes to 200,000, it's a big number. And every single individual is a
[21:54] is a human life is is individual who for some reason couldn't proceed. So, how can we help them to be more successful? And that's where our we we worked on these data frames to be made available ahead of time. Um a classical example that I can quote here would be
[22:10] there was an instance and it will come down later on. So, there was an instance where a student all of a sudden decided to drop or wasn't available. And then the mentor didn't knew for a few weeks. And then when they reach out, they said, "Oh, basically we had I had a financial
[22:26] problem. I couldn't proceed with making my payments and that's why." And then the mentor says that, "Oh, I'm so sorry we reached out late to you because we do have financial programs. You can You can declare a financial hardship or and you can you can pay it later. All sorts of
[22:41] options are possible, but the problem was it wasn't notified at the same time." Right. Or the right time. And so, as we said, data, we had a lot of data, but is that actionable? That is what we have we got help with Tealium. Yeah. So, obviously intervening in the moment as opposed after the fact. And
[22:58] so, obviously as you now think about this long-term journey in this this you know, the whole that that that whole student now that you've collected all this data and you're tying it together. One of the things that you guys are doing is that that interaction or that that intervention as students are. So, you want to maybe
[23:14] talk a little bit about kind of that that momentum. Absolutely. So, when this use case that I just talked about or particular student getting impacted with some financial hardship everybody was like okay, no, this is wrong. We we want to fix this problem
[23:30] and we just know about one such problem. There could be many such problems. The other possible problem we we figured out was like okay, there was a student who was properly doing all the stuff and then all of a sudden was didn't appear for an exam and then when the mentor reached out he said that oh, my laptop
[23:47] broke and I couldn't proceed with it. That's life, right? Everybody have faces this problem. They may not have internet, they may have some natural calamity or anything. So, how do we help them support them? Because of the timely intervention, the mentor was immediately able to say that
[24:03] okay, not a problem if your laptop broke. We will ship you a loaner laptop for a month. Finish your stuff. And it's pretty easy for us to like okay, send a laptop to the student. The student will work through it. When they are done they will send back. But how was that to be made possible and that's where the
[24:18] learner 360 came into play and we started with a very simple red yellow green RYG model. That was one of our primarily model that we we implemented which was keeping track of all the student interactions.
[24:33] The student interactions are in the form of when the student open a page, they click on things, they go to a course, they click read the course or they do a pre-assessment or they give an exam or they call to a mentor or they they ask technical questions to their to their
[24:49] colleagues. So, all these interactions are tracked and these interactions help us to keep track of okay, what is the progress of the student? At the beginning of the term, a student is defined that okay, he's going to work on 15 CUs, credit units, for the for the 6-month term. And if there is like 4
[25:06] months passed and if the student has not completed even a single credit unit, then we it's immediately an indicator that okay, now it's turning green, it's turning into yellow. And to make it even more effective, we do it more on a daily basis. So, every single day we track it
[25:22] and then the indicators will immediately put the mentor in charge saying that okay, I can see that in the last 7 days you haven't got an opportunity to study. Is everything okay with you? And why would they see it? Because the indicator is showing them red.
[25:38] That means that if you don't get yourself on track, you may lag behind and you may not complete what you have promised for or you you want to complete for. So, to help them with these things, we described all of these things as hero moments. We call them as hero moments.
[25:53] There are about 54 of these points that we created in our student 360, which is hero moments. Data is there, but now how we make that actionable? And that is where we were able to feed all of this information, all the phone call information with the
[26:08] students, all the assessment information that we got from the students, their Salesforce interaction, email interaction, text messaging, everything is available. Now we feed it into the Tealium system and then the Tealium and Databricks are connected in two-way directions. So, it
[26:25] is it's we are sending data from Databricks into Tealium and then we are also getting all the insights from Tealium back into Databricks. And so, yeah, this two-way communication has helped us to approach and get to a learner's order to student 360 program.
[26:41] And then those are all triggered interventions as opposed to, you know, somebody stumbling on a student who has a need and and and intervening kind of, you know, uh I would say haphazardly. Obviously now, you said 54 54 hero moments all orchestrated and and and managed based on on data and it had an
[27:00] impact, right? You know. It did have a great impact and as you can see the number stopped the we said the very satisfactory number for student feedback is was increased by 5% and the reason being is okay, you completed a course or you completed an assessment
[27:17] within an hour within 15 minutes or you would get a message saying congratulations on completing this. Now that you have crossed this line, now this is your next item in queued up. So, the students are constantly on their toes being monitored, being helped and
[27:32] helped with their progress. Now, because we're here at at the Data Bricks, you know, AI Summit, they take away our speaker badge if we don't talk about AI. So, let's talk about obviously all those human interactions that are happening. So, a student calls in, talks either as a prospective students or as a
[27:50] you know, current student they're talking to staff, there's an opportunity to refine and learn and iterate and you guys are doing some unique stuff around coaching. So, if you want to walk us through that, that would be I think everybody would be interested. Absolutely. So, here's the thing that there is data flowing in from different
[28:06] systems. These are all siloed systems all around. But, the one basic thing that we've done is we've got everything into the Data Bricks environment. So, when a prospect calls in and talks on the phone and then they say that on the phone that okay, I
[28:22] am actually seem to be interested in a cybersecurity program, okay? And then they have an interaction and everything. Now, that call interaction transcript data is available and we get it from Genesis, so it's all available in Data Bricks. Now, the next time the student calls or the next time the student is on
[28:38] the website, you would want to tell show them information that they have already talked to us. They have already communicated to us. And how can we personalize all this information? Because we have classified one of the key project that we are now implementing for the fiscal year 27 is segmentation
[28:56] of data so that we can be very much focused one student at a time. And so, how can we do that? Because there are all sorts of users all sorts of students. There are wanderers, they they don't know what they want basically. They are just trying to feel find out a
[29:11] way. There are trailblazers who are very clear focused that they want to do this. And and then basically there are other other users or other types of learners who come with their backgrounds different backgrounds. So when we provide this information
[29:27] as soon as the transcript data is available to us within a time interval of 4 hours all this information is also fed to Teallium which is building that CDP. Which is building that that entire lineage of the data combining all and making a profile
[29:43] or a persona of that person. So that the conversations are very focused and then they don't have to waste their time talking about the same thing again and again and they can then make a very confined decision about their life. So all of this is done by collecting all
[29:59] the information, transcribing it, making it available for AI to decisions. And these are the whole hero moments that I mentioned about like okay at every stage the the upper funnel I talked about then during the journey they need a lot of help when they are doing their courses and programs and and
[30:15] in the process we also found that okay there are some toxic combination of courses. Sometimes if you do a sequel course right after a statistics course that's not good. We figured that out by looking at the data and we know that okay whenever
[30:30] somebody comes in we see their profile and we say that okay yeah based on your history or or your prior learning experience you're okay to take this or no we would not recommend you to have a sequel course back to back with a statistics course. We'll have to keep it apart. And that's basic small
[30:47] changes helped us to make the student more successful, me more engaged, and and to get more out of that degree. It's great. And and obviously, you've learned a lot. So, as we wrap up here today, you know, I think we've compiled some lessons learned or some some, you
[31:04] know, opportunities for the the crowd. So, if you want to Yes. tell us what we should be doing and then ultimately what we maybe want to not Absolutely. And so, this is my slide where I talk about data literacy mainly. And we practice, because we are an education institute, we practice this
[31:19] literacy within ourself. And how do you do that is you you do these basic things. You need to trust your data and you certify your data. So, having a single version of truth. Now, we talk about data governance and all now, but back in the day when we were building
[31:34] our first data warehouse also, we we kept this into the mind. This has to be an integral part of your data. Certification, the validation. So, we created two types of audits. One is the process audit and we had a a simple business audit. A process audit was
[31:49] source system has 100 records, did it move all the way to the target with 100 records? Yes or no? That's the process. If the process is broken, we should be the first one to know, not when the dashboard is broken, we come to know about it. And then the business audit was we worked with our business partners
[32:05] and said that what do you mean by right? What do you mean by wrong? Give us that threshold. And so, if the V-sat or if the drop rate is more than this percentage, that's wrong. So, we could track all that information. So, always and this is by just making it open and
[32:21] communicating all the changes and challenges, everybody likes to participate. Everybody likes to contribute to a success. So, the other thing which I say is like data literacy. And one of the thing which I'm currently, as I mentioned earlier, was the cost of that I'm making everybody
[32:37] aware. I'm not going to stop you. You won't believe we have 600 plus data bricks users. I'm not stopping you from doing what you want to do, but I want you to be vigilant. I want you to be aware of what's going on. And that's where the literacy comes into play. Treating your data as a product, meaning
[32:54] that yes, you have to put the the data as your prime importance. And and all the tools and technologies that you use around it is secondary. So, that's one of the thing is being agnostic to tool. Whatever you have, your main focus
[33:09] should be okay, I want to achieve this. I want to be able to reach out to my student at the right time with the right message. So, those are the things which I would recommend to do. But at the same time, I also have a list of okay, these are some things which I would say okay, we should be cautious about and we should care
[33:24] about is when we started looking AI. AI is not a tool. I I treat that as an intern, a new intern, a new team member who has joined you. He's learning. So, treat them that way. Treat them that they are helping you to speed up your work, amplify your work.
[33:42] Things that you would take 4 hours to do it, you're now able to do it within an hour. So, treating AI as a tool and constantly monitoring and supporting that new intern in your team. Fragmented data sources and reporting. That that is one of the key thing.
[33:57] I when I first built a warehouse and with my experience, I said, "I'm not going to be in the business where you will bring up a a Tableau report and a Power BI report that oh, this data doesn't match. You are showing me this number here, that number here." Nope. That is not what we do. Why? Because everybody's serving
[34:14] from that same data layer. So, your numbers should match. If something is wrong, you have a different business logic, you have a different purpose for which you are reporting this data. So, make sure that you communicate that to your end users that there is no such thing as the data duplication or anything. And then, more importantly,
[34:32] avoiding any kind of AI lock-ins or every vendor, as we heard in the keynotes also, that every vendor is providing with some AI support. So, how do you have a centralized AI support or AI governance model? That is one thing that you want to keep track of. And then, as I said, one of the key thing
[34:48] that had helped me throughout my career with WGU is my leadership supporting me and saying that, "Okay, yes. Fail fast. Move fast, fail fast." And that's that's basically the core of like what what your data literacy program can do for you. Wonderful.
[35:03] Well, I think that that that wraps up our our time today. I want to, you know, I want to thank you. Obviously, this you are the distinguished person on the stage today. Uh I just get to uh tee you up for the good the good the good uh content. But, um you know, we're going to hang out for a little bit if you have any questions, but if you want to
[35:19] understand how uh Tealium can apply to your business and maybe try your uh you know, find some of those triggered interventions, uh you know, for your brands, we're at booth 239 uh in the expo hall. So, you can also stop over there. Um but, uh you know, thank you all for joining and enjoy the rest of your conference.
[35:35] Thanks, everyone.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.