Transcript#

This transcript was generated automatically and may contain errors.

Perfect. Well, thank you everyone for joining us today. This is our first academic webinar. My name is Amy McCorkle. I am an account manager here at Posit. I'm on a public sector team, and I solely focus on our academic accounts. That includes universities, school districts, ministries of education, departments of education, all around the globe.

And over the past year, I've been hearing a lot more interest in how R is being used in the institutional research and advancements. So I've invited three of our customers to share what they've been building with the Posit tools. So to kick us off, we're going to have Inger and Lou from Harvard and Georgetown, both from the institutional research side. We also have Jason from the University of Pennsylvania from the advancement side.

We're going to end with a demo of Commons, which is a new open source package that was just released last week at Posit Conf. John, our solutions architect, will be walking you through that. Like Allison said, if you do have any questions, we're leaving room at the end for 15 minutes of answering those questions. So make sure you feel free to pop those in the chat.

And lastly, one of the goals of putting this webinar together was to really help you guys build this community. Building community helps drive innovation, culture, and technology adoption. So we're going to be adding a link to our data community Discord server. It's a great place to meet other people, share your work, and have fun there. You can look around at different channels. There's one for introducing yourself and then another one for educators, so you'll be able to find your people there. And then at the end, stay tuned because we also want to introduce you. We have a new R and IR group that will be meeting next week. So we'll give you all the details at the end there. And then I'm going to be putting my email to the chat as well.

So if you have any questions, feel free to reach out to me directly. So first, we're going to be going through some introductions and we will start off with Inger.

Hi everyone. My name is Inger Burgum. I'm Associate Director for Data Science and Analytics at Harvard University within the Office of Institutional Research and Analytics. And just kind of concisely what we do, we develop systems and tools and analyses that support university research and planning. Thank you, Inger. Next, Lu, you want to go next?

Sure. Hi, everyone. I am Lu Hao. I'm the Associate Director of Data Science and Analytics at Georgetown University. My office is the Office of Assessment and Decision Support, so we mainly handle institutional research and strategic analytics to support a different decision making process for the university. Thank you, Lu. And Jason.

Hello. Yes. Jason Vale. I've got quite a few years of experience in non-profit and higher education advancement, mostly at the University of Pennsylvania. So I support the advancement division and that's fundraising and alumni relations for the most part. So I support the, or my group supports reporting dashboards, data analysis, data engineering, and most recently, some AI agentic solutions. So POSIT has been a cornerstone tool for us for at least the past decade, if not longer. So thank you.

Thank you. Well, I will go ahead and pass it on to Inger to get us started.

Harvard: role-based report delivery with Shiny

Thanks so much, Amy. And thank you everyone for being here. It's great to be with you. So I work in the Office of Institutional Research and Analytics at Harvard, which is part of central administration. And our office supports decision making and reporting for the university as a whole. But of course, Harvard, like other large research universities, is made up of many different schools and colleges, each with its own priorities and data needs. So when our central office does a university-wide analysis, the results may need to be presented differently depending on the audience and what they care about.

One example, say we have, if we do an analysis of high demand courses. So the business school is going to care probably about business school courses and less about medical school courses and vice versa. And then for sensitive data, something like a climate survey report, each school needs to see only its own results. So the challenge that I want to talk about and how we approached it is essentially a role-based permissions puzzle. Like how do we securely deliver school-specific views of basically the same analysis? In other words, how can we create a single report that kind of knows who the user is when they open it and presents only the data that that person needs to see?

So we did have a way of doing this, and it just wasn't ideal. But just to kind of give you a sense of where we were coming from, we might do something like create a PDF and Excel file for each school, upload them to SharePoint, create school-specific folders and links, and then get the right link to the right person. And we tracked this all in a spreadsheet, and technically it was working and getting the job done, but it was really not efficient, and it created opportunities for mistakes like people getting the wrong link and us getting the dreaded email of somebody saying my link doesn't work. So what we really wanted was one report to maintain, one link to share, one pipeline instead of many different reports to manage, and a fragmented pipeline. But we also wanted to solve this using tools that we already had in our toolbox. We didn't want to have to go shopping for something new if we could help it, and we were already using R and Quarto and Shiny to create the data products, and we were using Posit Connect to deploy and share them.

And the key for this solution that we ended up going with turned out to be something that was already available within Shiny, the session argument, and more specifically the session groups object. So if you're familiar with Shiny, you might have worked with the input and output arguments within that server function. So session is an optional argument that you can also pass through, and where you can pass information or metadata about the current user session. So when an authenticated user is running that app, that information can tell us something about the viewer that we can then use to tailor or kind of control what they see or how the app behaves.

And there are actually a lot of different pieces of metadata that can be passed through that session argument. This is a list of different pieces of metadata that are available that you can pass through. The one that really mattered for us in our solution was the session groups object, which lets us pass the groups that are associated with that logged in user. In the case I was describing, this would be the group that identifies the user's school. So that's what allowed us to move from managing multiple school-specific reports to maintaining one single Shiny app on Posit Connect, but one that can serve up different views based on the school associated with that logged in user.

And the key for this solution that we ended up going with turned out to be something that was already available within Shiny, the session argument, and more specifically the session groups object.

Building blocks and code walkthrough

So far I've talked about kind of the why and the what of it all, but now I want to focus more on the how. So what are the main building blocks behind this approach, and what are the key pieces of code that are making it all work?

So a first step is creating the groups that are relevant to the project and assigning people to those groups. So you can do this directly within Posit Connect, or you can manage groups in an external application and then sync that information with Posit Connect. And the latter is what we ended up doing because in our particular environment, we already use something called Grouper for access governance. So that's where we can create the groups and assign people to them. So Grouper then is synced with Posit Connect so that group membership information is passed along to Posit Connect, and then Posit Connect has that group information and can pass it to the Shiny app through that session groups object. And the app can then use that information to determine what each viewer should be seeing when they log in.

So putting all that together in our particular setup kind of looks like this. We have these three main building blocks. First is Grouper, where we create the groups and assign people to them. Second is Posit Connect, where we deploy the Shiny app. And third is the Shiny app itself, which uses the viewer's group information to kind of customize what they see.

Now, to make this a little more concrete, I'm going to share this sample Shiny app that I made using Hogwarts course data that I made up just to kind of keep things simple. And this particular app has three parts that are customized based on my group membership. I'm a Gryffindor, so you'll see just Gryffindor-specific information here. And those three parts that are customized are the greeting, the course table, and the download file. So the same group information is being used in three different places.

Now, just briefly, I'm going to take a look at kind of what's behind the curtain with this code and see what's making it work. The first step, and probably the most important step here, is to create a reactive that returns the user's house. So the session group object returns actually all of the groups associated with the user, whether or not the group is relevant to this particular report. So what's going on here is we're identifying the group that applies to that app and then assigning that to this reactive. So once we have that reactive value, we can just use it throughout the application.

So for instance, this opening, this welcome text string, we can kind of plop that reactive value in that string and customize what the user sees in that welcome message. Second, we can use it as a filter to display courses in this table that only apply to the user's house. In this case, Gryffindor classes. And then finally, we can use that same reactive value to generate a house-specific download. So you don't have to necessarily read all this code, but basically what it's doing is it's using the user's house to determine which file to download.

So that's the basic pattern. We use the viewers group membership to create one reactive value, and then we reuse that value throughout the app to control the greeting, to filter the data in the data table, and to select the appropriate download file.

Results and benefits

So before and after here, we were going from managing separate files and links for each school, and now we have one Shiny app that provides each school with its own view. So what is this bias? Was it worth it for us? Yes, it was definitely worth it to kind of re-engineer this. For one thing, it gives us a single link to manage rather than 10 or 20, which means fewer opportunities for distribution errors and less headache overall. It also means a workflow that can scale. So for instance, we'll soon be using the same approach to distribute department-specific reports to about 50 to 60 departments within a single school. And then finally, everything is in one pipeline now. So we're no longer moving 10 PDFs and Excel files over to SharePoint, handling permissions there. So it's all in a single pipeline. So the benefit overall is not only a simpler workflow for us, but ultimately a more reliable and consistent experience for the campus partners that we're sharing these data products with. Thank you.

And I will pass it over to, I think, Lu.

Georgetown: transforming analytics workflows

Hi, everyone. My name is Lu. I work at Georgetown University in the Office of Assessment and Decision Support. Today, I wanted to share how we've transformed our workflow to handle data analytics more efficiently. A quick intro of our team. We're a small team of eight analysts with different areas of focus covering IR, survey research, BI, data science, data management, and course evaluation. And we all directly report to the Associate Vice President of Institutional Research. Since the fall of 2023, we have used

Posit Connect to publish and share data insights with stakeholders across our campus.

Before we had Posit, we were faced with some challenges, which ultimately drove us to rethink how we work. First is staff turnover. When people leave, they take most of the institutional contacts and knowledge they kept in their heads. So we realized the need to document things directly in code so that our work can remain consistent and reproducible.

Then there's this increasing demand for more data, more up-to-date dashboards, more predictive modeling, and scenario simulators. It quickly became clear that our old workflow, which relied heavily on Excel and PowerPoint, just wasn't going to meet those needs.

Then last but not least, our team handles many ad hoc requests on a daily basis. And a lot of them are really the same type of work, just with small variations. So we needed a more streamlined approach to handle them more efficiently.

A new automated workflow

And how do we adjust those challenges? We come up with a new workflow. So this diagram shows how the workflow works. I'm going to start with the box in the top right where it says Google Cloud Platform, or GCP. This is where our enterprise data warehouse is built. And our Posit Connect is also hosted in a virtual machine in GCP. So having them in the same environment makes authentication much simpler.

The actual workflow starts in a local IDE. We create a portal document or a QMD file. So a portal document is a notebook where you can combine regular text and executable code within the same place. And in that portal file, we write the code to pull data from BigQuery, and that is our data warehouse. And we then do the data wrangling and then produce a clean, analysis-ready data set. Next, we use an R package called pings to take that cleaned data set and publish it to Posit Connect, or ping it to Posit Connect. So basically, pinging is putting the data in a central location where other applications can access it. And then the middle layer is where the automation takes place. We publish the portal file to Posit Connect and set a schedule. It can be every day, every Sunday, or whatever cadence makes sense. Posit Connect then runs the code automatically on that schedule. So every time it runs, it pulls fresh data from BigQuery, cleans it, and then updates the pings. Everything's automatic, so no one has to do any manual work.

Then the bottom layer is where we have user-facing applications or data products. So anything downstream here that uses those pinned data sets gets updated automatically. So dashboards, reports, they all get updated. Everything is in sync. I know there are a lot of boxes and arrows here, but the underlying code is actually very straightforward. So here is a sample QMD file that handles the authentication and also sets up the pings. So this block shows the actual authentication code to establish a secure connection between Posit Connect and BigQuery. So once that connection is established, the rest of the code can pull the data from BigQuery without anyone having to manually log in or provide credentials. And this is what allows this automated workflow to securely access the data and run on its own.

And here is the actual pipeline. At the top here, we're just pulling the data from BigQuery using regular SQL queries. Then at the bottom, we take that cleaned up data and pin it to Posit Connect. And this is really the whole pipeline.

And once we have this automated data pipeline in place, that's where our team can get really creative. So we've built all kinds of products on top of it, dashboards, simulators, report surveys, and many more. So one pinned data set can serve multiple downstream applications. And then when the data updates, everything downstream updates automatically.

Examples of data products

So now I want to show what this looks like in practice with a few examples. So here's an example of a dashboard analyzing some synthetic iPads data.

This is a scenario simulator where users can set different parameters on the left side to project seed capacity for a new campus. This is a static report on course evaluation results with searchable and sortable columns. So department chairs can quickly find their schools' data without needing to submit an ad hoc request.

And this is a survey report of a very long questionnaire with over 100 or so questions. This report was built in Quarto with very detailed analysis that is searchable. And also, this makes things very easy to reproduce for future years.

Benefits of the new workflow

So to bring everything together, this workflow gives us a couple of benefits. First, it gives us a workflow that is documented, reproducible, and even version controlled. So the knowledge is now in the code rather than sitting in someone's head or Excel files or PowerPoint stack. Second, it gives us scalability. We can build the data pipeline once and use it to support many different products downstream without having to start from scratch every time. And finally, it makes us much more efficient with ad hoc requests. Instead of recreating similar analysis over and over again, we can now often point the requesters to an existing data product that already gives them what they need.

Instead of recreating similar analysis over and over again, we can now often point the requesters to an existing data product that already gives them what they need.

Thank you for your time. I'm happy to take questions during the Q&A. Now I'm going to pass it over to Jason.

Hello, everyone. So thank you to Inger and Lou. I've enjoyed your presentations. I made a note about session groups. I'm going to try that out. Okay. So I, again, I'm the Senior Director of Advancement Analytics here at the University of Pennsylvania. Today, I'd like to take you through how we've been using Posit here at Penn, what our environment looks like, how we've built our platform, and some of the exciting things we've been able to accomplish as a result of using Posit. So I hope you come away with a clearer picture of how Posit can support analytics in a complex higher education setting.

The challenge for us, as I think for a lot of you, is we operate in a highly complex data environment. So we work across multiple systems. They don't always talk to each other very well. We've got academic data, financial data, and of course, our own advancement data. So the siloed nature of the systems creates challenges. At the same time, we are expected to move quickly, of course, to deliver results. And we also want to find space to innovate within our projects. And layered on top of that is our institutional data policies and our responsibility to keep our data secure, of course. So Posit really meets all of those demands for us.

So the tools we're using here, a lot like Ingrid Liu, we're using Posit Connect. We use Posit Workbench for our development work. We have Posit Workbench and Connect running on Linux on-premises servers here that we administer. We also have recently a Posit team on Snowflake Snowpark container services. So just as of January this year, so nine months. I'll dig more into what we're doing with that tool a little bit later.

So again, why it matters, what benefits do we get? Well, we get speed. Can you make sure you put your slides on the screen? I'm not seeing anything.

Okay. No, you're totally fine. I was like, hmm. So yeah. So I'm just rattling away. Okay. Not too long. You're good. We caught it before it became too much rattling, you know? Okay. All right. There we go. Sorry about that. How's that? It is... Now I see it. Perfect. Thank you. Okay. Thanks, guys. So I don't think you've missed too much.

So again, here are the tools we're using. And so the Posit platform really does, it gives us some... It gives us speed. It gives us some flexibility. We have access to a lot of different tools. We use primarily R. We have been getting more into Python. Python, we find it to be kind Python, we find it to be kind of what Snowflake prefers, but R works well as well, of course.

But yeah, the ability to really host a lot of different dashboards. We built some custom APIs using the Plumber package. We can host those on Connect. And we do a lot of data integration and data wrangling and ETL work as well using... Leveraging the R markdowns, and in some cases, the Python notebooks, and then Connect's scheduling capabilities really help us out with providing data for various applications.

Posit team on Snowflake

So Posit team and Snowflake. This has been a huge upgrade for us. It's allowed us to really accelerate our development, accelerate our projects. And it's very secure. All of the data that we're working with, it stays within the Snowflake perimeter. So even with Posit Workbench, we're accessing the power of the Snowflake Cortex agents within the Posit Workbench, Posit team development environment. And we also get the benefit of the AI-assisted development. So command line, cloud code development experience as well. So all of that is done for the most part in the Positron IDE, which we've started to use after being really loyal users of the RStudio IDE for years and years.

So yeah, so a significant improvement over our prior development environments, much more streamlined workflow. It really... Everything integrates in with Git and GitHub very well. If you haven't taken advantage of deploying your applications on Posit Connect using GitHub, it's really worth a look. It's a game changer in having automated application updates.

And, you know, the Positron IDE is very familiar to a lot of programmers who've been using VS Code for years.

So, again, just faster time to value, faster development, broader capabilities. We have access to a lot of tools, a lot of AI agents within the Posit team. So, those are some of the benefits. These are some agents in action. So, we have put these agents that we've built into production. This is a particular agent for annual giving. So, if you're familiar with advancement and development, you know how important annual giving is to the institution. It's, you know, sort of the lifeblood of unrestricted annual giving for Penn in particular. And so, this agent allows all of the annual giving programs. We have, you know, many, many schools and centers here at Penn, a lot like Harvard, 18 schools and centers. We also have central fundraising units as well. This is a conversational interface, a way to access the agent that we built. This is a Python shiny dashboard. And so, they have a prompt down at the bottom. They have some text to help them fill the prompts out on the top. And it's a really well guardrailed and compliant way to provide mail, email, and text marketing appeals for our annual

giving folks. So, they can have a conversation and, you know, depending on how much information they provide initially, they may be prompted to answer, you know, additional questions. But it's really, it's delivering consistent, compliant output for these appeal channels in every case.

And it's flexible enough to accommodate most of our annual giving programs here at Penn. So, athletics, the vet school, the dental school, all of those programs can use this agent. And we feel good knowing that we know what data they're getting. It's a predefined template that has been developed to work well with all of these different annual giving communication tools that we have. So, once they're done their discussion, they can go down to the left-hand side and download their results and use the appeal data.

So, the second agent, this is the first one we developed this winter. This is one we had sort of floating around in our minds for a while. Anybody, again, in advancement who knows what contract reports are, these are the bread and butter of fundraisers, sort of one of their tools for just keeping track of interactions with our constituents. We have some contract reports going back more than 40 years for alumni. These are a varying sort of, I would say, quality and format and length. And it's really never been possible in the past to synthesize all of those contract reports into a coherent summary. So, I would say the possibility arrived with the access to these large language models, particularly for us in Snowflake Cortex and the Posit Workbench environment, we were able to get the agent set up to work with some semantic layers of data in Snowflake and really do a good job of producing a comprehensive, well-written, concise, accurate contract report summary.

I have a very highly redacted screenshot of one over here to the right, just to give you a flavor of what the output looks like. But again, they can get a bulleted or a more comprehensive executive summary of a few pages. And this is also being made accessible with a Python Shiny dashboard on. In this case, we're putting our agents on Posit Connect, Posit team on Snowflake. So, again, we feel good about the sort of the security perimeter that that provides us and having those agents made accessible to our users from the Posit Connect on Posit team has been working really well for us. So, yeah, thanks for your time and looking forward to any questions that you might have. Thank you. Sorry for not bringing my slides up earlier.

So, I would say the possibility arrived with the access to these large language models, particularly for us in Snowflake Cortex and the Posit Workbench environment, we were able to get the agent set up to work with some semantic layers of data in Snowflake and really do a good job of producing a comprehensive, well-written, concise, accurate contract report summary.

Introduction to Posit Commons

Great. Thank you, Jason. Thanks. John, I think you're next. Okay. I think I need Jason to stop sharing. Yes. I know Jason's a time hog. Yeah, there we go. Yeah. I will say the slides are great. So, at least you get to see a beautiful slide. You would think I would have mastered this having only done thousands of these. But okay. Yeah.

Okay. All right. So, y'all should see my screen. So, very nice to be with you guys today. My name is John LaPlante. I'm a solutions architect with Posit. And I am really excited to share with you a brand new, highly experimental package from Posit that was just announced last week at Posit Conf, and it's called Posit Commons. So, let's talk about what Posit Commons is. So, I'm going to talk about Commons. I'm going to talk about the different pieces, parts regarding it, why we have Commons, and I'll do a little bit of a live demo.

So, first of all, what is Commons? Commons is a framework for building self-service data analysis agents in R and Python. The package is built on ellmer, which if any of you know what ellmer is, ellmer is an R package that connects your programs to LLMs. chatlas is a similar package, but in Python. And then shinychat is what displays the connections that ellmer and chatlas make to a Shiny application. And I'm going to show you what those look like in just a little bit.

So, those are our open source LLMs stack. Behind ellmer and chatlas, you can use whatever LLM you want, including locally developed LLMs, as long as they're written to the open AI standard, which almost all of them are. So, why did we do this? So, there are two ways that AI is currently failing resource teams. First of all, they're either too locked down that they're very narrow, they're very prescribed, it only answers certain highly regulated questions, or they're completely untrustworthy, where you just release Clod or Gemini or ChatGPT on your data, and you have no idea what they're doing, how they're doing it, or why they're doing it.

So, Commons takes a third path. We have the flexibility of an agent anchored to the validity of the code your team already trusts and the definitions you already abide by.

We have the flexibility of an agent anchored to the validity of the code your team already trusts and the definitions you already abide by.

How Commons works

So, how it works. So, the user, starting from the left, the user asks a question. Those can be researchers, PIs, deans, vice presidents, board members, whomever you want to have ask a question. What Commons then does is search their trusted calculations. Those trusted calculations, that semantic layer, you create in code. So, it's a, if you think about a Shiny application where you have the server UI and the app.UI, you have a document that explains exactly your trusted calculations. If a user asks a trusted calculation question, Commons will find it, run the code, and then give them the answer, along with a little check mark saying this is a verified calculation and you can trust this.

Now, in the low trust area, you've got two different possibilities. One, you have a cited calculation, which means it didn't find a calculation specifically, but in your documentation, it found a description of what a calculation might be. And in that case, it'll run the calculation for you, give you the answer, and then it will give you a cited tag. And then the last step is if it finds nothing having to do with either a trusted calculation or a citing, it'll still run a traditional AI code, but it'll give you a little note saying, hey, this is untrusted. You really need to research this before you use it for anything of interest.

So, how does it work? So, first of all, you have your client, any LLM via ellmer or chatlas if you're using Python. So, AWS Bedrock, Anthropic, Azure, GitHub Copilot, whatever LLM you want to point it to. Then you have your data sources, anything you can connect to, whether it's any kind of DBI connection, Databricks, Snowflake, anything you can connect to regularly using Python or R, you can connect to a commons agent. Then, as I mentioned earlier, you have your semantic layer, validated R functions or Python functions tagged with an at measure tag. This is your trusted code that you have used and have agreed upon. Then you have your context layer. This is traditionally what you would have an MCP server for, a model context protocol. And then you can also have instructions that you create as well. And then that runs through commons, and then you get your governed agent.

Live demo: Campus Lens

So, let's see what this looks like in action. So, I have a very simple application here. This is called Campus Lens. It's Meridian State University. I just used some Fall 22. This is a fictitional company or a fictitional university. So, here you'll see attrition by program. What percentage of students in each program withdrew or were dismissed? This is a verified, and this is just an example. You can type whatever message you want. I'm going to run attrition by program.

So, what this is doing, it's searching for a trusted calculation. It's going to find that trusted calculation, and then it's going to run that. So, here is each program, the end, how many observations, how many withdrew, what percentage. And because I see this little green check mark, that's going to tell me that this answer comes from a trusted calculation.

By contrast, if I run a program GPA calculation, so this says, what is the standard deviation of term one GPA by program? If I run that calculation, the commons agent is going to look through its semantic layer, see if it finds a trusted measure. It didn't, but it's going to run the calculation for me anyway. So, this is going to give me an untrusted calculation. And then the last one is, how is academic probation defined? If I come and run that, it's going to look through trusted calculations, but it's not going to find anything. But it is going to look through and it's going to a, academic probation is defined by a term GPA falling below two. And then this gives me the explanation of where it found that particular definition. So, it's in the term GPA table. It's also in an ACAE table.

So, just as part of a live demo, that is what commons gives you, is it gives you the way to create these three different types of calculations, either verified, cited, or untrusted calculations. I'm going to drop in the chat here. Let me grab this. Put that in the chat here. I think I can put this in chat to everyone.

This is the blog post that you can go check out. Again, this was just released. This blog post was nine days ago. So, this was released at Conf last week in Houston. So, again, this is highly experimental. I encourage all of you to go out and give it a try, but realize this is experimental. It can change, as you'll see by the 0.1.0 package version. But it is really exciting to start to put some guardrails and some expectations around AI.

R in IR community

All right. The next thing I want to talk about, we mentioned this R in IR group. This is an institutional research focused R community. It meets once a month. We talk about how we're using R in institutional research, ask questions, share ideas, learn from each other, build connections with other higher ed R users, and then who we are. I thought that was a little clever. Who we are. IR professionals from across the U.S. IT, finance, business ops, registration, assessment, anyone curious about R, IR, or higher education. This is a small group. They've only met twice.

I'm going to drop this link in the chat. Grab both of these links here. Let me copy these. I'm going to put these both in the chat. So, the first link, it's a bit.ly link. That will actually be a Google calendar invitation. You can click on that. You can join that meeting. If you, for whatever reason, if that bit.ly doesn't work for you, I've also included Kevin's information. Kevin's leading the creation of this group. So, if you have any questions whatsoever, Kevin is a great contact. Or if you can't get that Google calendar invite to work for you, reach out to Kevin and he can assist.

I think with that, I think that of all the slides I have, I'm going to stop sharing my screen and I think I will turn the time back to Amy. Thank you all very much. Yeah, thank you all for joining. I know there's a couple questions that were in the chat that we haven't addressed yet. So, we can start addressing a few of those and great presentations today. Thank you for your participation. The first one that we have

is from Keith. Any suggestions for making a case to senior leadership for Posit Connect and other tools like these to invest in? I have my thoughts on that, but Jason, Inger, or Lou, is there anything that you wanted to add from your perspective on getting these tools approved? Yeah, I mean, it's a great question. I would say Posit is just a really good value proposition for us when we compare it to some of the other tools that are available out there. So, depending on, you know, of course, what your budget is, what other tools you're using or looking at, I mean, we just really found that it compares quite well. And, you know, for instance, we're able to do unlimited installs of Connect and Workbench. So, that took care of us being able to install it on Linux and also onto the Snowflake container services. So, you know, for us, it's been, and there are other reasons as well, but those are the top of mind for me.

And what I would say, what I think has worked really well is if you have just like a small use case, building those applications, showing those to leadership and building off from those and proving the value, that's what I've seen be successful. Inger, it looked like you were coming off mute. Was there anything else you wanted to add there? I was just going to add that we had already been using our, you know, all our workflows were in ours. So, it made sense as, you know, a product for helping us share the things we were creating. So, that made sense. The other thing I wanted to mention was like, I think it makes sense to bring in your IT folks as early as possible to make sure you like, you work out kind of who's responsible for what and how to get things set up on the back end and make sure you have a kind of buy-in from the IT side as well.

Yeah, that's a great point. There's a lot of moving parts on making sure that these tools work well. So, IT is a big partner in helping not only getting those approved, but also getting their perspective from that as well. And I did put my email in the chat. So, if you're interested in exploring that and wanting to talk to us about those use cases that you have and help us build your use case, we can always do that.

Community, K-12, and code vs. no-code

Greg was asking, is there a way for an analytics group for K-12? Every education group I've explored, it's more people teaching R. So, absolutely. Like I mentioned before, I work with a lot of R education accounts. One of them is universities. We also have a lot of the K-12 and school districts, departments of education. This was just the first webinar. We wanted to see who was interested. So, please send me an email. We can discuss on putting together a separate group for K-12 as well. And John, is there anything else you wanted to add there? Yeah, Amy, let me, and Greg, I think that's a great question. So, one of my, in my previous life, I went through Harvard's strategic data project, which teaches a lot of K-12 data sciences across country. So, I'd love to have that discussion with you as well. So, make sure you reach out to Amy and we can set up a follow-up to discuss that more in depth, but love the K-12 focus as well. So, thank you.

And then the other question is the R and IR group also interested for people from Europe. I think yes. This is, like I said, this is a small group that Kevin actually started. So, it's not even hosted from Posit. It's, I think we're just trying to build up this community. We've seen this built up in different industries and we see that, you know, there's a lot of you out there that are looking for support and looking for other ideas of how to get started. So, I would say go ahead and join that group and get connected there.

Yeah, Logan's got a question about code versus no code. And I think that's a, it's a question that we get a lot. And it really, Logan, it comes down to reproducibility and trust as to the decisions that are being made with the analysis that you do. And we really talk about it as the decision is really, is this a business analytics thing or is this data science?

You know, if you're just doing business analytics and you're doing, you know, Excel, you know, bar charts and pie charts, and you just need a quick little answer, that's one thing. But if you're writing a report that decisions are going to be made around, if you can't reproduce, if you can't stand behind, if you can't explain 100% where that came from and how to reproduce it, that's where the decision between business intelligence and data science comes in.

But if you're writing a report that decisions are going to be made around, if you can't reproduce, if you can't stand behind, if you can't explain 100% where that came from and how to reproduce it, that's where the decision between business intelligence and data science comes in.

And Logan, if you'd like to have a more in-depth conversation, Amy and I'd be happy to sit down and talk with you and show you a little more about how that works.

Closing remarks

Thanks, John. And then, is there a link to join the R and IR group? Do you guys have access to the chat? We did put it in the chat. Allison, they should have access to that, right? Okay. Yes. I was going to confirm. I think it's in the chat. And yeah, everyone should have access to it. And I'm going to repaste that message real quick. That is going to be a verbatim repeat of what Libby, who initially sent that message, is. So, I'm just going to go ahead and copy and paste that. So, if anyone didn't see it before, they should be able to see in the chat now.

Yep. So, there's two of them. One was for the Discord, and then the other one John also put in there was for the R and IR group. That's one. John, do you mind linking that for me in the chat if you have a second? I think he already did. So, they should have access there.

And again, this is just our first one. So, if any of you have anything else that you want to share, if you think this was helpful, we'd love to get your feedback. And we'd love to host something like this again if you guys are open to it.

Yeah. I wanted to say thank you to everyone here who presented. And this was a really awesome first jab at something like this. So, very happy that we were able to pull this off. And so, a special thank you to Inger, Jason, and Lou as well. Amy and John, thank you so much for your time here. Now, for everyone else who's in the chat, I know I've mentioned this a few times, but you will be receiving a recording of this webinar within 48 business hours. It will be on YouTube. So, we will send you that link. And again, just reiterating, if you have any additional questions for anyone or more specifically Amy, she's been putting her email in the chat. Otherwise, I just want to say have a great rest of your day, afternoon, evening, from wherever you're joining. And yeah, thank you for spending your time today with us at Posit. And yet again, I just want to thank you, Inger, Jason, and Lou for your time. It was really awesome to see what you guys have done. Yes. Thank you all. All right. Well, have a great rest of your day and happy Thursday to everyone. Thank you. Thanks, everybody. Bye. Bye.