School officials at the Plano Independent School District hoped to eventually roll out access to its new SAS-based business intelligence (BI) suite to more than 4,000 teachers and administrators.
So when just 90 principals brought the whole system crashing down within 15 seconds during a test deployment in the fall of 2005, Jim Hirsch, an associate superintendent who oversees the school district's IT operations, knew he had a problem.
"We found out very quickly working with our SAS team that we had underestimated what it means to have 4,000 users, even beginning with a handful of just 90 users," Hirsch said. "This did not come off without a hitch. This came out with a fantastic, catastrophic failure backing us up by six months."
The problem, Hirsch said, was that though typical corporate BI deployments often deliver reports to hundreds or even thousands of workers, they actually support significantly fewer users who directly access and manipulate the data and data models themselves, as would be the case in Plano.
"Companies themselves have thousands and thousands of users, but what everybody neglected to think about early on was that what was being delivered to those end users wasn't direct access to the enterprise intelligence suite," Hirsch said. "It was reports out of HTML pages or even just static reports. In our case, this was live data, and we were having the individual teachers in their classrooms accessing that live data as it ran through the model."
Hirsh's staff, which had experience building other applications and services for such a large user base, worked with SAS to reengineer the infrastructure to handle that type of concurrent use, including moving from a single-tier to a three-tier architecture to spread Web, stored procedures and database interactions over more processing, Hirsch said. The work took around six months, during which time "we limped along providing very basic reports to principals and asking teachers to limit their use," Hirsh said.
With the infrastructure enhanced, the SAS system finally went live in the April 2006, with the 2006-2007 school year being the first full year it was operational. The school district essentially lost a half to a full year's worth of use thanks to the initial "fantastic, catastrophic failure," as Hirsch described it.
Since then, however, the system has worked as expected, and the school district has steadily added more models, analytics and users without a hitch. Now, Hirsch said, even students and parents can access the system to track grades and test results.
"The good news was [the deployment failure] allowed us to do some additional fine-tuning, and the system no longer fell to its knees," he said. "We actually had done enough reengineering homework to realize what it took from an infrastructure standpoint, and SAS realized what it took from our environment standpoint, for this thing to be ramped up and ready to go."
As for the system itself, integration software collects student-related data from a variety of sources, including student schedule information, attendance records and test scores from the school district's student information system, and deposits it in a SAS data mart. Teachers access the role-based system via a portal on their laptops, where they are presented with data relevant to their particular students and classes.
Relatively simple visualization tools allow teachers to then examine student progress by a number of different "core" categories, Hirsch said -- and, more importantly, develop lesson plans to address specific problems.
A teacher may notice that a third grade student has lower-than-average math scores, for example. The teacher can drill down into the data to discover exactly which math areas are giving the student the most trouble, then the SAS-based system can generate recommended lesson plans to meet those problem areas.
Once a lesson plan -- which teachers actually implement via a separate curriculum planning system -- is developed, teachers can check back into the SAS system to track student results against the developed objectives on a weekly or even daily basis.
"We're able to give our teachers and principals predictability models for likely student results on the state assessment tests, and that's done early in the fall," Hirsch said. "When students return in the fall, all the variables that teachers need are basically there, so we're able to run them through the model and they can begin targeting those students [in trouble] early in the school year."
Going forward, Hirsch's team is working on connecting data from the curriculum planning system to the SAS system so teachers can dig into student data and create and carry out lesson plans in one place. That job has been made easier, ironically, by the initial deployment failure back in 2005, which forced internal school district IT workers to really understand how the system works.
"Since we became a more active partner after the meltdown, our own internal knowledge has grown to a great degree in terms of what we know we can connect to," Hirsch said. "We had to take more ownership of understanding what we could do with the tools and what was going to be possible…. We really are masters of our destiny now."
Most importantly, since deploying the SAS system, test scores are up throughout the school district now that teachers can better analyze and respond to student performance.
"Over the last three years, we've probably decreased the number of failures [on state-wide assessment tests] by over 10%," Hirsch said. "The bottom line is it's been a huge success for over two years now, but there was some really quick learning done by both SAS and my team as we first brought this in."
Dig deeper on Business intelligence best practices