Return to flip book view

In Predictive Analytics does the Technology Platform Make a Difference Data Driven Decision Making in Higher Ed Part 2 Predictive analytics is a data science that uses new and historical data to forecast trends Applying statistical analysis techniques analytical queries and machine learning algorithms to existing data sets predictive analytics creates a model on how and when a particular event will happen In the mind of many analytics is an arcane black box in which some sort of artificial intelligence AI makes predictions The term is mis applied to many sorts of applications currently offered to the higher education community This is especially true when applied to the analysis of existing university data sets relating to recruitment retention and financial aid The term appears in the description of everything from an LMS or SIS to the tools utilized by outside consulting groups on campus So if a system claims to provide data your institution can use in making business

decisions does it matter where that data comes from How the system itself works can impact how actionable the output actually is when it comes to making decisions about the use of scarce resources Let s look first at a primarily non machinebased example of a predictive analytics platform Consulting firms often use predictive analytics terminology when referring to their onsite staff using data imported into spreadsheet models While this may provide some results using such a manual analysis strategy results in a lengthy turnaround time losing valuable time to act on insights each semester These solutions are costly in man hours due to a less automated process and less rigorous quantitative methods and a limited analytical toolbelt On the other side of the ledger companies rooted in software applications tout large scale platforms that tie in to existing campus software systems to analyze student records for signs of those who may need interventions in order to be successful How are these institution wide applications implemented and what are the potential problems you may encounter First these software products are typically run in a data center when installed in the schools data center usually referred to as on prem for on the actual premises It might also be hosted meaning that the actual software was running on the software company s hardware often in a 3rd party commercial data center This is the SaaS or Software as a Service model which is considered a subscription to the institution that must be accessed via a web browser In the on prem model the university IT department is responsible for uptime and installing upgrades and patches which can result in the university not running the most current version of the software It also means that university IT is responsible for any connections to other databases where relevant data is stored In the SaaS model the University accesses the software platform via a remote login Due to the shared nature of servers in an external data center your software sessions are typically shared across a number of servers used by other clients If there is a network interruption or a server malfunction your data may be inaccessible for some time In addition IT security must be involved in SaaS implementations as the data is housed off campus SSO single sign on

with encryption is necessary so that data is protected And if your provider is using a commercial data center they must provide security certificates that guarantee that data is kept safe This adds complexity to the installation and maintenance of the software platform In either case the cost of software installation hosting and training can be substantial This is an ongoing often longterm arrangement between the provider and the university the cost of which are funds no longer available to improve actual enrollment and retention outcomes With both on prem and SaaS platforms a substantial implementation process is required Typically a team from the vendor will work with your internal teams to manage the process of installing the software training users and managing connections to the data housed within your institution

One EdTech company stated that they would have their system providing results in 100 days That s the length of a typical semester and assumes everything goes smoothly and according to plan Another consulting firm in this space states that their client should see results within the first year Again not the most reassuring statements Effective use of predictive analytics should incorporate a variety of data sources that the institution already has in a process that produces actionable results in a time frame that makes acting on those results more effective There is now a hybrid option that avoids many of the pitfalls of having consultants on site trying to manage a large on prem system or fighting to maintain security and operations on a SaaS system This new hybrid predictive analytics solution balances the benefits of stellar customer service and ease of implementation offered by consultants with state of the science analytical methods offered by SaaS platforms In this option the process is streamlined A simple implementation process identifies key stakeholders who will utilize the results a single IT resource to pull the correct data sets from existing databases and a secure file exchange methodology that does not put university data at ongoing risk Within weeks not months or a year this process produces clearly focused actionable results for recruitment retention and financial teams Additionally the models can be easily updated using a new semester s data without having to start the process over due to internally automated code and programs Finally the data can be interpreted for individual schools or departments within the University that need better information on which to base their planning The cost differential is substantial as the University is not purchasing software or a long term subscription service but paying only for specific data and specific recommendations Sightline Data provides results oriented predictive analytics using existing data using the latest data science modeling without the cost of either on prem or SaaS software installation