Archive for the ‘Load Management’ category

Correlation Street

September 10th, 2009

I mentioned to a few people at the recent AAIR Load Management SIG that I believed there were some leading indicators that might be useful in gauging demand and therefore load way ahead of the actual enrolment process.  I’ve had this hunch for a long time but hadn’t quite managed to find the time to see if it was true.

Well, yesterday I did a little bit of digging and I think the results are quite interesting, almost exciting even.  Just to back up a little, the issue we have is being able to understand the main variable of load forecasting - the commencing student intake (we can get the continuing numbers through the use of retention statistics).

Right now we’re concerned with the main 2010 first semester intake, the census date for which is way off into the future (March 31st next year).  At that point we will know what our commencing load and income will be.  Until then it is all down to the black art of load forecasting - or is it?

If we look at admission applications and compare the numbers with the eventual enrolments we find a pretty good correlation.  Here is how things ended up for us last year in terms of commencing students, each dot represents a course.  This data however shows all admissions for last year and all eventual enrolments.  I need to do an ‘on this day last year’ comparison to see if this is truly a leading indicator this far ahead of time, but either way it gives us a high confidence level at least a month or so ahead of census date.

picture-13

But here’s where it gets really interesting.  What if we look at leading indicators of admission activity?  We’re talking about demand generation here, the interest in a subject that ultimately becomes translated into an enrolment if we achieve conversion in Marketing speak.

So what might it look like if we took the number of course searches carried out ahead of the up-coming admission and enrolment period?  Might this show some relation to the eventual enrolment?  Well I think it does.

picture-11

The above chart shows the number of course searches in the month of September 2008 compared to the subsequent final enroloment in that same course at March 31st 2009, that is over 7 months into the future.

At this stage the figures are pretty rough and the process is quite manual but it only took a couple of hours yesterday to do this and I think the chart speaks for itself.

The other great thing about course and indeed unit searches is that they are high in volume.  We’ve been capturing this data since September 2007 and have over 9.2 million individual searches.

So the next step in our quest for automated and dynamic load forecasting is to somehow use these leading indicators to bring perhaps just a little systems input to bear on the mysterious black art.  Hopefully, (much) more (very) soon…

4 comments »

Posted in Business Intelligence, Load Management, Visualisation

Load Forecasting at AAIR

August 28th, 2009

The AAIR Load Management Special Interest Group met for the first time at QUT Gardens Point Campus in Brisbane this week.  The turnout was, I think, beyond the expectation of most people with great attendance from universities on both sides of the Tasman.  Organised by Jeff Holmes of QUT, this will no doubt become a regular event.

The stories on how load is managed in New Zealand were both revealing and alarming for me - imagine a scenario where you were obliged to enrol students irrespective of their prior education achievement yet were still expected to manage your load within a capped model.  What was also fascinating is the range of methods and approaches adopted by various institutions to both forecast and monitor load throughout the year.  Furthermore there is variance in terms of where this function is located in each organisation including Planning, Statistics, Finance, IT and the Vice-Chancellor’s office.

board-swing-8396-960-sm

What I’ve taken away from this and the first day of the AAIR Data Warehousing Special Interest Group where I caught a great presentation from Griffith (and missed what I understand was an equally great one from QUT), is that load forecasting is a bit of a black art. If load is only forecast at a high level then the ’swings and roundabouts’ approach tends to iron out things that are gaining and losing in popularity and forecasts can be expected to be within 1%, but the real art is in being able to provide a forecast down to award level.  Not everyone is doing this.

Load Forecasting at the highest level of an organisation needs to be a process that considers the following:

  • Historical data, particularly with respect to trends in certain disciplines or courses
  • Environmental data - economic situation, incentives to study certain disciplines
  • Continuing Students - treating these separately using history to determine rates
  • Commencing Students - here there is more variance in terms of enrolled courses
  • Faculty and School opinion - use the local subject matter expertise to inform and refine estimates

I think this latter point is the most significant and this is supported by opinion at other institutions which have devolved forecasting (or at least sanity checking) in place.  Having pondered on this a little bit, I think the following additional considerations would be worthwhile when delivering forecasts at the award level:

  • Provide a confidence level along with the award forecast
  • Revise the forecast continually based on current actual load
  • Retain the original forecast to assist in process improvements
  • Investigate correlations between admission and enrolment data for commencing students
  • Investigate correlations for other leading indicators such as level of interest in awards

I’m sure there are many more things to consider as no doubt I will discover as I quietly embark on yet another BI/DW voyage of discovery - more as it happens…

1 comment »

Posted in Higher Education, Load Management