The smoothing model is one of our more powerful models.  It is smarter than regression in that it can predict when the next referral batch will come in as well as the number of days in that batch.  It is also smarter in that it can weight newer data much more than older data.  This is the one we look at most seriously when putting together our projections.

Our first forecast was for the May referrals.  We predicted referrals would arrive on May 24th and cover through June 16th.  Pretty close!

Here is a table showing projections from this method.  Note, the method can only assume things will be constant in trend from the last data point.  Therefore, only the next few months are truly relevant. 

We will be updating this monthly as each new referral group comes out. 

 

Projected Referral Date	Last LID Date in the Group
10/24/06	8/21/05
11/22/06	9/2/05
12/31/06	9/17/05
2/5/07	10/11/05
3/5/07	10/27/05
4/8/07	11/6/05
...
7/8/07	11/12/05
8/5/07	11/19/05
9/1/07	11/26/05
10/3/07	11/29/05
11/1/07	12/4/05
12/5/07	12/13/05
1/13/07	12/18/05
2/2/08	12/23/05
3/5/08	1/1/06
4/4/08	1/9/06
5/31/08	1/17/06
7/12/08	1/24/06
8/6/08	1/28/06
9/13/08	1/31/06
10/1/08	2/9/06
10/28/08	2/20/06
12/5/08	12/20/06

Please note: this is a time based method only!  It can not accommodate vacations and poll variations.