Bill is right, Austin has entered the building!
Pam has come up with the same solution I use I think. At least we are on the same playing field. This all has to do with the proper sequence of adjustments. The correct sequence of adjustments comes from the market and you have to have a system to extract the adjustments from the market. I have used this system for a long time and it has never failed. First, forget all of that crap about location, design and appeal, functional utility, etc. adjustments. That stuff was put there for number hitters. Work the cost approach in reverse in the sales comparison approach. Adjust for major dollar items first like basements, garages, porches, etc. I keep a graph at the bottom of the page that moves every time I make an adjustment and guides me in the selection of the proper sequence of adjustments. One graph is the GLA vs actual price and the second on the same graph is the GLA vs adjusted price. Once you have made all adjustments for physical items only except GLA, I program my spreadsheet to make a size adjustment. For example if the subject has 1,000 sf and sale # 1 has 1,100 square feet, I program it to make a 100 sf + adjustment. Do this for all comparable sales. Then program the sheet to a grid square where you put your guess for size adjustment. For example, enter $50 per square foot and watch the trend line move after it makes a size adjustment at $50/sf. If the trend line still has a positive slope, use a lower $/sf number. Keep changing the adjustment $/sf until the trend line for the adjusted prices has a zero slope or is flat. That defines a perfectly adjusted sale. If anything else is influencing price you can see it in the variance of the adjusted sale points about the trend line. Normal random variance is generally plus or minus 7%, so if you are that close, no use looking for further adjustments. If you use this method you will find that you don’t want clone comps, you wanted two comps that bracket the subject and only one very similar. This helps judge the gravity of the value influencing factors.
This system has one draw back: You cannot cheat. If an adjustment is justified, it is shows up right on the graph. If you make an adjustment that is not justified or pick bad comps, there is no way you can make it work. You are trapped in to doing it correctly. Which is probably why my system will never be accepted by the industry. Basically, what I am doing is doing a manual-least-sum of squares regression analysis by using size to avarage out all unknown value influencing factors. If after I am finished and I have missed something, it will show up right on the graph. Then I go looking for it. I don't just assume it is there and adjust for it like view for example. If view matters and one comp has a view, you can guage the view. If view doesn't matter you can clearly see it on the graph.
