How many comps do you use?

[OSU Beavers]

Four sales and two listings.

 

[Tom D]

I can't imagine anyone having it more harder than me. Big urban city. Always 10 comps within 3 city blocks. I could almost use a stamp to paste in the same row home looking comps. Only bracket the sale price usually. Easy going direct lenders. I actually have a rubber stamp template with 90% of it filled already.

But, i can complain about the 3.1 time devourer, i mean faster done, per the 3.1 hustlers.

 

[J Grant]

J.G., your comments remind me of a thought that crossed my feeble mind this morning....today's assignment is of an extremely complex SFR in a neighborhood that ostensibly would appear to be homogeneous. It would be soooo easy to devote 30 minutes to report 3 comps that support the selling price... but big old dumb country boy like me will devote 3 - 6 hours to locate appropriate comps that reflect market reaction and also satisfy USPAP and lending guidelines.....but WHY DO SO, regardless that integrity might be all that one has left at the end of the day, although integrity just don't pay the bills or keep the insatiable wolf from banging on the door. SO WHY DO SO????

Why do so...because our asssigment ( appraisal purpose) is to find the market value of a property, not just match the sale price of the property) .It takes a lot longer and can be a lot harder to find market value, as you note it is a longer and more thorough comp search. And when the reserach and development has your MV opinion match or exceed a sale price,then your appraisal is credibly supported.

Other reason to do so... a forensic review can be done any time in the future and it can reveal if an appraiser skipped good comps and cherry picked to hit a number. And as you said, our integrity.

 

[J Grant]

Grouped data can support an adjustment, but is it the right adjustment ? ( the same can be said of bracketing. Imo, then both is ideal. However, bracketing is still the better because the sale or sales are right there on the grid. And that comp or comps was competition for the subject, not some #20 data point sale that a buyer for the subject would not consider.

If grouped data is good enough to include comps, then why not put some of those comps them on the grid? And if grouped data does not contain any comps, or just one very weak comp, then it brings into question how reliable the adjusment it indicates is.

Did the appraiser vet the sales and pick the sale in grouped data, or did they let software do it?

 

[DWiley]

And that comp or comps was competition for the subject, not some #20 data point sale that a buyer for the subject would not consider.

Was it though? That is the question. I cannot count how many times I have seen a very dissimilar sale thrown into a sales grid with the report noting that it was done for bracketing purposes. When that happens, there is no support for the adjustment, just the appearance of support. Using pairing and/or sensitivity analysis on comps that bracket can provide good support. But if that comp isn't really a comp it adds nothing. It is popular because it is easy for a reviewer/underwriter to see and it takes less time for the appraiser.

By the way, if an algorithm picked the sales used in a group data analysis with no input from a human, then that is no better than using a list that someone gave you.

 

[J Grant]

Was it though? That is the question. I cannot count how many times I have seen a very dissimilar sale thrown into a sales grid with the report noting that it was done for bracketing purposes. When that happens, there is no support for the adjustment, just the appearance of support. Using pairing and/or sensitivity analysis on comps that bracket can provide good support. But if that comp isn't really a comp it adds nothing. It is popular because it is easy for a reviewer/underwriter to see and it takes less time for the appraiser.

By the way, if an algorithm picked the sales used in a group data analysis with no input from a human, then that is no better than using a list that someone gave you.

Very few appraisers use a "list that "somone" gave them, and if they do, it is probably from a mentor who spent lots of time developing it. Most appraisers do develop a working knowledge of market and might have a general range in mind about what a feature or defect might be worth - then have it change when they do research.

Saying using an algorithm to pick sales with no input from a human is no better than using a list that someone gave you is a pretty poor excuse for using the algorithm! The algorithm is more dangerous because the slick support it shows makes people not question it - at least with the lousy comp , you saw it. And the problem is that relying on algorhtims without meaningful human input and further research can derive bad comp choices and from there bad adjustments on a mass scale.

You keep providing the example of a very dissimlar sale thrown on with no support - surely there must be many more times you see bracketing done well? Should bracketing be discarded just because a few do it poorly? If the majority of appraisals you see are doing it poorly, ask why they are still being chosen, when experienced competent appraisers are leaving the field because they can not get work or will not work for low AMC fees.

 

[alebrewer]

It's been my experience that utilizing a dated (but comparable) sale to bracket elements of comparison is much more meaningful than utilizing a non-similar recent sale - most especially in a stable market.

 

[DWiley]

Very few appraisers use a "list that "somone" gave them, and if they do, it is probably from a mentor who spent lots of time developing it. Most appraisers do develop a working knowledge of market and might have a general range in mind about what a feature or defect might be worth - then have it change when they do research.

Saying using an algorithm to pick sales with no input from a human is no better than using a list that someone gave you is a pretty poor excuse for using the algorithm! The algorithm is more dangerous because the slick support it shows makes people not question it - at least with the lousy comp , you saw it. And the problem is that relying on algorhtims without meaningful human input and further research can derive bad comp choices and from there bad adjustments on a mass scale.

You keep providing the example of a very dissimlar sale thrown on with no support - surely there must be many more times you see bracketing done well? Should bracketing be discarded just because a few do it poorly? If the majority of appraisals you see are doing it poorly, ask why they are still being chosen, when experienced competent appraisers are leaving the field because they can not get work or will not work for low AMC fees.

Don't conflate algorithms with tools. If an algorithm is picking the data, deciding what analysis will be conducted and spitting out a rate, with the appraiser having no understanding where that number came from, that is very different from an appraiser using something to analyze data that the appraiser selected. When I did group data analysis I had spreadsheets set up to analyze data I extracted from my local MLS. The fact that I did not do the math by hand does not diminish the results.

 

[J Grant]

Don't conflate algorithms with tools. If an algorithm is picking the data, deciding what analysis will be conducted and spitting out a rate, with the appraiser having no understanding where that number came from, that is very different from an appraiser using something to analyze data that the appraiser selected. When I did group data analysis I had spreadsheets set up to analyze data I extracted from my local MLS. The fact that I did not do the math by hand does not diminish the results.

I agree, however, back when you did it, there probably was no software with an algorithm for picking the sales and spitting out a rate. You have not indicated planning any specific oversight about it, at the GSE end or client end. Lack of specific certs and oversight will allow the "cheaters" a considerable advantage. Imo ,the low fees and fast turn times encourage cheating, and the software now makes it easier to do , and harder to detect, because slick "graphs, charts and stats, (for a result that can be bad,) looks authoritative.

The fact that you did not do the math by hand did not diminish the results because you analyzed the data, and selected the data.
Todays sofware programs, can use algorithms and autopopulate, which means an appraiser does not have to select the data, or can do so at the most superficial insta click level.

The difference is ,the appraiser is not using a calculator or software to crunch the math, they are using the software to do the appraiser's work of comp selection and vetting and/or analysis.

 

[NC Appraising]

Was it though? That is the question. I cannot count how many times I have seen a very dissimilar sale thrown into a sales grid with the report noting that it was done for bracketing purposes.

That would be me. My only difference is that I add paired sales or additional data to support the adjustment and that it is typical for the market in my report.....not just in my imaginary files.

Don't forget that it is not only about the adjustments.

Marketability or is it typical for the market?

AMCs and UWs (and the GSEs per their guidelines) are obsessed with this.

O wait...no closed sale with a similar feature.....their head explodes and the red light goes off at the UWs computer. Not typical for the market...therefore no contributory value and it affects the Marketability.


Just because there hasn't been any closed sales from within the past 12 months doesn't mean the feature is atypical or doesn't add value.

Although some lame appraisers thanks so. It's their lazy way out from not having to do additional research and analysis. Not typical for the market and no closed sales from within the past 12 months....therefore no value and not typical for the market. Easy peasy