What Data Is The Most Valuable?

Most Popular Reply

Charlie MacPherson
China, ME · Member since 2014 · 3k+ posts · 4k+ votes
6y
Done - and a note of caution on comps. 

Without deep local knowledge, it's not possible to deliver anything better than the random dart throw of Zillow - and those are often so far off as to be laughable.

In the Boston area with an average sale price of $765K, being within 20% of the true market value 79.3% of the time means that the predicted error is $153,000 or less, except for 20.7% of the time where the error will be MORE than $153,000.  Not only is that a useless number, it's dangerous when people rely on it.

This is something that we Realtors have to struggle to overcome.  We talk with sellers who have an inflated view of their home's value because "the Zestimate said $XXX,XXX", but we know it's not.  Investors who rely on these numbers can get hurt too.

I still run comps manually.  It takes me 30-60 minutes to do them, but when I tell a seller what their home will sell for, I'm almost always well within 5% of the actual sale price.

The bottom line on comps is that no data is better than bad data.




21Reply
See this reply in the discussion

37 Replies

Jump to latestLatest
  • Cara Kennedy
    Rental Property Investor · Member since 2019 · 189 posts · 62 votes
    6y

    @Dave Meyer - Done. Thanks for asking your customers! We appreciate the work you do!

    3Reply
  • Charlie MacPherson
    China, ME · Member since 2014 · 3k+ posts · 4k+ votes
    6y
    Done - and a note of caution on comps. 

    Without deep local knowledge, it's not possible to deliver anything better than the random dart throw of Zillow - and those are often so far off as to be laughable.

    In the Boston area with an average sale price of $765K, being within 20% of the true market value 79.3% of the time means that the predicted error is $153,000 or less, except for 20.7% of the time where the error will be MORE than $153,000.  Not only is that a useless number, it's dangerous when people rely on it.

    This is something that we Realtors have to struggle to overcome.  We talk with sellers who have an inflated view of their home's value because "the Zestimate said $XXX,XXX", but we know it's not.  Investors who rely on these numbers can get hurt too.

    I still run comps manually.  It takes me 30-60 minutes to do them, but when I tell a seller what their home will sell for, I'm almost always well within 5% of the actual sale price.

    The bottom line on comps is that no data is better than bad data.




    21Reply
  • Jacob D Cockerell
    Appraiser · Leitchfield, KY · Member since 2018 · 54 posts · 42 votes
    6y

    @Charlie MacPherson you make a good point about the comps. Personally I wish Zillow would just get rid of the zestiment. I'd rather see a list of all houses sold within the past 6 months maybe categorized by sqft or something. 

    15Reply
  • Dave MeyerPro Member
    OP
    Head of Real Estate Investing at BiggerPockets · Seattle, WA · Member since 2015 · 224 posts · 826 votes
    6y

    @Charlie MacPherson Very good perspective, thank you! We're still in the exploratory stages of what to do here, but this is really helpful insight. Would something like raw comp data be helpful? An easy way to see local comps, without the 'estimate' piece of it?  A tool that allows investors on BP to manually find their own comps? 

    As a side note of curiosity, do you think that Zillow tends to skew too high? Low? Both? I have some training in data science, and am always curious about the predictive quality of things like the Zestimate. Insight on what they're not doing well can really help us down the road -- not that we're currently planning to provide our own estimates of value or anything -- more of a professional curiosity. 

    Podcast Guest on Show #186
    BiggerPockets
    3Reply
  • Charlie MacPherson
    China, ME · Member since 2014 · 3k+ posts · 4k+ votes
    6y

    @Dave Meyer  I don't think raw data is helpful.  There are rare instances where it might be, like in a condo or tract home development where every unit is the same location, age, size, beds, baths, etc.  In my experience, that's about 1 out of 30 properties.

    There are way too many details that require you to be INSIDE the home to take a good run at determining value, especially if any level of rehab is necessary. 

    Very few listings will discuss things like knob & tube wiring, fieldstone foundations, cedar post supports & beams, damp basements, horse hair plaster, uneven floors, condition of HVAC equipment, mold in the bathrooms and basement, brands of kitchen appliances, quality of light fixtures and cabinets, height of ceilings, quality/age of windows and doors, etc.  Even the appearance of the rooms and floor plan are suspect when photographers abuse wide angle lenses - or are really lousy photographers.  

    You also need to know the area to do the job right.  For example, I showed a home about 6 months ago that to all appearances was on a quiet side street.  But knowing the neighborhood, I could tell the seller that this was an extremely busy cut-through for commuters trying to avoid a traffic light 1 block ahead.  Another example, in Scituate Mass where I used to live (an upscale seaside town), it wouldn't be obvious to an outsider (or to AI) that homes on the east side of Route 3A sell for more than those on the west side - even if they're 500 only feet apart.

    I've seen countless houses that look good on paper, but inside make me think the best outcome is a well-placed lightning strike.

    My thinking is that bad data is much worse than no data because it attacks the credibility of whomever provided it.  Ask any Realtor - Zillow is widely perceived as a bad joke when it comes to valuations.  They are much more of a problem than they are an asset.

    As we commonly tell clients, the "A" in Zillow stands for accuracy.  

    As to whether Zillow skews high or low, I haven't quantified it, but I've seen both.  My initial reaction when given a Zestimate is to ignore it completely and run my own comps.  

    Raw comp data could be marginally useful if it were accompanied by enough information to better assess the house, but as MLS is the best source there is - and it's nowhere near good enough, I think you'll need a different solution.

    7Reply
  • Will BarnardPro Member
    Moderator
    Developer · Santa Clarita, CA · Member since 2008 · 15k+ posts · 10k+ votes
    6y

    I can agree with Charlie on what he is stating. To his point, I have an area in my market here in So Cal where the west side of the Blvd will sell for significantly more than East side of same blvd and we are talking just 50 yards apart here! So what is on paper or some algorithm will NEVER be as accurate as a true CMA from a real estate agent or professional. That said, and one who built a system with algorithms to compute values for investors several years back, I got close on much of it, but there are just to many things that can sway it and no algorithm can pick all that up.

    Podcast Guest on Show #130
    7Reply
  • Russell BrazilBusiness Member
    Moderator
    Real Estate Agent · Washington, D.C. · Member since 2012 · 17k+ posts · 30k+ votes
    6y
    Originally posted by @Will Barnard:

    I can agree with Charlie on what he is stating. To his point, I have an area in my market here in So Cal where the west side of the Blvd will sell for significantly more than East side of same blvd and we are talking just 50 yards apart here! So what is on paper or some algorithm will NEVER be as accurate as a true CMA from a real estate agent or professional. That said, and one who built a system with algorithms to compute values for investors several years back, I got close on much of it, but there are just to many things that can sway it and no algorithm can pick all that up.

     Yup....we can in some specific locationa see a $200k variance on identical houses across the street from one another.

    [email protected]301-893-4635Podcast Guest on Show #192
    District Invest Group546 Reviews
    WebsiteView Page
    4Reply
  • Greg Moore
    Rental Property Investor · Tampa, FL · Member since 2019 · 128 posts · 166 votes
    6y

    @Charlie MacPherson

    I can’t argue with your numbers, or dispute your expertise as a professional seller of real estate. I can say, however, that data+machine learning will beat out human computation. Personally, I think that time is now, and those that don’t see it that way also don’t see the meteor speeding toward Earth. Zillow - love ‘em or hate ‘em - are sitting on mountains of data with sophisticated algorithms. All of that data+ML will continue to make its predictions untouchable by Homo Sapiens.

    I come at this from the buyer/investor perspective and I’m continuously annoyed at having to pay a ridiculous amount of money (and time) for an Appraiser to tell me the market price that is within a few dollars of what Zillow told me it was before I ever decided to buy the property! From the investor side, this Appraiser thingy feels like a job that is ripe for transformation and automation.

    #indatawetrust

    6Reply
  • Charlie MacPherson
    China, ME · Member since 2014 · 3k+ posts · 4k+ votes
    6y

    @Greg Moore There are things that AI just can't do.  Drilling down to the single property level is one of them. 

    Can they predict major market swings?  Maybe - but that relies on human inputs and it's yet to be proven.  In fact if it were that easy, AI would be turning astronomical stock market profits on a daily basis.  It's not.

    AI is going to have to get INSIDE each and every property.  And it's going to know sights, sounds, smells and have a deep knowledge of repair budgets, comps, ARVs, neighborhoods, etc.

    Ham-handed attempts?  Sure.  AI is great at ham-handed attempts at predicting reality. But when it comes to reliable data - the kind you can bet your own money on?  Not a chance.

    It's just not going to happen in the foreseeable future.

    Call me an old fogey.  Call me a dinosaur.  Call me whatever you wish, but until Zillow knows a cheap "contractor-grade" kitchen cabinet from a high end one with nice soft close drawers, all I can say is "good luck to you and the Red Sox". 

    You'll both need it.

    4Reply
  • Derek Hutson
    Denver, CO · Member since 2017 · 38 posts · 14 votes
    6y

    @Greg Moore You are right that technology is the way of the future... but there will always be a place for the human brain!

    Example: If you continuously feed a computer data from the cost of renovations in a neighborhood (gathered from contractors, construction companies, etc...), broken down into subcategories, you can maybe see that exterior renovations generally cost more than kitchen/bathrooms etc for the majority of properties in that area...

    Then you could start to come to more educated conclusions of maybe what material to use for a deck foundation because it turns out the extreme humidity wears out certain materials faster than others, and who you should hire to do the work based on a calculated average of cost vs. positive reviews for each contractor..

    This is a pretty simple and limited example, but I guess my point here is that data and ML will absolutely help you identify certain trends with more precise accuracy, however it still requires a thinking brain to figure out why that trend exists, what direction that trend is going to go, and execute a plan based around your newfound insights.

    4Reply
  • John Collins
    Investor · Tx, GA · Member since 2019 · 313 posts · 337 votes
    6y

    Took the survey.

    Wanted to add I am more than happy to upgrade memberships if some form of consolidated comps are provided. What houses in the neighborhood or surrounding area rent for , all sales in the past 3 years, median home value for the surrounding area (or block if it's in a crowded city), etc. Right now, it's stuff I have to click on each home manually to figure out. 

    I wouldn't ever want an algorithm to do any kinds of predictions or forecasts for me - that is my job and my judgment. This information allows me to determine whether a house is worth investigating further and accurate bids to make on it. 

    1Reply
  • Dave Van HornPro Member
    Fund Manager · Wayne, PA · Member since 2009 · 1k+ posts · 1k+ votes
    6y

    Done and done!

    Just throwing in my two cents here, we look at A LOT of data in my business and in my opinion neighborhood data is the most valuable. I say this because you can really only find the best of it from the boots on the ground.

    Sure comps can be pulled from MLS data or public record but something like general neighborhood trends can usually only be determined by someone with their ear to the ground. Someone who knows the area and it's history well.




    Podcast Guest on Show #273
    3Reply
  • Darren SagerPro Member
    Investor · Tampa, FL · Member since 2013 · 2k+ posts · 1k+ votes
    6y

    I think being able to utilize data and manipulate it in the way that you need is probably the most important way to determine trends. The NAR aggregates data far better than Zillow ever could, because it's fed by info put into the MLS' and not just a court filing and algorithm like Zillow. It allows us to view data in ways that makes Zillow look antiquated. The issue with the NAR's system is that we can't play with the dates over a period of time that we want to see. They limit the findings to the last 12 months only. I want to see trends visually over the last 90 days, or specific time periods that I put in, say from January 1-April 15. Allowing us to see specifics on time would in many cases be as good or better than boots on the ground. There are no systems that allow that of which I've come across, which is unfortunate. The NAR could put those features into their system but they won't. I already asked many years ago.

    862-208-2287Podcast Guest on Show #48
    1Reply
  • Brock Mogensen
    Real Estate Syndicator · Milwaukee, WI · Member since 2018 · 1k+ posts · 907 votes
    6y

    Population growth, median household income growth, median home price value growth, job growth, crime stats, and tenant turnover rate. 

    4Reply
  • Victor Vella
    Flipper/Rehabber · Cleveland, OH · Member since 2014 · 173 posts · 121 votes
    6y

    @Charlie MacPherson

    AI has been turning fastastic profits in the stock market for the last 10 years

    https://www.amazon.com/Flash-Boys-Wall-Street-Revolt/dp/0393351599

    Data is the new oil, that being said for bigger pockets yes local market data isn’t the right fit.

    2Reply
  • Shawn Coverdell
    Investor · Clatskanie, OR · Member since 2014 · 212 posts · 233 votes
    6y

    Your bank balance. Quit your job, Go on a long vacation. when you return, recheck your balance. if it went up your most important data is good. if it went down. Buy more properties and try again. 

    2Reply
  • Ellie Adams
    Real Estate Broker · Seattle, WA · Member since 2018 · 27 posts · 16 votes
    6y

    Accurate data need to be accurate :)

    1Reply
  • Mary White
    Rental Property Investor · Klamath Falls, OR · Member since 2016 · 146 posts · 213 votes
    6y

    My biggest frustration with Zillow is it's inability to account for multi-family properties. I have a 4-plex listed as being worth what one unit would be worth. For some reason they won't allow me to adjust the square footage to the honest amount. It's also been fully remodeled, but Zillow won't adjust the value to reflect this. This has greatly affected me getting a terribly low (over $100,000 less) appraisal value on the property since the appraiser directly referred to the Zestimate. It's a complete waste of my time to even look at it. 

    1Reply
  • Account Closed
    Wholesaler · Phoenix, AZ · Member since 2014 · 97 posts · 10 votes
    6y
    Originally posted by @Charlie MacPherson:
    Done - and a note of caution on comps. 

    Without deep local knowledge, it's not possible to deliver anything better than the random dart throw of Zillow - and those are often so far off as to be laughable.

    In the Boston area with an average sale price of $765K, being within 20% of the true market value 79.3% of the time means that the predicted error is $153,000 or less, except for 20.7% of the time where the error will be MORE than $153,000.  Not only is that a useless number, it's dangerous when people rely on it.

    This is something that we Realtors have to struggle to overcome.  We talk with sellers who have an inflated view of their home's value because "the Zestimate said $XXX,XXX", but we know it's not.  Investors who rely on these numbers can get hurt too.

    I still run comps manually.  It takes me 30-60 minutes to do them, but when I tell a seller what their home will sell for, I'm almost always well within 5% of the actual sale price.

    The bottom line on comps is that no data is better than bad data.




    Sounds like you really know your stuff, mind recording your screen and sharing with us how you do it ? 

    There is free programs that screen record like Apowersoft program

    0Reply
  • Account Closed
    Investor · Milwaukee, WI · Member since 2013 · 1k+ posts · 1k+ votes
    6y
    Originally posted by @Dave Meyer:

    If BiggerPockets could deliver you current, accurate market data, what would be the most important to you? Let us know in this 30 second survey. Thank you! 

    https://www.surveymonkey.com/r...

     I abandoned your survey because I couldn't get past the first question.

    I have real estate appraisal experience and like to think I have a pretty good insight to the question/problem.

    In fact, I could probably write a book on what I (like to think I) know that other people don't really entirely understand when it comes to real estate data, real estate data analysis and forming reliable conclusions from that data. (Not sure how many copies that book would sell - not exactly an exciting or fun topic)

    The answer to the first survey question is all of them are important and I would not even want to start an analysis without each of those data elements and others.

    To answer your question in another way, the data needed is accurate data. This will be problematic for bigger pockets and whoever decides they want to collect and sell data. Or more accurately, will be problematic for the users of that data, which of course turns into a problem for the seller of the data. Obtaining accurate, ongoing and comprehensive data is not easy because you can't simply write a program to go do it for you.

    Currently, the two most reliable sources for real estate data are MLS data and Appraisal data. Each of these compilations of data are produced and reported by a real estate professional with a professional obligation to personally collect the data and report it accurately. Do these compilations of data always get collected and reported accurately? No. However, they are both collected and reported more accurately more of the time than any other source, by a long-shot. The same could be said for other sources too, for instance a government office where the persons job is to accurately report a sale price in the public record.

    Is BP planning on competing with Corelogic?

    0Reply
  • Account Closed
    Investor · Milwaukee, WI · Member since 2013 · 1k+ posts · 1k+ votes
    6y

    Data is only as good as the analyst that forms a conclusion from it. And at the same time, an analysis is limited by the quantity and quality of data employed.

    1Reply
  • Account Closed
    Rental Property Investor · Houston, TX · Member since 2018 · 106 posts · 56 votes
    6y

    @Dave Meyer I am a senior data scientist in financial firm for my W2 job. We use all sort of data from yelp reviews to counting numbers of trash cans in bigger city to predict the rent growth of MFs in zip code or even the block level. 

    The prediction is as good as how clean and accurate the data is. I think the best way to approach is having the outlier detection system where you can map the rental growth sudden change YOY and population growth changes. 

    1Reply
  • Matthew Ewald
    Flipper/Rehabber · West Des Moines, IA · Member since 2019 · 27 posts · 9 votes
    6y

    @Dave Meyer just finished. I really appreciate how you are always trying to provide more information,tools ,and networks to allow people like me to better my real estate investing business. Thanks so much!

    0Reply
  • Adam Leitman Bailey
    Attorney · New York, NY · Member since 2019 · 72 posts · 27 votes
    6y

    I think competitor data is the most important. Competitors in the market, those fighting for the same house you're trying to buy, your companies competitors, anyone competing against you.

    0Reply
  • Rob Jones
    Rental Property Investor · Raleigh, NC · Member since 2016 · 109 posts · 59 votes
    6y

    Actual rental comp data.  It's very easy to see what someone is "asking" in rent for a property but what someone wants in rent and what someone actually receives in rent could be two different numbers.  Knowing this will be able to allow RE investors to have better actual rent data prior to making a purchase in a market they may be less familiar with.

    2Reply
Join the conversationCreate a free account to reply, vote on answers and follow this thread.
Log InJoin for Free