Domus Analytics

A Love/Hate of Polygons

September 02, 2021 03:04 PM Comment(s) By Amy

Polygon maps are great for search, but fall short for stats

One of the questions potential clients ask us is "do you have polygon maps?", and the answer is no.  For some groups, that's a must-have on their vendor evaluation checklist, and that's a mistake.  While polygon selects can be a great tool for helping a client search for a home, they too often come up short for market statistics.  I thought I'd use our recent home shopping experience to illustrate this point.

We're what you might call "regular home shoppers".  That is, we aren't actively looking for a new home, but we love open houses, and if we ever happen upon the perfect house then we just might buy it.  We've found three houses this way in the last 10 years.  The first one was just out of our budget, and the second house was already under contract by the time we saw it.  We found the third house a few weeks ago.  We ultimately decided it wasn't exactly the right house for us, but it was that experience that crystalized our position against using polygons for market analysis and statistics.

Here's how it went down.  During a lengthy chat with the listing agent, she offered to create a comp analysis for our current residence in case selling was a contingency for us to make an offer.  To give you some background, there are 3 neighborhoods right next to each other where we live.  The listing agent's primary business is in one of those neighborhoods, the house we looked at was in the second neighborhood, and our home is in the third.  These are not subdivisions of different builders that no one but the owners can tell apart - these are three very distinct neighborhoods with very different pricing.  It is very obvious, especially to the agent who works in one of them ALL THE TIME, where each neighborhood starts and ends.  However, the MLS doesn't capture neighborhood data, so the agent pulled comps for us using a polygon map.  Unfortunately, the map she drew accidentally included a street in her primary (and much higher priced) neighborhood, dramatically skewing the results and making the comp analysis useless.

Now you may be saying "it was a mistake, just redraw the map."  But that's the whole point.  It should have been trivially easy for an agent who was an expert in our area to draw the map correctly.  If she can't get it right--an experienced agent who knows the area incredibly well--what hope is there for an agent who isn't as well versed in this side of town or technologically savvy?  They may seem cool and powerful on the surface, but from a technical perspective, polygon maps are a lazy work-around for incomplete data management.  Don't let the razzle dazzle fool you - the MLS platforms are responsible for maintaining quality, persistent, useable data that can be consumed by all agents.  We've lived in our house for 18 years, and our neighborhood hasn't changed one street in that time--why make every agent draw a polygon every time they want to analyze the neighborhood?  That local, specific knowledge should be defined once in the MLS and made available to all agents.  Polygons force agents to execute a low-value manual step that is unnecessarily complex, repetitive, and time-consuming, which is obviously prone to errors.

What are the best practices for stats if it's not Polygon Maps?

Market Stats are fundamentally different than Search.  Polygon maps can be an incredibly useful tool for searching for a specific property for a specific client in a very specific area where the goal is to narrow down the criteria as much as possible.  Let me say that again: Polygons are useful for narrowing way down.  Stats are the exact opposite: you want a well-understood selection criteria, including the most listings possible that are economically similar.

Here's how to create an economic analysis of an "area" not easily selected from the data:

  • Start with one or more pre-defined areas: MLS area, township, city, zip code, school district, etc.  Whatever is specific, economically similar, and well-known.  This will enable comparison, repetition, and ease of use.  This type of analysis can even include multiple non-congruent areas.  You can't do that with a polygon!  Maybe some of those similar areas are just a few streets over, or even on the other side of town.
  • Narrow down your analysis using one or more filters like price range, year built, or size (lot or living area).  In many situations, listings in the same geography are economically different.  Using the additional filters accounts for those differences.

We haven't even touched on other use cases for market stats where polygons are useless:  How to compare two areas against each other?  What about 5 areas?  or 10 areas?  Can you imagine having to draw 10 different polygon maps just to create one chart?  How do I create a selection area of discontinuous space?  What about excluding an area inside an area, like a donut hole?  I could go on, but you get the idea.

The next time you hear someone say "polygons are a critical feature for stats", ask about their use cases.  I bet they'll be hard pressed to come up with a situation that can't be solved in a much easier, more stable and replicable manner like we've described.  Polygons have a place in today's real estate tools, but that place isn't the stats.

In the meantime, our search for the perfect house continues...