In this episode, the automated comp search infrastructure for LandPricer is explained, connecting to both Redfin and Zillow APIs with a 5-mile search radius that returns maximum 150 properties (50 each: sold/pending/active), then filters to a final 20 comps using distance, acreage range, and recency prioritization based on five years of operational comp selection experience.
Key Takeaways:
- Sold Comps Prioritized Over Pending Despite Market Activity While pending comps show current market activity, sold transactions provide definitive price data, making them more reliable for software-based pricing despite pending comps being valuable for trend analysis.
- Hard Acreage Filters Eliminate Irrelevant Comparables A 5-acre subject property automatically excludes 100+ acre comps from consideration, tightening the dataset before applying secondary distance and recency filters.
- Pending Properties Over 6 Months Are Hard-Filtered As Errors Any pending status exceeding six months gets automatically eliminated as likely listing engine errors from Redfin or Zillow rather than legitimate market data.
Listen to the full episode to understand the complete automated comp selection and ranking methodology.
(Podcast transcript below)
Welcome to get serious wanted to provide another update on what my engineering team is working on for land Pricer and really how we are automating the comp search and identification process. So I know there’s a number of tools that will kind of auto sort comps for you and they’ll have different approaches on how to handle this. You know, we try to introduce our own
nuance for selecting which features are going to be the most relevant for, for comps. And, there’s always room for improvement, but I’ve tried to use our past, you know, five plus years of, operations, know, successful operations to inform how we’re selecting the most reliable comps. so to boil things down,
I know there are various different sites for the first iteration of our…
software is we’re only going to be connecting to Redfin and Zillow. And generally for sold and pending comps, I prefer Redfin just their overall UI tends to be a bit better. It’s also easier to sort for pending comps on Redfin compared to Zillow as well. And then I prefer using Zillow for active comps because you get the views and the saves.
which tends to be pretty informative for activity, though definitely not a hard and fast rule. So it needs to be weighted accordingly. So we set up a scraper for both Redfin and Zillow. If you input a particular address, ideally the target property, but if it doesn’t have an assigned address, then you have to use a nearby property that doesn’t have assigned address. really
doesn’t matter, you we do that internally either way, like it’s going to be close enough. We just need kind of a identification point. then we can, build in a search radius for what we’re looking for. And so on the top end of the backend, you know, we’ll have a roughly five mile radius. That’s usually good enough for most areas of the country.
to find at least a handful of land comps that are going to be more relevant in more rural areas. It’s possible you could go out further to like 10-ish miles radius. So it’s possible we might need to adjust, but when you’re in super infill crazy areas like say Lee or Charlotte County in Florida, mean, you’re get hundreds of comps to look through.
So we had to kind of pick a middle ground to start out with that we can refine further and Then we are going to be searching for sold comps pending comps and active comps and Because you know the API calls are basically how much we’re gonna have to pay to scrape all this data can get More significant if you’re pulling in, you know
thousands of properties for some of these high volume comp areas, you need to limit that, those number of hits. So I was thinking, you know, like if we could get, you know, a max of 50 solds, pending and actives, like 150 total, that would be more than enough for, you know, any possible property that you’re reviewing. And then we would, we would shorten that list even more to only do the highest priority comps. And so for the sold comps,
in terms of identifying, okay, what of those top 50 that we should look at, we’re setting it up to be by distance and same for the pending and the active comps. So then we can ensure that, you know, we’re reasonably confident we’re going to get the most.
reliable comps for that particular property that we’re looking at. And even if we have to exclude a number of other ones in the area, they’re probably not going to be as relevant. Again, if you’re getting like 50 plus comps for any property, like you probably have a surplus of data to work with anyway. So I’m not too concerned that it needs to be absolutely perfect in that regard.
Compared to more rural areas that you’re just going to have less comps to work with and that’s where like these filters are going to be Even more applicable in terms of like how we prioritize the comps So, you know once we decide okay, we have like a max of 50 comps for our initial list Then we actually need to decide on a final 20 comps max that we are providing to the user for landpricer to consider just as
comps for the particular subject property they’re looking at. at minimum, one comp needs to be found. If there are truly no comps in the area, I mean, it’s pretty rare, but occasionally that is the case. like, you can’t price properties without comps. So at least one needs to be found. And ideally we want at least five for any given property and max of 20. So then the user, if they really want to go in more depth,
more complex property, they can choose to review those additional comps that are found for the property or they can add their own. So that was our thinking behind it. I’ll skip some of the details, but like the display is going to be easy to work with. So you can quickly tell what type of comps you’re looking at and various key factors here. But in terms of how we’re deciding, okay, what 20 comps to display if there was.
that amount or even if there’s less than 20, how we decide to rank order them. I’m always going to default to sold comps first. I know some folks will prioritize pending comps as the number one to look at. And I do love to look at pending comps whenever they’re available. The only thing is, is that with pending, we’re not certain what that final close price is. Usually we can get a reasonable estimate and we have that baked into our backend.
what that price is going to be. But it’s not as reliable as a property that has fully closed. know, pennies are great to see the activity in the market, but until it again, it fully closes, we don’t have that final data point. So I always prefer, you know, solds as kind of more definitive data. I know in non-disclosure states that can be a bit trickier. So you still have to estimate what that might be. then, know, pending comps can be more similar to sold comps.
But ultimately when you’re dealing with software, you got to draw the line somewhere with what you think is most valuable. And to me, I would just prioritize sold comps as the number one. And generally in areas that it’s going to be tricky to review properties or there’s just not that much data to work with, you’d be lucky to find 20 relevant comps in an area. you’re almost certainly going to have…
sold and pending comps to look at regardless. So as we’re determining what those final 20 comps are to look at, first we just have a hard acreage filter. So, you know, I won’t get into the granular details just reading off these things, but, depending on whatever the subject property acreage is or what the proposed child parcel average acreage would be, if you’re looking at a possible subdivide, that acreage is going to
be part of the filter process so that we only look at certain acreage ranges for the target property or potential child parcel property that is under consideration. So it’s like, you know, if we’re assessing a five acre property, it doesn’t really make sense to review comps that are north of, you a hundred acres. So we tighten up those filters, get rid of all the comps.
comps that showed up in our search that don’t fit that acreage range. And then we have to sort them from additional factors. So for sold comps, really the three things that are most important to me looking, if I was just filtering down my comp list are how recent was the sell date? What is the actual acreage of the property? How close is it to the target property?
acreage that we’re looking at and how close is the property to the target. So, you know, over time, we’ll probably develop a scoring system that kind of blends those three together. But, you know, to keep it relatively simple and in most cases, like this is what I’m going to be prioritizing anyway, I really want to look at that last or most recent sell date.
He has, especially in again, a shifting economy, land changes, you know, especially over the last five years, relatively quickly in a lot of areas or, you know, especially over the last like six to eight months or so, you know, we that kind of mini bull market for, you know, end of 2024, early 2025 and, you know, post early April, things got a lot tougher. I think a lot of people would agree with that. so like, I just, value.
recency of sales to be my most reliable metric when I’m looking at sold comps. And then if I was to prioritize after that, I would consider how close it is to the target property and then acreage as the third option. You acreage, already did a hard filter anyway to put us in range. So, you know, we’ve already kind of done some of the work to orient around that. So, you know, we would
set up our filter for our sold comps and, you know, prioritize by last sell date. And let’s say there was a ton of sold comps available, you know, 20 plus that fit within that acreage range. Then we are only going to display those sold comps to the user because, we only, don’t want to overwhelm the user. So, you know, 20 comps is already probably more than enough for pretty much any property. So if there’s
tons of sold comps available, great. We’re gonna just display those. But if there is a deficiency of 20 sold comps that we can filter out to display to the user, then we would move on to pending properties. And obviously, pendings haven’t sold yet. So we would sort by distance. I guess you could filter by
date the property went pending, but you know the fact that it is pending anyway, like usually it’s going to be within the last 60 days roughly, and we eliminate any property that was pending for longer than six months. So that is another.
Actually, we can probably build that in. that I’m actually, whenever I do these podcasts, sometimes I’m thinking, okay, yeah, we could actually improve this a little bit. I would probably have another hard filter for pending properties to eliminate any that have been pending for longer than six months, because that’s usually an error on behalf of the listing engine, know, Zillow or Redfin. In this case, we’d be using Redfin. So we’ll have that built in, but just oriented towards distance to the target property as the primary. And then same for active properties sorted by
distance that already had the hard filters there. And so then we’ll just go in priority, solds, dependings to actives and display that to the user with the top five auto selected for additional review on behalf of the user. And then they can add additional ones if they want, or they could check mark extra comps if they would like, assuming there is more than five comps that.
are found for a particular target property. So I just wanted to, again, give a little bit of a behind the scenes look at how we are designing the the back end of LandPrice or here. You know, again, it’s based on how I’ve reviewed properties successfully for many years now and how we would do it in our own business, even outside of this. Obviously, the first
Version of Land Pricer didn’t pass the user test just because it was way too slow. So it’s like, okay, I just have to speed up all this, just automate everything, the Comp Search, et cetera. Make it super easy for the user to just move through it. And it follows our criteria that we would be doing manually anyway. So that is the name of the game. So you’re getting an exact look on how we would do this regardless. If you wanna keep doing it manually, that’s the playbook for you. But…
If you want to give them more automated route a shot to be more reliable with your Pricing make a bit easier. Hope you give us a shot once we release soon with that in mind serious land capital for any of your funding needs land daily diligence Facebook group for zero cost review of your land deals and landpricer.ai, which I just talked about today Subscribe and share I will talk to you all next time
Take care now, bye.
