Route from the data or route from the windshield? I ran both on the same 40 minutes of streets

The part I'd want to know before voting is what your skip trace and mail cost per address actually is. If it's a dollar or two, the 16 noise logs cost you thirty bucks and the argument for filtering first mostly evaporates. If you're at eight or ten dollars all in per address with a real mail sequence, that noise is the whole budget. Your two lists aren't really competing on lead quality yet, they're competing on how expensive it is to be wrong.

@marlow not necessarily. A house with a 22 year owner and a tax flag can look completely fine from the street and still be a seller, because the distress is in the person and not the siding. I've walked into houses that looked crisp from the sidewalk and had a failed main line and forty year old panel behind the drywall.

What I'd add from the building side is that your blind pass is undercounting. Curb distress reads roof plane, paint, gutters, yard. It misses everything at the back of the structure and everything mechanical. So the 16 you wrote off as noise, some of those are owner occupied houses with a repair bill the owner can't face, which is a real motivation even though your filters and your eyes both said no. I'd keep them and just put them on the cheapest touch you have.

I've been going back and forth on the same thing and I think the geography answer swamped the lead quality answer for me. Pins four to a mile is what killed my second attempt. I was averaging maybe 14 pins an hour driving filtered scatter, and 30 to 35 addresses an hour just working streets in order. Even if the filtered addresses are twice as good per address, I'm still behind on volume, and I'm not established enough to be picky.

Where I landed, tentatively, is that the filter belongs at the neighborhood level and not the parcel level. Pick the subdivision by data, then use your eyes inside it. That way you never drive four minutes to a single pin.

@kestrel has the shape of it and I can put numbers on it because we run three part time drivers and I've had to pay for both versions.

We tested parcel level pins for six weeks and street level blind for six weeks in comparable areas. Pins produced fewer logs per hour by a wide margin, about 12 to 15 versus 30 plus, and the response rate on the mail was better, roughly 1.6x. Multiply it out and blind driving won on contracts per driver hour, though not by as much as I expected, and it lost on cost per contract because we were paying data and mail fees on a much wider list.

The deciding variable for us wasn't lead quality at all. It was data quality in the county. Where the assessor file is clean and current and delinquency is published in a usable form, filtering first is genuinely good and the pins are worth driving. Where the county is behind, or where sale prices aren't public at all, and roughly a dozen states don't publish them, your filters are running on stale mailing addresses and guessed equity. In those markets the windshield is the better sensor and we drive blind.

So I'd stop asking which method is right and go look at your county's last file update date and what fields are actually populated. That answers it for your market.

One process note. We route the spine off flagged pins only when the pin density is above roughly 8 per street mile. Below that the driver spends the day at intersections and the whole thing falls apart. Your scatter at four per mile is under the line, which is why Sunday felt bad.

I had to read @cobalt twice. So the argument isn't about eyes versus data in general, it's about whether the records in your particular county are good enough to be worth trusting. That reframing helped me a lot, thank you. Does the file update date usually show on the assessor site itself, or do you have to call and ask?

On land this whole question inverts and it might be useful as a contrast. There's nothing at the curb to read on a vacant parcel. No roof, no gutters, no mail piling up. So the physical observation collapses down to access, brush, dumping, maybe a fence line, and the data has to carry almost all of the weight. I drive to confirm access and topography after the list tells me who to look at, and I'd never drive blind for land.

Which makes me think the blind versus filtered split isn't really a philosophy. It tracks how much signal the structure gives off. Houses radiate a lot. Dirt radiates almost none.

@juniper the update date is on the site for most counties I've dealt with, usually buried in a footer or a metadata page, and it varies county to county. Where it isn't posted, calling works and they'll usually just tell you.

@windlass one thing that would change my vote is what you do with the 23 pins that looked fine. If you drop them, filtered driving is pure waste on those. If you log them as verified-not-distressed and keep them out of the next six months of routes, you bought yourself a permanent exclusion list and the drive paid for itself in time you don't spend later.

I'm building a service around this and the exclusion data is the part clients keep underrating. Half the value of a driver isn't the addresses they hand you. It's the streets they can tell you not to bother with again.