Item — HN Reader

You are correct, it is a correlation, not a causation. I think I mentioned that in the disclaimers.

It is really hard to control for all the variables (price, motivation of seller, location, who the listing agent knows, what style of home is hot), but all other things being equal, it's better to have more photos for a listing. In fact, since it is one of the easiest things to control (it's a lot easier to snap 25 photos than change the price point of a house) I'd argue it is one of the things home sellers should expect.

Would love to hear your suggestions on making this more rigorous. Off the top of my head, it'd be interesting to list a set of same homes twice, once with few photos and once with many photos, and see what the difference in response was. Unfortunately, that is a bit beyond my scope for the time being.

1 comments

> Would love to hear your suggestions on making this more rigorous.

That's easy:

1. A scientific study differs from an unscientific one in a number of ways, one of which is the presence of a control group.

2. To create a control for this hypothesis, create one MLS listing with photos, and another MLS listing without (if that's possible). See which one produces the desired result.

3. In lieu of (2) above, if that's not practical, do a classic Web A/B test -- create two versions of a web-based advertisement for a property, one with photos, one without (or one with more and one with fewer photos), each with a click-through. Then simply count clicks. This way, you don't have to track the outcomes, only logged clicks, which means you get results faster.

Item (3) above can also test varying advertising content, not just the photo issue. It's quite common to use A/B testing in Web design, so some of the infrastructure may already be present in the server. The key to the method is to randomly choose between two or three page designs when an URL request arrives at the server, and track the outcomes keyed to the designs.

This approach would give you a more scientific result, because it's based on a comparison in which the only thing changed is the thing being measured -- same property, different number of photos. Same property, different description.

Before closing, I have to tell my favorite real estate joke (which I invented). An agent visits a "distressed" property and is almost overwhelmed by the smell of the interior -- it literally smells as though there are mushrooms growing in the walls.

But our hero always thinks positively. He returns to the office and begins writing the listing. A long pause as he tries to think of a way to put a positive spin on the property. Finally a light bulb appears over his head, and he begins typing. The first word: "Breathtaking!"

Thanks for the joke!

I appreciate your comments. I like the idea of option number two, because it still gets to the end goal--a sale. It'd be a bigger project than I want to take on, though.

Option number three gauges interest in a scientific manner, but assumes that interest is a valid proxy for the end goal of a sale. I'm not sure how true that assumption is.

I've window shopped homes that I had no intention (nor capability) of buying. My employer one had a record day because one of the homes on the site was a football player's and had gotten mentioned in an ESPN blog. In both cases, interest was only vaguely related to purchase capability.

Testing all the way to the end of the funnel (home sale) is key, and difficult.

> It'd be a bigger project than I want to take on, though.

In that case, use the above outline as a litmus test for those who claim real evidence for a particular strategy -- just ask, "Have you done any A/B testing?" It has the advantage of putting the conversation on a quasi-scientific footing, and it impresses people too. :)

> Option number three gauges interest in a scientific manner, but assumes that interest is a valid proxy for the end goal of a sale. I'm not sure how true that assumption is.

Yes, fair enough. But remember that online advertising pays by the click, so someone, somewhere believes it's a valid measure of future sales, even if the click is all there is.

Maybe 500 people click, and only one buys. The outcome is still related to the click-through rate, and the pay-per-click scheme still works. If there was any serious doubt about the value of paying for clicks, people would refuse to pay -- they would try some other approach to online advertising and metrics.

> Testing all the way to the end of the funnel (home sale) is key, and difficult.

Absolutely. You would have to interrogate the buyers, and I venture to guess no one will want to push that idea.