Why Automated Domain Appraisals are Unreliable (An illustrated example)

[Joe123]

Experienced domainers know automated domain appraisals from Estibot, GoDaddy and others are not reliable. The associated comparable sales (or lack of them) are more valuable as they are real data points. Here is an extreme example of a wayward valuation:

(NOTE: I have no relationship to this domain name)

tyre.co.in
co.in is a second level Indian extension that, until 2005, was India's only commercial extension. co.in is now overshadowed by .in although many companies still use co.in. "tyre" is what people in the USA call "tire". Let's see GoDaddy's valuation:

That seems rather cheap but does reflect the low number of reported co.in sales. Namebio has 124 co.in sales versus .in's 945 although many sales go unreported.

Now let's see the appraisal's comparable sales...
.
.
.
.
.
.
.
.
...drum roll....
.
.
.

What the...!!!!!!!! The same domain previously sold for $20,000????

That late 2017 sale is 140 times the current appraisal value (and the co.in market has not imploded in this time).

Relying on the appraisal alone would have left so much money on the table. Experience, knowledge, patience and gut feel make the difference.

 

[biggie]

Hi

about unreliability, I agree

but some reported sales, may not have happened.


I know, first hand

imo...

 

[Joe123]

Hi

about unreliability, I agree

but some reported sales, may not have happened.


I know, first hand

imo...

I agree with that although the sale in the example did happen.

 

[lock]

Just because someone paid 20k once doesn't mean they will again the next will likely offer less than half. The data offered at estibot eg searches is completely accurate. A bot is only a mathematical equation that left his wallet in the car. The price well that is only cpc x estimated traffic if you can sell keywords it will find a value. Have my own valuation.site and the price won't be right.

 

[Joe123]

Just because someone paid 20k once doesn't mean they will again the next will likely offer less than half. The data offered at estibot eg searches is completely accurate. A bot is only a mathematical equation that left his wallet in the car. The price well that is only cpc x estimated traffic if you can sell keywords it will find a value. Have my own valuation.site and the price won't be right.

Yes, the valuation bot results are only the result of their inputs and equations. Domains, especially for extensions with less reported sales, need human input to provide a better value. Even then, nobody knows exactly what an end user will pay for it.

 

[lock]

Yes, the valuation bot results are only the result of their inputs and equations. Domains, especially for extensions with less reported sales, need human input to provide a better value. Even then, nobody knows exactly what an end user will pay for it.

What a human paid for it one time doesn't influence the equation it isn't calculated that way and it really shouldn't be an influencing factor. It doesn't need to be accurate just indicate value.

 

[Joe123]

What a human paid for it one time doesn't influence the equation it isn't calculated that way and it really shouldn't be an influencing factor. It doesn't need to be accurate just indicate value.

And the only real indicator of value is how much a domain sells for. Everything else is conjecture. If a domain sells again for less or more, then that is its new value.

 

[lock]

And the only real indicator of value is how much a domain sells for. Everything else is conjecture. If a domain sells again for less or more, then that is its new value.

But it simply isn't the case I would be one for locking a price but history shows if you abandon a project the next isn't quite as generous. It is an appraisal which is a speculation you acknowledge it shouldn't be updated based on actual data as it isn't based on that but traffic

 

[Joe123]

But it simply isn't the case I would be one for locking a price but history shows if you abandon a project the next isn't quite as generous. It is an appraisal which is a speculation you acknowledge it shouldn't be updated based on actual data as it isn't based on that but traffic

Traffic used to be a big valuation factor when parking was lucrative but now it isn't anywhere near as important. What evidence is there that current valuations are based on traffic?

 

[lock]

Why mention cpc in result of estibot if it isn't a factor.

 

[Joe123]

Why mention cpc in result of estibot if it isn't a factor.

CPC is unrelated to traffic

 

[lock]

cost per click multiplied by the estimated traffic. It is only an algorithm it may not suit you or anyone but it is how it is given an appraisal. If you can't over look the price and gain from the data it will never be of benefit. You can always ignore them like 90% of domainers and make your own price.

 

[xynames]

See:

I find the GoDaddy appraisal tool to be a good start for pricing two word exact-match low to mid four figure value domains. It tends to come in with a good starting point and sometimes isn’t even that far off.

It also, for these types of domains, often gives you some great closed comparable sales.

When GD’s valuation is off it is often because it couldn’t find proper comparables. Look at the comparables GD spits out - if they make sense and look comparable to you, then you may put more trust in the appraisal. If they appear unrelatedly wacky, then the appraisal is probably off too.

Even after GD gives you what appears to be a trustworthy valuation still there is some old hand manual valuation research you must do before you decide on a domain’s price.

 

[twiki]

Interesting subject.

So I am the proud owner of sort of a nextgen appraisal tool (private) designed for/by a .com domainer. (Considering making it public but haven't decided yet, there are significant cons to that.)

It is better than the tools out there for what I need . I can rely on it in most situations, but not on the public ones that in most cases bring up completely junk results. What I can tell you though is that it uses a far different appraisal method than what's out there.

I recently ran a thorough check against a large master list of domains and their actual sales. My tool has been 98% accurate overall (as in total value) which was a tremendous surprise to me (I knew it's rather good but didn't expect this much). However there are differences of course from domain to domain. Out of 4 domains, 2 are very close to the real sales value, while 2 are either too small or too big. Usually the difference is about 50% to 75%. So I'm getting to an overall 50% precision which is not that bad, I would say.

The tool doesn't appraise anything though. It is precise on .coms of 2 or more domains, which is what it was intended for. Can also do .net and .org with a far lesser precision.

My takeouts:

- It is entirely possible to make a better appraisal tool (I've done it for my own usage), but NOT for everything out there. There are significant differences in algo/design based on the type of domains involved. ngtlds and .coms are completely different in this respect. cctld's are also far different from each other.

- Building such a tool (properly) involves much more work than one would imagine. It took me one year of hard work and will still have to improve it further.

- Brandables are close to impossible to be appraised by an automated tool. That is for humans only.

- Public / known automated appraisal tools just haven't figured it out yet.

- Single word domains (either precise or with modifiers - "ify) cannot be precisely appraised by an automated tool. The reason is simple, each of these domains is unique so the price is unique, so data learning will not appraise them correctly by any reasonable margin.

- Edit: It is close to impossible to have an accurate appraisal of a domain in the high 4-digit or 5-digit range. Such high value domains are unique and if you ever look at any existing sales data, you'll see how different they are. It's all in the eye of the beholder. But for lower values, it is definitely possible. My tool works precisely for anything under 4K, including $200 domains if you want.

- Some interesting aspects from the data run were, although short domains make the bulk of the sales and overall value (domain length under 11 or 12 characters), lengthier 2-word domain are often sold for more. My tool adjusts for that, and I was expecting longer domains to be very cheap, well, it's not the case. 12 to 18 letters seems to be a high-priced range for 2-words or 3-words.


Note, there is a general impression that all appraisal tools are (and will always be) worthless. Which is an incorrect assumption. It's just that existing market appraisal tools are not yet to what would be a new standard in the future, in this line.

 

[twiki]

What the...!!!!!!!! The same domain previously sold for $20,000????
That late 2017 sale is 140 times the current appraisal value (and the co.in market has not imploded in this time).

This shows my point above. The appraisal tools haven't quite gotten it, especially GD here, they aren't even consistent with their own data as precisely visible in this example.