Saturday, July 7, 2018

Should Automated Domain Worth Tools Be Extension Specific?

Automated Domain Worth Estimation


As those who follow my posts on NamePros and elsewhere will know, unlike some domain investors, I do see value, despite their  limitations, in automated domain name appraisal tools such as Estibot and GoDaddy GoValue.  Even though these should never be the only or even primary estimates of worth, I think that domain investors and potential purchasers should know the values for domain names they are considering. Therefore, in most cases I show these values in my listings at NamesThat.win (along with the caveat that values are not static, and potential purchasers should check for the most recent estimates if this information is important to them).

As artificial intelligence, and in particular machine learning, progresses, these automated tools will become even better. I simply do not accept that a human is always better at the task of evaluating the worth of domain names and always will be. Even today Google Maps does a pretty amazing job of finding a good route from point A to point B in a complex city, whether I want to go by walking, public transit, car, or bicycle. Is evaluating a domain name inherently a more complex task? A tool that has access to all of the data on past domain sales, Google searches, etc. is well placed to make a good estimation of domain name value.

As behavioural economists remind us, humans are subject to numerous cognitive biases and blindspots. Yes, I value the opinion of domain name investors, occasionally asking for them on the Appraisal section of NamePros, and not infrequently leaving my opinions for others. But I also place value on objective estimates from automated tools. Both human and automated estimates are sometimes very wrong, or at least predict values far different than sale prices.  That is because ultimately the worth of a domain is what someone will pay for it, and that will never be an exact science.

As a scientist I know that an estimate is improved by taking into account multiple independent values, and that is how I see automated tools and opinions of experienced domain investors co-existing. This is a long lead in to a post on a new extension specific approach to domain name evaluation.


New APP Specific Domain Name Estimation Tool


NamePros user un1 on July 6, 2018 posted an offer to use an Excel based tool developed to estimate the worth of APP extension domain names.  The tool works only for the single extension APP, since it takes into account

  • how related the domain name is to the names of apps on Google Play or the iOs App Store
  • how many downloads those apps have

Among the other aspects of the evaluation the poster lists the following.
  • search result performance
  • TLD penetration (how many extensions the word is registered in)
  • traffic rank of the most similar domain with the highest rank (for example, for the appraisal of abc.app, the traffic rank of abc.com)
  • synthesis of appraisal results from different services (the poster does not specify which, but I presume at least GoValue and Estibot, possibly others)
  • length of the domain name
  • trademark strength  
  • meaningfulness
The scale is a mix of quantitative measures automatically assigned, and ratings assigned qualitatively.  For example, the developer explains that the trademark strength is assigned using the following Likert scale.

1 -> generic phrases or random letter combinations which are not used as trademarks (except possibly in a very regional micro company). .
2 -> deformed versions of strong trademarks (e.g. "itun") and/or out-of-context trademarks (e.g. "kandle")
3 -> well-known trademarks with no relevant apps released and/or no "mobile", "app-related" orientation (e.g. "baze") or generic names of mobile apps (e.g. "squadd", "dancingline" etc.)
4 -> well-known trademarks with apps released (or at least app-related orientation): e.g. "virgingames", "storspelare" etc. Rule: The sole activity is not the mobile app; the app is just the extension to the existing business.
5 -> well-known "mobile-first" trademarks (i.e. the main area of activity is 'app-related' services): e.g. "heimlich", "bambo", "xpro" etc.

Dollars or Scores


A number of people have commented on NamePros at various times that they find the automated tools worthwhile in expressing relative worth (i.e., whether name A is better than name B), even though they have no confidence in the dollar values given by the tools.  One aspect of the new tool is that it does not try to give a dollar figure for the evaluation, but rather a score on a 5 point scale, with a more valuable domain name being near the
  1. host.app = 4.73 (reported sale: $14,162) 
  2. RN.app = 4.42 (reported sale: $15,000) 
  3. HedgeFund.app = 3.13 (reported sale: $251) 
  4. YouProductive.app = 1.63 (reported sale: $1,300)
He does point out how his tool ranks those APP domain names that already have a NameBio recorded sales price.  As .app extension domain name sales increase, there will obviously be more data points both to evaluate the quality of the tool and also to convert points ratings into dollar value ranges.


Concluding Thoughts


I see a lot of value in a domain extension specific approach to automated evaluation.  GoValue currently rates (almost all) new domain extensions equally, and clearly this can't be right.  For example, when I checked just now it told me that pepperoni.pizza and pepperoni.science are worth exactly the same, $566.  That is clearly wrong!

Estibot do try to evaluate different extensions differently.  Although they do not disclose their detailed methodology, at least the way they structure their reports seems to indicate that in some cases (e.g., extensions like science) they span the dot and for their search statistics look at searches for the two parts together.  That is, for a word like planetary.science they would look at how searched (and advertised) "planetary science" was (as well as simply the word planetary).  While this is sensible here, for the new .icu (I See You) extension they seem to be doing the same as for extensions like science and spaace, and searching on the combination across the dot.  This does not make sense for this extension.

If the automated tool restricts itself to one extension, it can look at parameters important to that extension only.  For example, when evaluating .loan or .loans, it could look at how frequently that type of loan appears in commerce.

With well over a thousand extensions, even if country code extensions are not included, it would become chaotic to treat all individually of course.  However, some could be grouped together, and for some other extensions with limited registrations and sales, it is perhaps not sensible to try an automated worth estimate.

I welcome this new contribution to semi-automated domain worth, and look forward to seeing results from its application. I think several of the features, concentration on one extension only, use of a hybrid automated and human approach, and use of a scale rather than a dollar value, make it innovative and interesting.

Original post July 7, 2018.
Additional information added July 9, 2018.
Slight formatting change July 30, 2018 (no content change).

I have no association with the developer of this, or any other, automated domain value tools. The information on what parameters are used is taken from the original post.  You are responsible for your personal use of the information from any of these tools. This post is for information purposes only. The image that accompanies this posting is by Pixabay user geralt.

1 comment:

  1. I have extended and refined the original post using some additional information provided by the developer on NamePros.

    ReplyDelete

Featured Posts

Why Would Anyone Want a New Extension Domain Name?

Background A NamePros user asked this question: "Why (would) an end-user would want to invest in a ngTLD versus a .com domain nam...