Sunday, August 12, 2018

NNNNN Domain Names

While it is challenging  to estimate the worth of many domain name types, short numbered domains are somewhat easier to predict.  In the NameBio report for August 11, 2018 there were 30 sales of NNNNN (five number) domain names.  With a common day and sales venue (NameJet), this offered an opportunity to study a homogeneous dataset.  In this report I look at the sales prices, and place them within a historical context. I also see how effective  GoDaddy GoValue is in predicting sales prices for domain names.

Recent Sales

With so many NNNNN sales in a single NameBio daily report, I decided to take a closer look at the prices paid.  There were 30 sales of NNNNN domain names in this daily report, ranging in price from $490 to $1500, with an average price of $654.  The majority of sales were at $500, and that was also the median price. You can see just the list of just the  5 number  domain names from that day at this link.

Several numbered domains did go for prices significantly higher than the average on this day.  The two that went for the highest price ($1500), both contain the sequence 23 and also the number 7.  Many people and cultural traditions find the number 7 to be magical or lucky, although in Chinese culture it is also treated with suspicion.  It is easy to see why the lowest priced domain went for a bit less - it starts with a 0 which is generally less desired both aesthetically and because of possible confusion with letter O.

I do not have the expertise to comment in detail on coded messages in certain numbers.  This fascinating article on Chinese messages in numbered domains will help you understand that phenomena.

Historical Valuations

We thought it would be interesting to take a  brief look at how prices for these recent sales of five number domain names compare to prices for similar digital assets in the past.  The average prices paid for domain names is given below.  The information has been obtained from the NameBio database.

  • 2018 YTD  $952 (range  $151 to $22,383, with 2232 sales)
  • 2017 $959 (range $134 to $25,500, with 2454 sales)
  • 2016 $1098 (range $100 to $34,501, with 2460 sales)
  • last 5 years $1056 (range $100 to $245,000, with average 2274 sales per year)
  • all time $1051 (9343 sales)
The clear picture is that several thousand NNNNN domain names sell each year (at least as listed on NameBio, there are undoubtedly many more that sell on Chinese venues not reported to NameBio) with the average price in the range from  $900 to $1100.  Within this context the recent  sales are on the low side, but the average prices in the historical data are heavily influenced by sales of highly desired sequences. There is some suggestion in the historical data that the average price of NNNNN domain  names is decreasing slightly, probably a reflection from the general  cooling in the Chinese domain name market.

All Time Top NNNNN Sales

Let's take a  look at the highest sales of all time for NameBio recorded domain names.  You can find the full  list at this link
  1. 88888  sold at Uniregistry for $245,000 in Dec 2013
  2. 17176 sold at Uniregistry for $55,000 in Feb 2015
  3. 10023 sold privately for $50,000 in Feb 2015
  4. 83333 sold at Uniregistry for $43,000 in Aug 2015
  5. 70777 sold at NameJet for $34,501 in Jan 2016
  6. 88808 sold  at Uniregistry for $30,000 in Aug 2015
  7. 77787 sold at SnapNames for $26,055 in Oct 2015
  8. 79000 sold at GoDaddy for $25,500 in May 2017
  9. 55588 sold at DropCatch for $25,050 in Mar 2016
  10. 97088 sold  at NameJet for $22,500 in Feb  2016

While the largest sale was in 2013, the peak in numbers of high value NNNNN sales was in 2015. While 4 of the high sales include at least 88 (a lucky number in China), and the highest sale all 8's, various other combinations occur in the leading sales.  A number of  different venues have accounted for the sales, with Uniregistry leading, particularly in the early years.

GoValue Estimates of Worth

Since these are liquid assets, frequently traded, and  somewhat homogeneous (although with preferences for certain numbers and sequences), they should be easier to predict prices for than say multiple word (or even single word) domain names.  I had a look at GoDaddy GoValue estimates of worth for each of the domain names sold yesterday (with older sales it i difficult to be sure that the sale value did not alter the estimated worth).

What I found is in the table on the right (click on the table for a higher resolution view).  For each domain name sold (all on same day and venue) I give the name, the GoDaddy GoValue estimate of worth, and the sales price recorded.  I also compute for each the ratio of the two values – e.g. if a domain name sold  for $1400 and GoValue estimated its worth at $700 the ratio would be 2.

If you go to the bottom of the table, I compute the average and standard deviation for each column.  You can see that there is no evidence of bias, with the average prices in the GoValue estimates of worth and the sales values almost identical at $648 and $654.  This implies that the domain names recently sold did in fact trade at about the correct value.

If we look at individual values, most are at ratios of 0.50 to 1.50 indicating to me that the worths of NNNNN can  be estimated moderately precisely by automated means (although see next paragraph).  Many sales were within 20% of  their estimates of worth.  However, there are exceptions. For example GoValue predicted that 47601 was only worth $167 (it sold at the median price of $500), and the name 23757 that tied for the highest price ($1500) was only projected to have a worth of $700 by GoValue.

If I plot these data,  I obtain the result shown at the top of this blog post. There is only a weak correlation between the GoValue estimate of worth and the actual sales price.  As background an R2 value of 1.00 means perfectly correlated data, while 0.00 would mean no correlation.  This dataset with R2=0.11 is pretty weakly correlated. I actually tried a few different types of fits to the data (linear, polynomial, logarithmic, exponential) and while a couple were marginally better than the linear fit shown, not by enough that made them worth the complexity in my opinion.

For the mathematically inclined, I show at the top of the graph the equation for the best linear fit to the data, and it is repeated in words below.

Sales Price (US$) = $430 + 0.34 x GoValue Estimate

For example, if GoValue tells you that a certain domain name is worth $900 the best fit regression suggests that your sales value would be about $740.  Note that at best results are only meaningful to two significant digits, so I have rounded both the parameters in the equation and the final result to this precision.  Let me stress that this is only moderately more significant than simply using the average value.  Also, this equation is derived only for domain names and is not directly applicable for any other types of domain names


Christopher Beam has written a really informative article at The New Republic on why, particularly in China, numbered sequences are so popular and effective as domain names. I strongly encourage you, particularly if you invest in numeric domain names, to read the entire article which is full of  insights.

One thing  I noticed while using GoValue was that there were numerous comparator sales that came up that had not been in the NameBio database, and generally these were towards the lower end of sales prices.  This suggests that some of the venues not reporting to NameBio may well have lower prices, and that possibly the price range for the recent sales is indeed in keeping with the overall record. NameBio no longer  calculate the median, but almost certainly it is in the same range  as these sales. NamePros experienced  users have pointed out that sales to end users frequently have higher prices than  sales on venues such as GoDaddy Auctions and some of the registrar marketplaces that are dominated by sales from one domain investor to another.

Overall GoValue does a good job of estimating the value of these assets, with almost no difference between the average value predicted by GoValue for the dataset and the average sales price.  It seems only moderately effective in recognizing higher quality names however, so you should never depend exclusively on any automated estimate of domain worth.

As with conventional  real estate, prices go up and down over time.  I sense a very slight downward trend, but NNNNN domain names remain good value assets. Relatively short numbered domain names are among the most liquid part of a domain name portfolio, and therefore deserve to be part of your diversified portfolio, provided that you can acquire them at reasonable cost.


Original post Aug 12, 2018.

Disclosure:  I do not currently hold any NNNNN domain names, so have no conflicts of interest with respect to them.  I do try to use a fair and balanced approach in all of my analyses however.  I am  not associated with NameBio, but would like to acknowledge their incredible database and their generosity to make it available to the domain community. 

Fine Print

This post is offered for informational and educational purposes only, and should not be considered domain name investment advice. While an attempt has been made to be accurate, there is no implied or explicit warranty, and you are responsible for verifying any information of importance to you. You also accept full responsibility for any domain investing decisions you may make that use data from this post.

I try to be fair, balanced and objective in my analysis.  If you feel this post does not meet that standard, please express your concerns to me.  

In a very few cases there may be affiliate links on this blog. This means I receive a small amount if users visit or make purchases via the link. You do not pay any additional charge due to using an affiliate link, and in some cases your cost will be below the normal price. I receive no identifying information about who clicks, or does not click, any link. I never accept compensation to provide favourable review of any particular service or product.

The text of this posting is ©R Hawkes, all rights reserved. However, you may, without permission, use reasonable length portions of the post as long as a link to this post is also provided. If you wish to use the complete contents of a post, please request permission. I am normally open to reprinting, but will consider each request individually. 

The images used are either those associated with a product or service, my own images, or Pixabay images believed to be available for use without attribution. If you see any image that you believe is problematic, please let us know and we will immediately correct the situation.

No comments:

Post a Comment

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...