(A) INTRODUCTION
A domain name can create value to the owner
by generating earnings from an active Web site. In fact, even
if a domain name does not have an active Web site, it can still
be generating income as measured by cost savings associated
with preventing a cybersquatter or competitor from siphoning
away sales.
Domain-name ownership confers the owner with
two types of rights. The first is related to managerial flexibility,
such as when to develop an associated Web site, when to abandon
the operation of the Web site while retaining ownership of the
domain name itself, when/whether to renew registration and for
how long. The second right is associated with trademarks. If
a domain name is trademarkable, the owner has the right to legal
protection; that is, the owner can prevent anyone else from
claiming the domain name, if he/she wishes to exercise that
right. The options available to the trademark owner are to take
a trademark violator to court, to buy the domain name, or to
take no action. This flexibility has value, which is embedded
in an option. Thus, these options are valuable managerial tools
that cannot be ignored.
Thus, the value of a domain name comes in
two forms: expected earnings and options to take actions. The
latter component can be broken down into managerial flexibility
(referred to in finance literature as real options) and a trademark-related
option.
Hence,
the value of a domain name is the sum of the value of
its expected earnings, the value of its managerial flexibility,
and the value of its trademark option. |
(B) VALUATION APPROACHES
As recognized by the Uniform Standards of
Professional Appraisal Practice (USPAP), there are three
generally accepted approaches to estimating the value of all
assets: (1) the market approach, intended to reflect
comparative market prices; (2) the income approach, intended
to reflect economic value; and (3) the cost approach, intended
to reflect the utility characteristics of the asset. These approaches
apply to intangible assets and intellectual properties, as well
as to tangible property.
Each valuation approach emphasizes a different
attribute of a domain name. Using all applicable approaches
may increase the confidence level of value conclusions. Nevertheless,
poorly supported valuations result from the naïve use of all
three approaches. The information that is available for valuation
should determine the approach used.
The market approach examines the comparative
characteristics of reasonably competitive properties. When there
are sufficient market-driven transactional data from which to
estimate comparable domain names, this approach is appropriate.
If the selected comparable domain names are not, indeed,
comparable to the subject domain name, the market approach
is weakened.
The income approach relies on the cash flow
that the domain name is expected to generate over its life.
As such, this approach requires a reasonable estimate of future
cash flows and their risk. Thus, quality of valuation depends
on the accuracy of the estimates used in the valuation model.
The cost approach looks at the cost to reproduce
or replace an asset. This approach is not appropriate for domain
names and intangible assets, since the cost to replace
such an asset is seldom reflective of its value, except at the
inception of its life.
(1) Market Approach
A market approach to domain-name valuation
is both an art and a science. The art comes from knowledge and
experience in understanding the factors that influence the value
of a domain name, while the science involves statistical techniques
to quantify the importance of these factors.
(a) The Issues
What is the price of
xyz.net or xyz.biz? How different from xyz.com are they?
To answer
such questions, let’s consider the following examples:
Example 1.
Domain Name xyz
Extension |
Sale Price ($) |
com |
600 |
net |
300 |
biz |
200 |
Example 2.
Domain Name xyzz
Extension |
Sale Price ($) |
com |
300 |
net |
150 |
biz |
100 |
The
above examples suggest that .net is half the price of .com,
while .biz is one-third the price of .com. Thus, it seems that
if you know the price of a name with any of the above extensions,
you will easily be able to value the other extensions.
Unfortunately,
however, in reality, pricing different domain-name extensions
is not such a simple task. There are two major obstacles: there are not enough domain
names with different extensions sold, and the pricing relationship
between different extensions may not be a constant multiple
of each other.
To overcome
the above shortcomings, one can use statistical techniques to
predict the price of comparable domain names based on a reliable
database of domain prices.
We have developed
a statistical model to predict the price of a domain name. Statistical
models are a prerequisite to performing any meaningful appraisal.
With such a model in hand, one can measure how good the prediction
is and can strive to improve the model’s prediction accuracy.
(b) Statistical Modeling
Statistical
models use the relationship between these factors and prices
of domain names that have been sold to determine the price of
a given domain name at a specific point in time. These models
provide a scientific method of estimating value based on comparables.
Traditional
linear-regression techniques don’t yield satisfactory results,
as the relationship between market price and the extensions
is most likely to be nonlinear. Thus, in technical terms, “dummy
variables” for each extension would not yield meaningful results.
Moreover, because the domain-name market is relatively new and
not very active, more robust statistical techniques are needed.
At the
heart of our regression-tree statistical model is a set of predictors
of demand for domain names and historical prices of sold names.
The estimated model yields the best relationship between the
specified predictors and value. This relationship is used to
predict the price of any domain name. One of the obvious predictors
is the domain extension. Another of the 12 quantifiable predictors
we use is the number of searches in Google for the keyword embedded
in a domain name. The predictors are selected based on statistical
techniques as well as on our extensive knowledge and expertise
in the domain-name marketplace.
Typically, the basic
appraisal services do not take directly into account the contribution
of a trademark to the appraisal value, as trademark valuation
requires extensive domain-specific calculations and data collection.
An income approach is more appropriate for this class of domains.
We forecast the price
based on a statistical model for the form:
Price
= f (X_{1}, X_{2}, ..., X_{N}),
where
Value is the estimated market value of a domain name, f ( )
is a nonlinear function that also allows interaction between
the predictors, and X_{i }is the i^{th} predictor
of Value.
Before selecting
a predictor, we require that it make economic sense — i.e.,
it must be a meaningful predictor of profit (for example, even
if “the number of sun spots” were highly correlated with Value,
it would not qualify as a predictor). Moreover, the data available
should reflect the true relationship between the predictor and
Value (for example, if one expects a positive relationship between
them, the data should support such an assertion; otherwise, the predictor would not be used).
A regression-tree
model can be illustrated using Figure 1 below:
Figure 1.
A Two-Predictor Tree Example
In the above stylized
regression-tree model, factor 1 can represent, say, the number
of advertisers on Google for the keywords implicit in the domain
name, while predictor 2, say, if the .biz is registered (predictor-2
= 1) or not (predictor-2 = 0). Thus, for any domain name to
be appraised, if at time of appraisal it has more than 22 advertisers
on Google, its appraised value would be $11,400. If it has less
than 22 Google advertisers, the equivalent .biz domain name
is registered, and has Google advertisers greater than 10, its
appraised value is $8,220.
Thus, the model is
powerful to handle nonlinear relationships between the predictors
and Value. Also the predictors can be discrete variables (0
or 1).
Given the above estimated
regression-tree, a domain name that has Predictor-1 greater
than 22 units, would have an estimated market value of $11,400.
Tree-structure estimation
techniques are used to estimate the model, yielding a model
superior to the standard least-squares regression approach.
Given the set of qualified predictors, the final predictors
used are the ones that minimize the fitted deviance (the difference
between the actual sale price and the predicted Value). Because
such models don’t use the standard measures of goodness-of-fit
to provide you an idea of the quality of our model, we use the
R-squared from the linear model:
Value
= b_{0} + b_{1}X_{1} + b_{2}X_{2} + …+ b_{N}X_{N},
where
b_{i }is the estimated regression coefficient for X_{i}.
Using our predictors
in a linear model yields a multiple R-squared of .78, i.e.,
the model explains 78% of the variations in the prices of sold
domain names. Thus, the nonlinear model used in our Appraisal
Report yields a more reliable predicted value than the linear
model.
(c) Valuation Database
In estimating the
predictive model for Value, we use transaction prices collected
from publicly available auctions, as well as prices from proprietary-domain
escrow and sealed bid auction data. For each of the prediction
variables, data are collected from publicly available sources
at time of sale. Thus, the only proprietary data used are transaction
prices from domain-name escrow and customized auctions
through DomainMart.<<
The starting period
for which complete data on predictors is available is November
2002. As of April 2005, the database has over 2,000 observations.
(2) Income
Approach
(a) Methodology
This option is typically
used for predicting benchmark profit-potential of a website
and appraising active websites and domain names that involve
trademarks.
The income approach
to valuation, also called the Discounted Cash Flow (DCF) method,
is one of the tools used to value any asset, including domain
names; the earnings represent the additional or incremental
cash flows the domain name is expected to generate to the owner
over the life of the domain name.
(b) Estimation
DCF analyses require
access to information on traffic, revenue, and costs associated
with the business. To use this powerful valuation tool to estimate
a benchmark value for a website, we assume that the domain name
is parked to generate traffic revenue. The strength of the approach
is that it can be applied to any domain name irrespective of
whether it is parked or not.
The traffic income
business model focuses on domain names that generate clicks.
This is achieved by placing advertiser links on a webpage. Every
time a visitor clicks on any of the links the advertiser pays
the link manager a fee, i.e., pay-per-click (PPC).
The availability of
reliable public information on keyword searches and revenue
from PPC advertising has made valuation based on traffic income
a compelling domain-name valuation methodology.
The advantages of
the income-based over the market-based methodology are founded
on the following facts:
1. The median sales price of catalog listings is about $500. Moreover,
only a small number of sales are in the tens of thousands. Thus,
applying statistical models to value premium domain names will
not yield precise estimates due to the paucity of data.
2. For a specific level of website traffic, the extension of a
traffic-revenue domain name should be irrelevant, holding other
factors constant. However, sales data suggests that .com names
command a considerable premium, even after controlling for keyword
composition. Thus, using an income approach for such domain
names yields a more accurate appraisal.
3. Only a small fraction of domain names sold are hyphenated. Thus,
as in (2) above, they are undervalued by a statistical model.
4. Historical market prices, especially those for domain names
sold on auctions, suffer from asynchronous demand and supply,
whereby not all parties interested in the domain name would
be aware of its sale or willing to commit by the end of the
auction. Thus, the sale price might not reflect the market’s
true willingness to pay for the domain name.
5. The income approach allows various CF scenarios to be considered,
typically a “best case,” a “worst case,” and a “middle of the
road.” Such an analysis provides a more intuitive picture of
the range of possible market values.
The advantages of
DomainMart’s hypothetical parked domain name methodology over
using historical data from parked domain names are:
1. Grouping comparable parked domain names based on historical
data involves considerable classification error (incorrectly
including sales within a group of “similar” names or excluding
sales belonging to a group), especially in clusters with few
data points on sales. Thus, the results would be less reliable.
This error is magnified when classification is based on arbitrary
techniques.
2. Brokers tend to keep historical parking revenue information
private. Thus, diminishing the transparency and verifiability
of appraisals based on historical parking data.
The advantages of
DomainMart’s parked-domain methodology over a historical revenue
approach are:
1. It does not require access to private income data, which a domain
name owner might be reluctant to provide.
2. It does not require an active domain name. On the other hand,
without historical income data, a revenue model cannot be estimated.
3. It is considerably cheaper.
(C) LIMIATIONS
Neither the market approach nor the DCF technique
captures the value of flexibility options. Thus, an appraiser
can use the DCF method to estimate the earnings component and
use option-pricing-theory models to estimate the two option
components separately. Although in principle, an appraiser can
use decision-tree analysis to estimate flexibility options,
an option-pricing methodology can be much simpler to formulate.
Moreover, DCF techniques require estimating the risk of cash
flows, whereas the option-pricing methodology overcomes this
difficulty, especially when this risk is not constant, as assumed
by the DCF method.
Let's look at the trademark-option component
by considering the action of a cybersquatter (someone who registers
a domain name that constitutes a trademark infringement). Such
an action is equivalent to writing a put option on the domain
name, in that the cybersquatter is legally obligated to surrender
the domain name. However, the owner of the domain name has the
option to surrender the domain name, litigate, or take no action.
One could use discounted decision-tree methods to value such
a domain name by considering the different actions and counter-actions
that an owner of the trademark and a cybersquatter can take
and the consequences of each action. However, this process would,
at best, be cumbersome, compared with an option-pricing model.
In fact, even if a domain name has no associated trademark,
as long as that domain name is trademarkable, it has a higher
value (other things being equal) than a non-trademarkable domain
name of a generic word. Obviously, the higher the value of a
trademark, the higher the value of the associated domain name.
However, an analyst has to be careful in distinguishing between
the contribution to value from the trademark and that from the
earnings component.
In sum, taking account
of flexibility options and trademark options embedded in domain
names provides a more accurate domain-name appraisal than using
the DCF method alone.
Topic tags: appraisal/valuation
Connect & Share