Conversion Optimization Glossary

From a/b split testing to Z-score, learn the terminology that conversion rate optimization pros use everyday. 




A-B split testing

The simplest form of landing page testing. A new visitor to the page is randomly shown either the original version (“A”) or an alternative version (“B”).

Above the fold

The portion of a Web page that is seen without vertical or horizontal scrolling.


A company or individual that voluntarily chooses to promote the products or services of another company. Affiliates are paid based only on the measurable and trackable tangible actions that result from their promotion activities.

Affiliate Program

A performance-based marketing program set up by a company. Affiliates join the program and are compensated based solely on their performance. Typical payment methods include a percentage of sales revenue generated, or a fixed amount per specified action on the company’s website.

^ Back to top



Acronym for “business-to-business”. Refers to vertical industries or businesses whose clients are also businesses (rather than consumers).


Acronym for “business-to-consumer”. Refers to vertical industries or businesses whose clients are consumers or the retail buying public (rather than other businesses).

Back end

The portion of a website or Web application (including the underlying database) that does not interact with the end user.

Back links

A hypertext link from another website to a page on your site.  A sufficient number of properly constructed back links from reputable websites can increase your position on search engine results pages (see “SERPs”).

Banner Ads

Rectangular graphical ads of various dimensions that appear on a website. Banner ads may contain animation or other interactive features. Normally the website owner does not have control over the content or color scheme used in a particular banner ad.

Banner Blindness

The tendency of website visitors to ignore and tune out banner ads.


The original version of your landing page that is used as the benchmark against which other design variations are compared. Also called the “champion” (see “Control”).


A list of “junk” email addresses from which an email program will not accept messages. Blacklisting is one form of spam filtering.

Bounce rate

The percentage of visitors who land on a Web page and immediately exit without visiting any other pages linked from it.

Brochure ware

Websites that are static in nature and provide high-level descriptive information only - lacking in any interactive features.


(see “B2B”)


(see “B2C”)

^ Back to top


Central Limit Theorem

An axiom of probability theory that states that regardless of the original distribution of a random variable (such as conversion rate), its average will conform to a normal distribution.

Clickstream analysis

A capability of Web analytics software to display and represent popular sequences of pages that visitors navigate on a website.

Clickthrough rate

The percentage of Web page viewers who click on a particular link (also abbreviated “CTR”). CTR is often applied to the percentage of Internet users who click on a PPC advertisement and land on the advertiser’s landing page.


Programming functionality that takes place in the visitor’s Web browser software after the page has been loaded. Many landing page testing changes are implemented via client-side technologies.


The practice of showing different content to search engine spiders and human visitors to a Web page for the purposes of manipulating the ranking of the page in search engine results.

Continuous variables

Testing variables that can take on a range of continuous values such as temperature or pressure (see also “Discrete variables”).


(see “Baseline”)

Conversion Action

A desirable measurable action on a website performed by a visitor. Examples include clickthroughs to another page, form-fills, downloads, or product purchases.

Conversion Rate

The percentage of landing page visitors who take the desired conversion action.


A small informational file stored within the Web browser software that records information about current or past visits to a particular website.

Cost per thousand impressions

The dollar post to be paid by an advertiser for each thousand appearances of an advertisement on a particular Web page or set of websites (also abbreviated as “CPM”).


(see “Cost per thousand impressions”)


An abbreviation for “customer relationship management”. This type of software is used to track the whole history of a company’s interactions with a particular person across multiple channels including the Internet, telephone, mail, and in-store visits.


(see “Clickthrough rate”)

Customer relationship management

(see “CRM”)

^ Back to top


Deep link

A hyperlink to a page with very specific information within a website.

Deep linking

The practice of linking to the most relevant landing page possible within a website from a PPC keyword.

Descriptive statistics

A branch of applied statistics that is used to describe and summarize attributes of the data collected in an experiment. Typical descriptions include the mean, variance, and standard deviation of the data.

Design of experiments

The methods used to collect and analyze data from multivariate tests, sometimes applied incorrectly to refer only to fractional factorial testing approaches such as the Taguchi method. Abbreviated as “DOE”.

Discrete variables

Variables which can take on a set of distinct enumerated values.


(see “Design of experiments”)

^ Back to top


Error bars

A measure of uncertainly regarding the true value of a sampled quantity. Error bars can be represented graphically, or expressed as a numeric “plus-or-minus” range around the observed value.


In probability theory, an event is the set of all possible outcomes to which a probability is assigned. In landing page testing the event is commonly defined by the probability of converting, and the corresponding probability of not converting.

Experimental studies

A statistical method in which you observe a control condition and then modify the environment in a pre-planned way to see if the modification resulted in an observable change in the desired outcome.

^ Back to top


First-party cookies

Cookies set by the particular Internet domain or website that an Internet user is visiting (see also “Cookies”).

Fractional factorial

A subset of DOE and multivariate testing that seeks to cut down on the proportion of recipes sampled from the total search space in order to extract the most useful information from the smallest number of recipes. Fractional factorial methods (such as the Taguchi method) make simplifying assumptions about the underlying model (see also “Full factorial”, “Main effects”, and “Variable interactions”).


A method of designing Web pages in HTML by which information is pulled from a number of distinct sources and constructed in a collage-like fashion into the final page.

Front end

The portion of a website of Web application that interacts with and is seen by the visitor. Often used interchangeably with the term user interface.

Full factorial

A subset of multivariate testing and DOE that collects data evenly across all recipes in the search space. Allows for the most complicated and accurate models of the underlying data by taking variable interactions into account.

^ Back to top


Gaussian distribution

A distribution of sampled data that can be approximated by a “bell curve”.

Gross margin contribution

In accounting, the dollar amount that a purchase adds to the profit after subtracting all variable costs.

^ Back to top


Inferential statistics

The branch of applied statistics that is used to predict or model the behavior of an underlying system based on an observed test sample.

Information foraging theory

A branch of applied computer science and that describes the behavior of people when faced with a lot of available information in their search for a specific solution to a current need.

Information scent

The extent to which a person’s attention can be kept on a particular task or desired outcome based on the visual clues such as text or links placed on Web pages.

Input variables

The variables in a landing page test that are assumed to have an impact on the conversion rate or other optimization criterion. Also called “independent variables”.

Interruption marketing

Marketing in which the visitor must be interrupted in the course of their normal activities. Examples include billboards, television commercials, and Web banner ads (see also “Permission marketing”).

Inverted pyramid

A website copywriting style in which important information is put at the beginning of the page and summarized, and detailed information is provided lower on the page or accessed via related links.

^ Back to top



A word or phrase typed in by Internet searchers. A portfolio of keywords related to a particular topic or industry is commonly used as the basis for constructing PPC campaigns.

^ Back to top


Landing page

A landing page is the first page that a visitor lands on as a result of a traffic acquisition activity. The landing page can be a stand-alone page, a part of a special purpose microsite, or a page on the company’s main website.

Latin squares

A fractional factorial multivariate testing method.

Lifetime value

The full economic value resulting from a particular conversion action as measured over the whole lifetime of that visitor’s relationship with a company. Abbreviated as “LTV”.

Likert Scale

A surveying response scale that measures affinity or agreement. Most commonly used with five response levels (strongly agree, agree, neither agree nor disagree, disagree, strongly disagree).

Linear models

A class of mathematical models that adds and subtracts the effects of all input variables and their combinations.


(see “Lifetime value”)

^ Back to top


Main effects

The effects of individually changing the values of a single variable in a multivariate data sample. If a model only measures main effects, it assumes that there are no variable interactions (see also “Variable interactions).

Managing by exception

A management principle that focuses on problems or deviations from normal behavior as measured by a set of performance indicators.


A copywriting style that embellishes the effects or benefits of a product or service in an attempt to make it more attractive to the target audience. Frequently uses superlatives and adjectives.


An abbreviation for Myers-Briggs Type Indicator. A framework for behavioral styles base on someone’s innate brain type and cognitive predispositions.


The sum of all measured variable outcomes divided by the number of outcomes. Also commonly called the “average” value.


A special-purpose small website that is designed to maximize conversion rates for an online marketing campaign or traffic source.

Multivariate testing

A type of landing page testing methodology where data is collected simultaneously across a number of different variables (see also “A-B split testing”, and “Design of experiments”)

Myers-Briggs Type Indicator

(see “MBTI”)

^ Back to top


Negative interaction

A combination of two or more variable values in a multivariate test that results in a worsening of performance (see also “Positive interaction”).

Normal distribution

(see “Gaussian distribution”)

Null hypothesis

In statistical testing, the assumption that there is no difference in outcomes based on changes to the test variables. If there is a significant observed effect, the null hypothesis is said to be “rejected”.

^ Back to top



A type of search engine traffic that originates from non-paid search results often as a result of search engine optimization (see also “Search Engine Optimization”).

Output variables

The measured quantities that are studied in a statistical experiment. The goal is to prove a relationship between modifying the input variables and any resulting changes in the output variables (see also “Output variables). Also called “dependent variables”.

^ Back to top


Permission marketing

Any marketing activity that is voluntarily accepted by a member of your target audience. Permission marketing must be anticipated, personal, and relevant (see also “Interruption marketing”).


A detailed profile of a hypothetical person representing an important class of visitors to your site. The persona allows you to empathize with them and understand their needs. Used as the basis for constructing a relevant and effective conversion experience.

Positive interaction

A combination of two or more variable values in a multivariate test that results in improved synergistic performance (see also “Negative interaction”).

Probability distribution function

In probability theory, the probability distribution function describes the set of possible outcomes for an event along with their likelihood (see also “Events”). Also called a “probability density function” or “PDF”.

Probability theory

A branch of mathematics that deals with the description and analysis of random events.

Promo code

A special alphanumeric promotional code that allows the holder to get discounts or special deals not available to the general public.

^ Back to top


Random variable

In probability theory, an event drawn from a larger population of events that is independent of all other events (see also “Event”).

Rate card

The rate sheet indicates current pricing for promotional advertising. Rate cards are commonly expressed in cost-per-impression, or cost-per-click.


A unique combination of variable values in a multivariate test. Defines a unique version of the landing page being tested.


The size of the improvements that can be reliably found in a landing page test with a certain number of total conversions sampled. The larger the data sample the smaller the effects that can be “resolved”.

Run of network

Exposure of an online advertisement across the whole network of available websites for a particular distribution advertising network.

^ Back to top


Sales force automation

A type of software that allows you to track all important interactions with potential sales prospects and clients. Abbreviated as “SFA”.

Saturated main effect

A type of fractional factorial model that only considers main effects and assumes that variable interactions are not strong (see also “Main effects”, “Variable interactions”).


Program code and scripts for constructing a Web page that execute on the Web server before the file is transferred to the Web browser for display (see also “Client-side)”.


(see “Sales force automation”)

Search Engine Optimization

The process for having your website appear as high as possible in the search results for relevant keywords. Also abbreviated as “SEO”.

Search engine results

The pages that are displayed by a search engine in response to a keyword typed in by the searcher. Also abbreviated as SERP.

Search engine spiders

Automated robotic scripts that visit and index websites for the purpose of classifying them in search engine results.

Search space size

The total number of unique recipes possible in a landing page test (see also “Recipe”).


(see “Search Engine Optimization”)


Acronym for “search engine results pages” (see “Search engine results”).

Service level agreement

Guarantees of a certain level of responsiveness and availability for a certain Web application or service. Also abbreviated “SLA”. SLAs are typically expressed in a latency delay to access the data, and percentage of uptime that the website should be operational.

Signal-to-noise ratio

The strength of a particular observed effect as a ratio of the background variance associated with it. Also abbreviated “SNR”.

Staging environment

A parallel implementation for a Website or landing page that is used for testing new features or quality assurance before the content is moved to the “live” or operational environment.

Stochastic process

In probability theory, a time series of independent random events drawn from the same underlying populations (see also “Time series”, and “Event”).

^ Back to top


Taguchi method

A specific type of fractional factorial design of experiments approach commonly found in manufacturing, and sometimes applied to landing page optimization (see also “Fractional factorial”, and “Design of experiments”).

Third-party cookies

Cookies left by websites other than the one that the Internet browser is visiting. These kinds of cookies are commonly used for tracking advertising campaigns and are often turned off by Web surfers (see also “Cookies”, and “First-party cookies”)


The practice of restricting the percentage of available traffic that is allocated to certain recipes in a landing page optimization test.

Thumbnail image

A small product image that is used on ecommerce website pages to show a large number of products. Larger images are subsequently displayed on the product detail pages.

Time series

A series of independent and random experimental observations drawn repeatedly from the same underlying population (see also “Stochastic process”).

^ Back to top


Universal Resource Locator

A method for describing specific content (such as a Web page) that is available on the Internet. Also abbreviated “URL”.


(see “Universal Resource Locator”)

User-centered design

A philosophy and practice that considers the needs and background of the intended user of an object or interface as central to the design process.

User experience

A series of specific user interactions with a website form a larger user experience. Also abbreviated “UX”.

^ Back to top



The specific setting for a variable in a landing page test (see also “Variable”). Also called a “factor” or “variation”. The number of distinct vales for a particular variable is called the “branching factor”.


A specific landing page tuning element or page section that is part of a landing page test (see also “Value”). Also sometimes called a “level” in statistical testing.

Variable interaction

The effect of a variable value that depends on the settings of one or more additional variables. In other words, the context in which something is seen will have an effect on its impact. Most fractional factorial approaches assume that there are no variable interactions (see also “Positive interaction”, “Negative interaction”).


The sampling noise during a statistical experiment. As the sample size increases, the variance decreases.

^ Back to top


Web analytics

Software for analyzing and tracking the behavior of visitors to a website.


A set of email addresses from which an email program will always accept messages, regardless of whether the messages fail other spam filtering criteria.

^ Back to top



A statistical measure of the difference between two quantities. The larger the Z-score the less likely the two quantities are to have been drawn from the same population.

^ Back to top