Common terminology in online advertising

There is some stable terminology within the online advertising industry that we will cover in this article. This post is part of a larger series of posts that I am writing on online advertising on HackerNoon:
https://hackernoon.com/u/rahoul

I am assuming that the reader is already familiar with general landscape of online programmatic advertising ecosystem, but if not - you might find this article useful:
https://hackernoon.com/programmatic-advertising-ecosystem-the-main-players

Events

Bid request - request saying “this user is going to view ad space of this size on this page, please suggest which ads you would like to show and how much you are willing to pay” that publisher sends to SSP and SSP sends to DSPs. Also might be called “bid opportunity”.

Bid response - reply to bid request that says “I will pay you $X if you show this ad to this user on this webpage right now” that DSPs send to SSP.

Ad auction - an auction that SSP runs to decide which ad to show. Usually it is either a first price auction or a second price auction. In the first price auction the winner is the highest bidder and the price is the highest bid. In the second price auction the winner is also the highest bidder, but the price is the second highest bid.

Ad impression - event of an ad being shown to a user. Publishers usually charge DSPs/advertisers on impression. There is some variety of interpretations of what “being shown to the user” actually means. For some publishers, even if the ad was in the bottom of the page and the user did not scroll enough to see it - it will still be considered an impression. To resolve this, some publishers introduce a “viewable impression” - ad impression when the user had X% of the ad visible for at least Y seconds. Modern Javascript and HTML allow to run special code in the page that measures and logs which parts of the page are displayed on the user's screen at each moment and for how long.

Click - user clicks or taps on the ad.

Conversion - target action that the advertiser wants the user to take after viewing an impression. Might be a user making a purchase on the advertiser’s website, or a user filling in some form and providing contact details, or just a user subscribing to an email newsletter. We will most frequently assume that the conversion is a purchase. There is also sometimes a funnel of conversion events considered:

• landing - user opened advertiser’s website page and have not closed it immediately
• page view - user looked at any page on advertiser’s website
• put to cart - user put some product to the shopping cart on the website
• initiate checkout - user started the purchase process on advertiser’s website
• purchase - user made a purchase

Post-view or post-click attributed conversion - also called view-through or click-through conversion correspondingly. Conversion that happened within X days after the user sees the ad or after the user clicks on the ad. Usually advertisers ask DSP to optimize for post-view and/or post-click conversions. For example, an advertiser might optimize for 1 day view-through plus 7 day click-through conversions - purchases that users make within 1 day after seeing the ad or within 7 days after clicking on the ad. If conversion falls under both definitions, it is usually reported as click-though. In naive view, we can consider that attributed conversions are the conversions that happen “because” of ads, not just “after” ads. We will discuss in the later articles the difference between conversions caused by ads and conversions just happening after ads and how advertisers might measure the former.

Metrics

CPM (cost per mille) - price per 1000 of impressions. Usually bid amounts and prices of impressions are computed in CPM for convenience. Typical price for impressions in programmatic advertising is $1 - $10 CPM.

Bid amount - how much DSP is ready to pay to show a particular ad. Basically bid reply consists of an ad to show and a bid amount. Also measured in CPM.

Cost, media cost, price, clearing price, spend - how much the advertiser was charged for impression. Supply side might also call this “revenue”. Also measured in CPM.

CPC (cost per click) - how much the advertiser paid for 1 click on average. Total media cost divided by total number of clicks. Typically within $0.01 - $1.

CPA (cost per action) - how much the advertiser paid for 1 conversion on average. Total media cost divided by total number of conversions. Usually within $1 - $100.

CTR (click-through rate) - rate of clicks per impression on average. Total number of clicks divided by total number of impressions. Usually measured in %. Frequently within 0.1% - 5%.

CVR (conversion rate) - rate of conversions per impression on average. Total number of conversions divided by total number of impressions. Might be within 50/1,000,000 - 1,000/1,000,000

Reach - also called unique impressions. Number of unique users who saw the ad. Potentially, the same user might see the same ad multiple times, so the number of impressions is always larger or equal to the number of unique impressions. For some advertisers reach is an important target metric if the campaign goal is brand awareness (users getting familiar with brand name), not direct response (users making particular action). Also comparing reach and impressions might reveal abnormalities in ad delivery.

Flavors of metrics

Effective - also called realized. What actually happened. For example effective CTR - total number of actually happened clicks divided by total number of actually happened impressions.

Target - also might be called "advertiser". What the advertiser asked DSP for when setting up the advertising campaign. Sometimes what the advertiser asked for is not what actually happens. For example tCPA - target CPA, how much advertiser would like to pay per realized conversion.

Expected - what DSP was expecting to happen. When replying with an ad, DSP was computing the probability of the user clicking on that ad and converting after that ad. All these probabilities can be logged and we can compute the total sum of expected clicks over all ad impressions, or average predicted probabilities per impression, and so on. Big difference between expected and realized metrics would usually indicate problems.

Let’s consider an example: an advertiser instructs DSP to spend $7000 in one week on ads and deliver at least 5% CTR. After 1 week, DSP spends $7000 and purchases 1,000,000 impressions that generate 40,000 clicks. In DSP logs we can see that DSP actually expected 60,000 clicks to happen from these impressions.

target CTR = 5%
effective CTR = 40,000 / 1,000,000 = 4%
expected CTR = 60,000 / 1,000,000 = 6%

In this particular example there was something wrong with DSP’s prediction engine so it was predicting too much and bidding too high. This is usually a sign of problems with Machine Learning models that DSP is using.

Sometimes people shorten metric names and say just tCPA or eCTR which might introduce ambiguity. For example eCTR may mean both “expected” and “effective”. We will try to avoid this, but the reader should be prepared for situations like these happening.

Advertising Campaigns

When an advertiser instructs DSP to achieve some results, for example spend some advertising budget and bring some conversions, the advertiser usually sets up a hierarchy of entities with DSP to specify what exactly is needed. The exact details and naming might differ from DSP to DSP, but the main idea will probably stay the same.

Ad campaign - an ad campaign is an advertiser's initiative to achieve particular results by spending a particular budget within a particular time. For the sake of our article, an advertiser sets up an ad campaign with DSP, instructing DSP to spend specified budget within specified time and achieve specific goals measured with metrics. Usually an ad campaign is a border umbrella entity that includes smaller entities called line items.

Line item - is a more granular component within an ad campaign that describes how a particular group of users should be approached. Each line item usually contains targeting information (which users to show ads to) and creatives (which ads to show). Line items may have their own budgets and maybe even their own optimization goals, similar or additional to campaign goals.

Creative - this is what an ordinary person would call an ad, the image or the video that is actually shown when the ad is shown.

Targeting - which users should see ads from particular campaign. When advertiser sets up an ad campaign, they frequently specify which users should see the ads. Targeting might be based on users location, gender, age, browsing history, interests, prospecting/retargeting and so on, similarity to some other group of users or might be even explicitly set to particular list of users.

Prospecting - targeting set up to show ads to only new users who never interacted with the advertiser before.

Retargeting - targeting set up to show ads to only users who have already interacted with the advertiser, purchased something from the advertiser or put something to the shopping cart on the advertiser’s website. When you see ads reminding you about products that you have seen on amazon last week - these are retargeting ads. Retargeting users are usually considered to have higher probability of conversion and therefore higher value compared to prospecting users - maybe because retargeting users actually are humans, have a bank account and can make online purchases, or maybe just because retargeting users have shown some interest in a brand.

Lookalike targeting - targeting set up to show ads to users who are similar to some explicit list of users uploaded by advertiser - for example similar to list of existing advertiser's customers. Lookalike targeting may be very powerful tool both to increase effectiveness of ads.

Let’s consider a particular example of a campaign setup for some local sportswear shop from Chicago, IL.

Ad campaign number 1000:
budget: $7,000
time: 2021-01-01 - 2021-01-07
goal 1: CTR above 5%
goal 2: maximum purchases

Line item number 1100:
targeting: prospecting, women, within 100km of Chicago, IL, US
creative: an image of a tennis skirt
creative: an image of female sport shoes
creative: an image of female swimsuit

Line item number 1200:
targeting: prospecting, men, within 100km of Chicago, IL, US
creative: an image of a pair of tennis shorts
creative: an image of male sport shoes
creative: an image of male swim trunks

Line item number 2100:
targeting: retargeting, within 100km of Chicago, IL, US
creative: big shop logo and text “5% discounts for returning customers”

Advertisers tend to set up multiple campaigns and line items, trying to segment the audience, assign separate budgets and show best creatives to each user group.

General terms

Ad inventory - for a publisher, all the spaces on publisher’s pages that can be filled with ads. Publisher’s goal is to monetize its ad inventory the best way - show ads in all the possible spaces for maximum possible price. Putting additional ad slot on the existing page increases ad inventory to some extent.

Marketing channel - might also be called “source of traffic”, large group of publishers, SSP or DSP from advertiser’s point of view. For example Google search, Facebook, news websites or even whole RTB might be considered separate marketing channels.

Audience - users who saw ads, or might see ads potentially. A publisher usually has fixed stable audience. An advertiser in general wants to reach as big or an audience as possible, as long as it is relevant. By using multiple marketing channels advertiser extends its potential audience. A publisher in general wants to monetize its audience as well as possible. If DSP connects to multiple SSPs it increases potential audience for its advertisers.

Advertising budget - amount of money an advertiser is willing to spend on ads in some time period. Advertising budgets are usually reviewed and distributed between marketing channels monthly or once in a number of months, depending on ads performance: if in this month ads on DSP Y worked well, advertiser might allocate a higher budget for DSP Y next month.

General goal of an advertiser is to buy as much relevant audience as possible for its advertising budget - and this way to drive maximum conversions. Goal of a publisher is to monetize its audience and ad inventory as well as possible by showing the most relevant ads to each user and therefore charging the most for each impression.

Discussion

All businesses are different and might put different meanings into similar words or use local lingo for general ideas. However, conceptually the ideas are usually close.