So far we have discussed the basics of adverting and players involved, this particular post will dig right into the detail and review how we measure programmatic advertising. It may be a little more mundane than other posts in the series, but the information in this post should be understood by everyone before a real understanding of Ad Tech can be gained.
Note: This is a selection of the most common metrics, it is not complete and so there are a variety of other metrics you may come across. However, understanding the below set fully will provide a good overview of advertising measurement.
Impressions, Clicks & Actions
At its core online measurement breaks down into three metrics these map to impressions — the delivery of an Ad creative to a publisher placement, clicks — The interaction of a user on the Ad by clicking through or actions — The user achieving a goal of the campaign other than click such as landing on a checkout page, note Actions are often also called Events and the most common type of action is a conversion.
Of course, these are just raw metrics, in reality the advertiser (and campaign optimiser) require much more refined views of their campaign these metrics are aligned with one of the key campaign types: Performance or Brand. In order to derive many of these metrics, its important to first review financial metrics.
Financial metrics
Financial metrics provide insight into the money side of a campaign, there are revenue metrics (Money received from the advertiser) and cost metrics (Money spend on the campaign). It is important we track both sets of metrics — to correctly charge the advertiser, and to ensure the campaign is profitable.
Gross Revenue: the top line monetary value agreed to be spent on the campaign. An advertiser will request a certain amount to be spent on advertising over a given period of time, with set goals. Note, if a campaign is proceeding well — the customer may apply an upweight which simply adds more committed budget to the campaign.
Discounts: There are a variety of discounts: Standard discount for a customer — a (typically yearly) fixed favourable rate to encourage them as a customer; Rebate discounts may be agreed upon certain committed spend levels to encourage large spends — for example- spend £1M with us this year and we will offer an additional 10% discount; and added value — a discretionary discount the sales team can use to sign a specific campaign.
Net Revenue: The Gross budget minus any discounts provides the net revenue or the figure that will be billed to the customer should campaign goals be achieved.
NOTE: Net revenue (given it is the billed amount) is the value used to derive all Cost associated campaign metrics.
Media Cost: Simply the cost to acquire impressions, what is the cost of the media purchased.
Data Costs: The amount of money paid to partners for targeting data used in the campaign. This can include our own data partnerships — but may also include data such as brand safety or contextual targeting on a campaign (reviewed later).
Tech Fee’s: We don’t get to use the DSP’s for free — each DSP charges a rate as a % of the bid price of media bought.
Margin: Margin is a standard margin model: What % of our revenue is profit. Easily calculated as Profit / Net Revenue (Or Net Revenue-Costs/Net Revenue). Margin is a very important metric as it shows how successful the company is at making profit independently from how well we have sold business in a given month. We sometimes discuss a short cut margin called media margin, this is simply the margin based only on media costs and gives an easy to calculate measure of campaign financial performance without having to worry about more complex costs such as tech fee’s. It is however, an incomplete metric so is used for campaign performance only.
Performance metrics:
Performance metrics focus on DR campaigns, given a majority of our campaigns are DR, those metrics which are used commonly for both Performance and Brand will be included here.
Impressions: The base unit of campaign delivery, how many times has a creative been displayed.
CPM: Cost per mille (Cost per thousand) is the standard way to discuss the cost of delivery. Given impressions are typically very cheap on an individual basis, costs are per 1000 impressions. (Net revenue/impressions)*1000. Incidentally, campaigns are often priced by CPM a fixed cost per impression, though this doesn’t allow for much flexibility in the number and type of impressions delivered to reach a target, so a more efficient mechanism is to bill campaigns as dynamic cost per thousand (dCPM) — in this case, revenue is set, the trader defines the bid range and generally the target performance is the important factor rather than a fixed number of impressions served.
Clicks: A click occurs when a user clicks (Or taps on mobile) the creative. For DR campaigns, this is a common campaign goal as it leads to the user being redirected to the content the advertiser is trying to sell.
CPC: Cost per click, like CPM is a derived metric of cost per click. Given Clicks are far more rare than impressions, there is no need to use 1/1000th’s instead the result is simply: Net revenue/Clicks
Conversions: The ultimate goal of many DR campaigns is to have the user take a specific action. This could be reach a landing page, a checkout page, fill in a form or some other event. Essentially, a conversion arrises due to a pixel fire on a specific page, or an SDK event on mobile such as downloading an app. There are two standard types of action — Post click and Post view:
- Post View Conversion (PVC): The conversion is proceeded by a view of an impression, but the user did not necessarily click on that ad. This may seem unusual at first read, why would an advertiser measure conversions that were not the direct result of an impression? Quite often users will gain familiarity over time from an ad, and not immediately click — but instead find the product through other means later on.
- Post Click Conversion (PCC): The conversion is proceeded by a user clicking on the impression. This doesn’t mean the user immediately clicked on the Ad and reached the conversion — the user may return to the form or checkout page at a later date and still be measured as a Post Click convertor. This all depends on the attribution mechanism in place.
CR (Conversion Rate): Is simply the rate at which users convert, much like CTR — it is calculated as Conversions/Impressions
Attribution: There are many attribution models, the objective of an attribution model is to specific how events are correlated back to impressions. The most common attribution model (Due to its simplicity more than anything else) is last touch. In this model, the last impression to be shown before the event gets full credit for the event. Other attribution models focus on multi-touch (MTA) where any series of impressions take a share of the event and each is given a fraction of the event credit. You see various MTA models: equal weight, time decay, or even first touch (which actually isn’t multi touch). For all attribution models — an attribution window is needed, this defines how long is allowed between the impression and the action — the most common timeframe being 30 days.
CPA: When measuring conversions the typical performance measure is Cost per Action, the total cost of each action. Some people prefer the term Cost per Acquisition, since you are acquiring a customer. The equation is simple: Net Revenue/Total actions. A lower CPA means better performance as the advertiser is getting more conversions for lower cost.
ROI: CPA alone will indicate the cost for an action, however — it does not incorporate the value of that action. If your action is of very high value — a high CPA is acceptable. In order to provide a balance between cost and value — Return on Investment is used. ROI is calculated as the advertisers revenue / advertisers cost. Revenue can be a fixed amount per action, but is sometimes a value passed in on the conversion pixel (such as shopping cart value) in which case a sum of conversion values is required.
CPL: A less common metric for performance campaigns is Cost per lead. This is essentially the same as CPA, but specifically the action is a lead generation — a customer signing up for a test drive or requesting a brochure.
CPI: On mobile devices it is common to measure the cost to the advertiser for each application installation. If the advertiser is promoting their mobile App, this is an easy way to measure the return on their investment — it requires the app to trigger a pixel fire on first load which is then attributed back to the impression shown within another app (Or in a cross device environment via other display advertising).
Brand metrics:
Brand metrics tend to focus less on the actions of a campaign and more on the number of people who saw the marketing message and for how long. Some of the above metrics are used, CPM for example is an easy pricing model for brand campaigns so is often used.
Reach: Given brand objective is to provide a message to a large audience, there is often no direct performance target. Instead Brand advertisers are interested in ensuring they reach as many unique users as possible. Reach is typically measured as a count of unique (or distinct) users.
Viewability: Clearly without a direct effect measure, it is very important to advertisers that users actually saw the ad. The objective of viewability is to ensure ads are in view — that is-not below the page fold or in some other way obscured from view. There are many companies that provide technology to enable viewability measurement which is a % of Ads shown that are viewable by the user on page. More and more brand advertisers are requiring a minimum viewability rating for their campaigns.
Survey Measures: With a lack of impact metrics, often brand advertisers use a panel survey. The survey will be shown to a small portion of the audience and ask questions around the advertising effectiveness. Clearly the measurement is limited to the survey size — even a fairly large survey audience requires users to respond to the survey truthfully so results are often questionable.
Video Specific Metrics
Video is most frequently used in brand advertising, it is less common to expect a user to take an action resulting from a video such as clicking through. In addition to the above brand metrics, the following are often used in video advertising.
GRP: Gross Rating Points has its origins in the Television industry. It is somewhat related to Reach in that it evaluates the audience covered, but is reported as a % of the total possible audience within the target set multiplied by the frequency users are exposed. For example, if a campaign targeted male adults in the London area, the target group may be 3 million users. If the campaign reached 1 million of these, but showed them the ad 4 times, the GRP would be 33% x 4 = 133%.
CPCV: Cost per completed view is the cost to the advertiser of each user who watches the video ad all the way through.
CPV: With Cost per View, the advertiser may be happy to measure based on a partial view, typically with a limited exposure (Such as the user having to watch at least 5 seconds of the video).
Audience Dimensions
Finally, the advertiser will be concerned not only in the performance or exposure of the brand, but who was shown the ad. Specifically they may target and measure a variety of audience dimensions:
Demographics: Basic demographic information about the user is one of the most common profile types. Typically demographic data focuses on Age & Gender. It may also include other less frequently used dimensions such as income or ethnicity. Demographics data usually comes from 3rd party technology partners who have publisher relationships to track users web browsing patterns and make assumptions based on their usage. Using known sample sets of users (based on surveys) they will create lookalike models to determine that certain web addresses skew towards certain demographics and assign demographic profiles based on these.
Geography: Probably the most frequently used dimension for targeting and measurement is the location of the user. Nearly all campaigns will ensure the user is within a certain set of countries, but often more granular geographic targeting will be used: Country, DMA/Region, State, City. For display adverting this typically relies on IP address lookup to fixed ranges provided by 3rd party vendors such as Digital Envoy, Quova or Maxmind. They use complex route tracing techniques and Wifi position databases to derive the approximate location. Accuracy to Country and State is generally very good, with it dropping off as we get more granular. For mobile — positional location is possible at the Lat/Lon level due to those devices positional hardware.
Frequency: An advertiser will want to show their ad to a given user more than once, but typically not more than 5 times. Performance and impact reduces as a user is exposed multiple times to an ad, though there is an initial performance increase for more than one exposure. Frequency will define how often a unique user see’s an ad within a specific timeframe such as a day or a month.
Time: Users perform differently throughout the day or week. For example, a user at work may be interested in different products from a user at home. As such — time is a common way to target and measure campaign performance. Day part (Hour or group of hours) & Day of week are the two most common ways to measure time.
Next…Discussing the DSP
