Background paper for the W3C Advertising Business Group. Draft, updated 2020-04-23
Contributors:
- Wil Schobeiri (wschobeiri@mediamath.com)
- Laura Carrier (lcarrier@mediamath.com)
- George London (george@survata.com)
This document describes the purpose and common uses cases for digital advertising, and attempts to connect them with technological solutions currently in use by the web ecosystem.
TODOS:
- Change links on introduction
- Clarify: How is this different than https://github.com/w3c/web-advertising/blob/master/support_for_advertising_use_cases.md?
- Commercial context / revenue overall (digital, programmatic), and per use case
- Approximate percentages of advertising spend supported per adverting use case, where possible
- Definitions are incomplete
- Complete constituent "we must" section
- Complete shortfalls section
"The aim of marketing is to know and understand the customer so well the product or service fits him and sells itself." (Peter Drucker, credited as founding modern management)
- What is the goal of marketing?
- What are their challenges?
- How does digital marketing today make those goals more tangible?
The goal of marketing is to change customer perception in order to influence behavior.
Practically, this typically includes 1) getting a customer to know that one's brand, product, or service exists, 2) consider it in the solution set for solving their problem or achieving their goal, and 3) selecting that product/service as their solution. This all should happen at the lowest possible cost to the company but also at the highest possible delight to the consumer such that long term loyalty is gained.
Along the path from awareness through purchase, there are many behaviors that a customer will take. Marketing activities attempt to guide a customer down that path to purchase, trying to incentivize the specific behaviors that have a higher propensity to lead to a sale while working to incentivize the actual sale. Once a sale occurs, this path (or subsets of this path) starts over, in an effort to continue the relationship between marketer and consumer.
Marketing’s goal can be boiled down to the "5 Rights":
- Right Message to the
- Right Person at the
- Right Time in the
- Right Channel and for the
- Right Reason.
Achieving this goal in the context of traditional marketing (print, live television, billboards, et al) is impossible. In digital marketing, not only can one achieve these goals, they can prove it happened. This proof is what enables marketing activities to continue, and is important for modern marketers to justify their advertising dollars, which ultimately finance the publishers sponsoring the underlying content being monetized. Without proof, finance teams are reluctant to release dollars, which leaves marketers, their partners and most importantly, their customers at the mercy of measurement – because the consumer is ultimately the biggest beneficiary when marketing is done right. They get to see the right message at a time when they are trying to solve their problem, in the right channel and based on solutions that are relevant to them (are they wanting to feel safe and secure in their selection of a power tool? and/or are they wanting to feel empowered and strong?).
The presence of certain data signals makes it possible for marketers to target more granularly and accurately, to tailor a relevant message to the customer, to engage with them at an appropriate or advantageous time, and across a medium that makes the most sense for the product or service.
Getting this wrong can be costly, damaging, or potentially even illegal. Advertising a product that a customer has already purchased, or airplane tickets alongside a plane crash article, or a new product to a customer who just called to complain, or even alcohol to a minor, are all examples of advertising gone wrong. The use of data helps modern marketers ensure these examples don't occur.
Once delivered, engagement and impact with ads in a digital context can be understood and measured in a way not possible in traditional advertising. Digital advertising enables marketers to demonstrate return on investment (ROI) and tangible business results that most modern businesses demand for any expenditure to be renewed. “Was my marketing effective? How effective was it? How much more revenue will we drive if I spend this much more on it?” are all questions that can be answered today.
Imagine digital advertising withered or disappeared, perhaps because of legislation or technological changes. Who would be hurt? Obviously the technology companies who profit off digital advertising would disappear. But if digital advertising were to contract to the point where it'd be no longer viable to operate businesses funded by digital advertising, there would also be other less direct casualties, in particular:
About a third (35%) of newspaper ad revenue now comes from digital advertising. Research shows that only about 11% of people are willing to pay for news subscriptions, and that half of all U.S. news subscriptions go to the Washington Post or New York Times, which means all other national news outlets (as well as regional and local news outlets) are disproportionately dependent on ad revenue.
Smaller news outlets have already been struggling over the last decade, and the business models they currently have are highly tenuous. The current COVID-19 pandemic has proven to be a grim natural experiment highlighting just how tenuous those business models are. The economic and media conditions surrounding COVID-19 have lead to reductions of 1/3 to 1/2 in advertising spending in news media, which has lead to layoffs and furloughs affecting ~36,000 news media employees (and counting). The current loss of revenue is still less than the decline of 50-60% in revenue that Google itself has estimated publishers stand to lose if current cookie-based advertising techniques are disabled and not adequately replaced.
A robust free press is essential for a free and open society, and the explosion of freely available information about every topic imaginable on the web has been an incredible boon to human development and happiness. For as long as newspapers have existed (all the way back to the 1700's) newspapers have relied on advertising to fund journalism and distribution. As more and more media moves online, non-digital advertising is waning rapidly. If all advertising moves to digital channels which independent or sub-scale publishers can't access then those publishers (and the numerous social benefits they provide) will simply disappear.
Much discourse about the health of the media focuses on consumers' willingness to pay for news; too little focuses on consumers' ability to pay. An annual subscription to The New York Times costs ~$191/year. A subscription to the Wall Street Journal costs $515/year. In other words, a subscription to just one single publication can cost 1.5% of the US median individual income. The fact is that paying for all the information necessary to be a well-informed citizen is prohibitively expensive for many people (even moreso for people in marginalized communities or in less wealthy countries). This is why so many news organizations have dropped their paywalls during the current pandemic, which of course only increases those news organizations' dependency on advertising.
In a real sense, consuming advertising is selling one's attention. For the many people who do not have disposable money to spend on news, there is real value in being able to sell their attention in exchange for information and services they'd be otherwise unable to access. If consumers are going to sell their attention, it's in their interest to get the highest price for that attention they can.
If digital advertising disappears, or the value of advertising drops precipitously, the wealthy will still find a way to get the information they need. Everyone else won't.
(not intended to be comprehensive)
Marketing strategy can be organized in many ways – by funnel, by audience, by channel type – which largely have to do with how a marketing organization itself is organized (segment, product, channel, funnel, region, function), so below definitions apply broadly, but are used in differing ways depending on company and campaign:
- Brand Marketing: an investment-based marketing strategy of marketing products and services in a way that grows the asset of brand equity over time. Brand marketing is geared towards raising awareness and changing consumer perception of a brand over time.
- Direct Response Marketing: marketing designed to evoke an immediate response and encourage a consumer to take action by opting into an advertiser’s offer. The desired response depends on the goals of the specific ad. Because of the immediacy of the consumer behavior, it provides immediate and measurable return on investment. Direct Response campaigns usually target specific audiences.
- Prospecting: marketing to new audiences with the goal of attracting new consumers who are likely to convert and become customers
- Retargeting: the repeat marketing of a piece of content to a previous website visitor on other web properties, in order to attract them back to the original content.
- Relationship/CRM Marketing: a marketing strategy designed to foster customer loyalty, interaction, and long-term relationships. It is designed to develop or reinforce a strong bond with existing customers by continuing a dialog with them that provides the customer with information that matches their needs, problems, or interests
- Consumer: I enjoy and consume content, and I and strongly prefer it to be free. I do not want intrusive or creepy advertising. I want choice. But I don’t really understand that my options are to pay with money, pay with data or pay with advertising. And I don’t understand that the players are the publisher, the browser, the device maker, the internet service provider, the advertiser and their agents; I only think about the advertiser and sometimes the publisher, but not the entire supply chain.
- Publisher: I need to support the high costs of creating and hosting content, either through advertising or through subscriptions. I am concerned about the rising difficulty to sustain my business.
- Marketer: I need to sell my goods/services in order to sustain my business and help it grow. I need to limit wasted ad spend, and I need to prove the effectiveness of my marketing spend in order to get make sure it's driving ROI.
Real-world example:
- Marketer: I am marketing for a retailer who sells diapers and I want to sell as many as possible, meaning I need to develop recurring revenue streams with existing customers and bring new customers on board.
- Consumer: I have a baby and I have tried 2 different diaper brands, but the diapers keep blowing out. My biggest problem is fit. It’s not the cuteness of the diapers. It’s not the lowest price. It’s not even getting the diapers here as fast as possible. It’s getting diapers that work with the way my baby is sized and moves. I get online on my computer and start to browse for diapers online, going to several retailers and websites to research diaper fits, and while I’m doing that an alert pops up that there is breaking news in my neighborhood, so I go to publisher.com to look it up.
- Publisher: In order to pay for the news content and the alerts service that I am providing, I need revenue. That revenue comes in the form of advertising in this case, so when the consumer visits publisher.com, I want that advertising to be effective (aka a conversion follows the advertisement), such that advertisers continue to want to advertise on my platform and such that readers continue to want to read my content.
So, how does this look from a digital marketing perspective? As a retailer of diapers, the marketer has the opportunity to connect with this consumer based on an expressed need that he has right now.
Scenario One: Branding - Prospecting
Marketer (consumer packaged goods brand): I know little about this consumer other than that they are looking for diapers with a good fit
- This consumer is targeted in my campaign through machine learning algorithms geared towards serving ads to probable customers. These algorithms are based on behaviors, demographic and/or contextual information that i.e. my DSP learns about the consumer prior to the ad opportunity being shown. These algorithms could be look alike models, or based on any number of other variables that the model shows tend to lead to conversions (i.e. researching diaper brands)
- Using dynamic creative optimization, I can tailor the message to this consumer and select my branding marketing messages that are geared around diaper fit, electing to send a video that explains why fit is so important and why ours is so good, in order to influence the awareness and perceptions this customer has of our brand
- I can also choose to follow up with display advertising on this consumer’s phone because the identity graph that my DSP has allows me to understand multiple of this consumer’s devices without knowing the name, address, phone number, etc of this consumer. I don’t need to know that type of PII—it’s not relevant.
- But, sequential messaging across channels is relevant – depending on the product/service being offered, customers will engage with a brand multiple times prior to purchasing (e.g. for most of retail, it’s 7 times), and each of those engagements are important in helping that consumer move towards solving their problem aka converting
- So, now I have the right person, the right time, the right message, the right channels and the right reason. But, how do I prove that it was effective? Since this was part of a branding campaign, in partnership with my DSP and a branding survey provider, I have a branding study set up on the campaign, which will tell me whether the campaign was effective in boosting awareness and changing perceptions of my brand
- Using an onboarding provider, my DSP onboards the list of exposed users to the brand survey provider and the onboarding provider then provides the matched list to the brand survey provider, who uses this information to send out the surveys (they will send out the surveys to a group of exposed and to a group of non-exposed users to compare the results and provide lift between the groups)
- When a campaign is multi-channel (which almost all of them are), the brand survey provider will accommodate for the different combinations of exposures across the various different channels and providers, which means that they will need all partners i.e. the DSP, the social providers, and the OLV providers to provide their exposed users using their respective identifiers and identity graphs
Consumer: I am looking for diapers that fit, so to see ads that solve my specific problem is not intrusive (assuming the ad format isn’t annoying). If these ads follow me around for a relatively short period of time, I’m okay with that. If I had to instead pay for the publisher.com content and not see any ads, that would be okay as long as that was my choice.
Large Publisher: I know a lot more about this consumer than the advertiser does. I know that they visit me regularly and the types of devices they own. I know that they never pay for content and tend to read a lot of human-interest and breaking news stories. I know that they are a 30-45 year-old male. And I know that they tend to be online sporadically, but spend a lot of time online in the overnight hours. I don’t know their name, their address, phone number etc. Based on the demographic and contextual information that I have, I can place a value on this ad slot.
Scenario Two: Retargeting
Marketer (Retailer): My data shows that this consumer has not yet made a diaper purchase with me, so I want to continue to market to him across the channels that he is engaged with. Based on the pages he visited from my website, I elect to offer relevant product and messaging ads to him across his online journey
- This type of marketing relies on following the consumer across the web using cookies to notify retargeting platforms to serve specific ads based on the specific pages visited from my website retailer.com
- I don’t need to know his name, home address, phone number, but it helps to know as much demographic, contextual, and pseudonymized data about him in order to ensure I am solving his problems in the right way
- Due to my limited budgets, I use both frequency capping and post-exposure attribution of my ads across different publisher. Individual publishers can limit the frequency of exposure within their specific websites, but require independent vendors to limit frequency capping across other publishers
- Ideal state for valuing the media spaces purchased is to be able to compare the costs incurred against the revenue gained (ROAS) and to measure the incremental impact of my marketing. Lift measurement allows marketers to understand pseudonymously who actually purchased and who didn’t purchase (using pixels for online conversions and onboarding of offline conversion data where possible), and then compare those purchases (and costs) between test and control groups. Randomized control trials, which compare test and control groups in order to measure lift, are the only mathematically accurate way to measure the effectiveness of marketing.
- Lift measurement is run in real time as the campaign is running, splitting users either into test or control groups, using the pseudonymized identifiers in the identity graph of my DSP and is based on the behaviors taken within the particular channels (video views, social clicks, display views, search clicks…). It’s important for the DSP to connect as many devices as they can to each individual so that they do not place different devices owned by the same individual in both test and control (which contaminates the results)
- NOTE: the type of KPI that is selected will be dependent on both the marketing channel and the campaign goal (i.e. a social ad that has the goal of a website visit vs a video ad that has the goal of awareness), but no matter what KPI is selected, measuring lift of that KPI is the key to measuring true marketing effectiveness
Consumer: I haven’t made my purchase yet, so these reminders are relevant and helpful as I am juggling so much on my to-do list / or at least not annoying as long as they don’t continue to follow me for a long period of time / they are annoying – it’s not creepy, yet, because I did visit their website
Small Publisher: While larger publishers will have built up significant 1P data stores, I only see small numbers of visitors today as my site is new, but my content is very good and I need a way to pay for it. Without advertising, I cannot stay in business, because most consumers who use content and services on the open web (including my site) do not pay for it. Effective advertising on the open web relies on the ability to understand some level of information about a consumer before the ad is served (contextual…demographic…CRM match)
Scenario Three: Relationship Marketing
Marketer (retailer): I have built up a relationship with this customer and I now have him in my database. I want to continue to market to him in order to upsell other products that will meet his changing needs as his babies grow. My campaign right now includes a segment on parents.
- I onboard my segmented CRM files, which include this customer’s information, into my DMP of choice, and into my DSP of choice. Based on the matches between my onboarder and my DMP and DSP respectively, I do not know if this particular customer is matched or not, but I have a high likelihood because I achieved a 90% match rate with my DSP and a 96% match rate with my DMP (I assessed the onboarder, DMP and DSP that I selected each based on match rates)
- Let’s say that I ran analysis of my data and it showed that customers make their second purchases of diapers between days 18-23. I can remarket to this customer from days 12-23, setting frequency caps during that period so as not to inundate the customer and annoy him (and waste my money), but ensuring that I am talking to him at a time when he likely needs to purchase his next set of diapers
- Then when this customer is online during this time period, my DSP will target him using the pseudonymized identifier that they have in their identity graph and decide whether or not to bid on him based on:
- price
- frequency caps (determined through the DSP’s identity graph across all known devices attached to each individual consumer)
- brand safety
- I will not learn whether this particular customer was marketed to, but I will be able to understand in aggregate the effectiveness of my marketing based on lift measurement results of the campaign (this understanding is vital to me as without it, it’s very hard for me to get my finance team to approve budgets)
Scenario Four: Re-engagement
Marketer: My customer has now not purchased with me in a year, which moves him into the lapsed customers segment. However, I would like to regain him as a customer. I have tried marketing directly to him via email and direct mail, but have not seen any resultant conversions. Because I know he is heavily digital in his day-to-day, I think that marketing to him on those channels will be more relevant to him
- I repeat the steps from relationship marketing, but this time elect to trial different messages using both dynamic creative optimization and adaptive segments because I want to react in real time based on his behaviors, tailoring my campaign and message based on what he interacts with or not
- I also set higher frequency caps and reach out to “him” on more digital channels than I have previously (via the segment – everyone in the segment is treated the same initiall,y and again, I don’t understand tindividuals tied to my CRM profiles). I can do this because my DSP has good understanding of the various devices attached to the individuals in my segment
- Based on attribution run through the DSP using the household level identity graph, the DSP is able to run lift measurement and understand that the ads led to a conversion at the household level, not at the individual level (the individual who saw the ads is not the one who made the purchase, but the HH did make a purchase)
- Using real time optimization and journey analytics, I can update my campaign parameters in real time (message, channels, context i.e. time of day, frequency caps…) to deliver the highest performance for this campaign across all of my programmatic campaigns. This relies on the strength of my DSP’s identity graph
Customer: What the retail marketer does not know about me is that I have gotten re-married and my new wife prefers another retailer. We have noticed that there have been some significant shipping issues of late that we are both not happy with. So, now we are open to finding a retailer who carries the brands of childcare products that we need (including, but expanded beyond diapers) and who can get us those products reliably and quickly. In the ads I’ve seen, this retailer messaged that they can now get me goods in <3 days. I mention this retailer to my wife after seeing a few of the ads, and she makes our next purchase from them
Scenario Five: Brand Messaging
Marketer: My company believes strongly in protecting the health of the next generation and the health of the planet they will inherit. We avoid using any potentially harmful chemicals in our diapers. Unfortunately, our competitors don't share our scruples and so our diapers are not the cheapest on the market.
We believe that most parents share our values but have never thought about the sustainability of their diapers. So they will make their purchasing decision based purely on fit and price unless they're made aware of the environmental and health implications of their decision. My company decides to run a large advertising campaign emphasizing the sustainable composition of our diapers. Our goal is not to drive immediate sales but rather to make new parents aware that not all diapers are sustainable, even if that might benefit some of our competitors who also limit harmful chemicals.
- Customers who already purchase our diapers are likely already aware of our message, so we want to target our campaign to new parents who are not already customers. We purchase "segments" containing lists of new parents and to keep costs down we use our CRM to suppress targeting parents who are already customers.
- Our CFO is skeptical that our ad campaign is a good use of money, particularly when we tell him it won't drive immediately observable sales. So we contract with a brand lift measurement vendor to run an brand lift study on our campaign. The measurement partner integrates a tracking tag into our advertisements and then uses that tracking information to survey parents who saw our ads and ask them "to what extent do you agree that diapers from {{My Brand}} are free of harmful chemicals?".
- Based on the results of the measurement study, I refine my creatives and targeting until I see a clear lift in consumer awareness. My CFO is pleased by the statistically rigorous results and authorizes a new campaign to run in the fall.
Customer: As new parents, we didn't realize that some diaper manufacturers cut corners and include chemicals in their diapers that are bad for the environment or even potentially unhealthy for our baby. I'm glad that I saw a video ad on Youtube that explained to me which chemicals to watch out for and which companies prioritize safety. I now research my diaper purchases carefully before making them, and I'm happy to spend a few more dollars on the sustainable option.
Scenario Six: Digital Marketing to Incentivize Store Visits
Marketer (retailer): I am opening a new store and I would like to let consumers in the neighborhood know about the opening and incentivize them to come visit the store.
- I ask them to target based on zip code for all zip codes within 20 miles of my store with brand messaging
- Additionally, I ask them to target all individuals in physical proximity to my new store during the first two months that it is open
- And, I take a seed audience from my customer base, onboard that to both my DMP and DSP and have them model out look alike audiences based on similar demographic information (e.g. location, device types, income…), which I use for more follow up messaging as I believe these audiences will be higher value than the zip code audiences
- In order to prove if these strategies had any influence on store traffic, in conjunction with my DSP and onboarder, I partner with a geolocation company, who uses store traffic to measure the effectiveness of my marketing against visits to the store
NOTE: Document is incomplete from this point