Digital Marketing I: Outbound

Innovation & New Business Planning - TUHH Institute of Innovation Marketing & Institute of Entrepreneurship, Hamburg · part of my Technology Management MBA · study notes for revision.

---

For most of the twentieth century a new consumer brand faced two walls it could not climb. The first was distribution: the big incumbents owned the shelf space in retail, so a startup product simply had nowhere to sit. The second was advertising: the only way to build broad brand awareness was television, and television was prohibitively expensive for a young company. Both walls have come down. E-commerce means retail shelf space is no longer the only route to a buyer, and digital advertising lets a tiny team reach a precisely chosen audience at a cost it can actually afford.

That is why the reading opens with a list of direct-to-consumer startups that reached serious valuations in only a few years. In May 2019 the razor company Schick announced it would buy Harry’s, then six years old, for 1.37 billion dollars - three years after Unilever paid about 1 billion dollars for Dollar Shave Club. The same list includes Bonobos in apparel at 310 million, Casper in mattresses at 1.1 billion, Allbirds in shoes at 1.4 billion, Glossier in beauty at 1.2 billion and Honest in baby, beauty and home products at roughly 1 billion. None of them out-spent the incumbents. They out-targeted them.

This chapter is the first half of the digital marketing picture: outbound marketing, meaning every situation where the firm starts the conversation and pushes a message towards a chosen audience. In digital channels that means three ad types - search, display and video - plus the uncomfortable question of how you prove any of it worked.

Tip

The one idea to carry through. Online advertising is not bought from a price list. It is bought in an auction, where your position depends on your bid multiplied by how relevant the platform judges your ad to be, and where the price you actually pay is set by the bidder ranked behind you. So the economics of a campaign live in only two numbers: what a click costs you and what a click turns into. Every metric in this chapter measures one of those two, and the hardest question in the whole field - attribution - is simply the argument about which ad deserves credit for the sale.

1 · Why the digital shift matters to a new venture

Digital technology changed three things at once: how people look for information, how they talk to each other, and how they buy. Cookies follow a visitor around the web. Content-curation and content-sharing sites gave individuals a way to publish and gave businesses a new place to advertise. Social platforms let a message reach a wide audience within seconds. And the unprecedented spread of mobile phones opened up whole populations, especially in developing countries, that never had fixed broadband at all.

The practical consequence for a manager is that market insight is no longer limited to surveys, transaction records and focus groups. You can hear directly what people say about your product, often know who is saying it, and sometimes to whom. That in turn allows better targeting and lets you link individual marketing touch points to sales and to customer lifetime value.

The money has followed. In 2019, digital advertising reached 129 billion dollars in the United States, which was 54 percent of total ad spend, and 333 billion dollars worldwide, about 50 percent of the total. Digital was expected to pass 60 percent of all ad spend by 2023.

Paid mediayou buy the exposure - a thirty-second television spot, a search ad

Owned mediachannels you control - exposure through your own website

Earned mediapeople share and discuss you themselves

Three sources of exposure a digital campaign juggles at the same time. Word of mouth and review sites sit alongside them and can either reduce the amount of advertising you need or amplify what you buy.

The volume of content is the reason managers feel paralysed rather than empowered. By 2018 an average internet minute contained more than 4 million YouTube video views, about 2,000 new Reddit comments and almost 500,000 tweets. Attention is the scarce resource, not media space.

2 · The framework: outbound versus inbound, and why the split matters

The reading organises everything around four elements. The first two are mirror images of each other, and getting that contrast straight is the single most examinable idea in this chapter.

Outbound marketing this chapter

• The firm initiates the conversation and sends a message to target consumers
• The direct descendant of traditional print, radio and television advertising
• Digitally it is search, display and video ads
• You are hunting for the customer, so you pay for every attempt

Inbound marketing next chapter

• The firm makes sure it can be found when a consumer searches for a product or service
• Requires aligning the website both to how people search and to search engine algorithms
• This is where search engine optimisation comes from
• Instead of hunting, you become the thing being searched for

Social media consumers create the content

• Platforms where consumers themselves create reviews and opinions that influence other buyers
• Firms take part actively to reach and serve customers in new ways

Mobile technology always connected

• Keeps the consumer connected anytime and anywhere
• Changes how people search for and buy things, and creates its own management problems

The reading gives an image that is easy to keep: outbound marketing is like looking for a needle in a haystack, while inbound marketing is like using a magnet so the needle comes to you. That is also why the split matters economically. Outbound spend scales with the number of people you try to reach, so it is a variable cost that never stops. Inbound work is closer to an asset you build once and keep.

Note

A hybrid worth knowing: native advertising. Some consumer and company interactions mix the categories. Native advertising is sponsored content designed to blend into a site’s editorial flow, distinguishable only by a promoted tag and a logo. The reason it exists is a numbers problem: ordinary banner ads have click-through rates of roughly 0.02 percent, while native pieces on one large content site ran at 1 to 3 percent. In practice an editorial list of beautiful places gets rewritten as a brand version of the same list. Native ads tend to use rich media, are usually mobile-friendly and are built to be shared. Regulators watch the format closely, because the line between editorial and paid content gets blurred.

3 · Search ads: buying a moment of intent that already exists

About 80 percent of all online sessions on PCs and laptops begin with a search of some kind, which has turned search engines into the gateway of the internet. Since people type words rather than specific URLs, the ads on a results page are unusually powerful points of interaction.

When someone types a phrase such as resort vacations, the engine returns two kinds of link. Organic links are ranked on relevance or user popularity and cost the company nothing. Paid links, also called sponsored links or search ads, are bought, and they normally sit above the organic results or in a column beside them. The same company can appear in both places at once. Firms can bid on single words as well as on multiword phrases, and different combinations of keywords produce different results.

1 · Decide the keywordswhich words to bid on, how much per keyword, and the total budget available

↓

2 · Design the ad creativethe keyword usually appears in the ad text itself, which lifts both placement and perceived relevance

↓

3 · Launch the campaignthe ad may show whenever someone searches a keyword you bid on, if it ranks high enough to make the page

↓

4 · Prepare the landing pagespurpose-built pages, typically not the homepage, holding exactly the content that search implied

The four-step search engine marketing process. Note step three: a displayed ad costs nothing. The advertiser is charged only when someone actually clicks, and even then never more than the bid.

The billing model is what makes search different from television. Search engines do not charge a fixed price. They run a generalised second-price auction and charge per click-through, and the higher the bid, the more likely the ad is to land in a visible spot. The reading’s simple illustration: if 1,000 people click and the bid is 0.10 dollars per click, the maximum cost is 1,000 times 0.10, which is 100 dollars. Because payment is triggered by a click rather than by a view, a search ad captures intent that already exists - the person told you what they wanted before you spoke. Bid size and budget therefore both follow from how effective and profitable the ads turn out to be.

4 · Keywords: curating a portfolio that can run to millions

Different people describe the same want in different words, so a keyword list is a portfolio that has to be curated and managed, not written once. The scale is genuinely large: one of the world’s biggest travel sites bought over 100 million keywords in 2010, and in the third quarter of 2018 a major hotel-booking site spent more than 1 billion dollars on Google Ads, a large part of it on search.

Four forces make the list multiply:

• Variations on one idea. Somebody looking for a hotel in Los Angeles may type Los Angeles hotel, Hollywood hotel, room in LA or three-star hotel in Los Angeles. Every modifier is another keyword attached to the same single search idea.
• Different languages, which reach customers with different demographics.
• Misspellings and typos, which firms sometimes buy deliberately, although autocorrect and spelling suggestions are making this less necessary than it was.
• Negative keywords, which do the opposite job: they stop your ad appearing on certain searches. A beauty company wants the searcher looking for skin treatment but definitely not the one searching skin cancer treatment. Ruling out the word cancer stops money being wasted on searches unlikely to convert.

Branded keywords bidding on your own name

• Your own brand terms already put you top of the organic results, so why pay at all?
• Defence. A competitor can buy a paid slot that sits above your organic link and siphon away your prospects
• Offence. One study found search ads raised the click-through rate of the same company’s organic link by 25 percent

Generic keywords bidding on the category

• A hotel chain bidding on a category phrase such as Bangkok hotels
• The searcher may not have had that chain in her consideration set at all, and the ad can pull her in to explore
• Even with no click, she now knows the chain has a hotel in that city and may look for more later

Both keyword types earn their place, for different reasons. Branded terms protect an asset you already own; generic terms buy entry to a decision that had not included you.

Whichever mix you run, the portfolio needs monitoring. Sophisticated firms use scripts, meaning programmed code that runs routine tasks automatically, but manual judgement is still needed to assess how search terms are trending, to keep up with changes in the way people phrase things, and to reposition keywords when a competitor changes tactics. Search engines help: a keyword planner tool reports the web traffic behind each keyword variant, which is how you find out that one phrasing of a term carries far more volume than another.

Caution

What this reading does and does not define. It builds keyword breadth out of concrete mechanics - single words versus multiword phrases, variations, other languages, misspellings and negative keywords - rather than out of formal match-type labels. So for this course, describe the breadth problem in exactly those terms, and be able to explain a negative keyword properly, since that is the one exclusion mechanism the text actually defines.

5 · The auction: what decides your position and what you really pay

Where an ad sits matters because of where eyes go. A 2005 eye-tracking study produced a heat map showing attention concentrated on the top left of the results page, a zone marketers named the golden triangle, which made it the most valuable and therefore the most expensive position; for ads down the right-hand side, the top ones drew most attention. A 2014 repeat study found the eye still starts top left, but now scans the page quickly and vertically rather than horizontally, which is attributed to smartphone scrolling habits and to changes in how ads are served on desktops. People now take 8 to 9 seconds to find the relevant result, against 14 to 15 seconds a decade earlier.

The useful implication is that capturing the very top result is less critical than it was, because people scan the top three or four anyway. Higher position still draws more attention and more clicks, but the top slot is not automatically the most profitable one: click-through rate falls as position drops, yet the conversion rate can rise, because somebody who clicks a lower link has implicitly expressed stronger interest in that particular company. Only continuous testing, measurement and analytics reveal the right position and bid for a given keyword set.

Position is decided by two things multiplied together.

Cost-per-click bidwhat you are willing to pay for one click on that keyword

×

Quality scorethe engine’s own 1-to-10 judgement of your ad; the algorithm is kept secret

→

Ad rankyour placement position on the results page

Ad rank is the product of the two, not the bid alone. That single design decision is what stops a rich advertiser filling the page with irrelevant ads that would annoy and alienate searchers.

The bid. What you are prepared to pay depends on competition, your search budget, the click-through you expect and the conversion rate, meaning the share of clicking visitors that results in a sale or another desired action such as registering. Keywords with heavy traffic cost more, and prices vary enormously by category: in 2018, travel companies paid on average 1.53 dollars per click on desktop and 2.43 on mobile, while educational institutions paid 18.91 dollars on desktop and 13.95 on mobile.

The auction rule. The mechanism is a generalised second-price auction, built on Nobel-prize-winning auction research. The highest bidder takes the first position but pays the second-highest bidder’s amount, the second bidder pays the third’s, and so on down the list. The point of the design is that it produces truth-telling: the best strategy is to bid your genuine maximum willingness to pay.

The quality score. The engine assigns each ad a score in order to balance the goals of three parties at once - the consumer, the advertiser and itself. It rests on three components.

Potential click-through rate of the adRelevance to consumersQuality of the landing page

Each component has a reason behind it. Click-through rate, clicks divided by impressions, is a measure of revealed preference: users do not click an ad unless they find it helpful, and it is also how the engine earns, since it is paid only when someone clicks. Relevance is the match between the words the person searched and the content of the ad. Landing page quality is measured by the bounce rate, meaning how quickly visitors leave after arriving - a factor included specifically to stop an advertiser baiting a click and dumping the visitor on a page inconsistent with the ad and irrelevant to them.

The worked auction from the reading

Four advertisers compete for three available slots on one keyword. The Price Paid column shows what each would have paid under a plain second-price rule with no quality score; the last column shows the real outcome.

Advertiser	Maximum bid	Price paid without quality	Quality score	Bid × quality	Ad rank	Actual cost per click
W	4.00	3.00	1	4	4	not shown
X	3.00	2.00	3	9	2	8 / 3 = 2.67
Y	2.00	1.00	6	12	1	9 / 6 = 1.50
Z	1.00	-	8	8	3	4 / 8 = 0.50

Advertiser Y bids only 2 dollars against W’s 4 and still takes first position, because its quality score of 6 beats W’s score of 1. With only three slots available, W is not shown at all despite bidding highest. The price each winner pays is the minimum needed to hold its rank against the advertiser below it, which gives the rule:

Price needed to hold your rank

P1 * Q1 = B2 * Q2

Rearranged for the price

P1 = B2 * Q2 / Q1

Applied to advertiser Y

Price(Y) = 3 * 3 / 6 = 1.50

Three conclusions the reading draws from this, all worth memorising:

• You almost always pay less than your bid. Y pays 1.50 against a 2.00 bid. The platform guarantees you are never charged more than your maximum, and usually you are charged less.
• A higher quality score lowers your cost per click. That is the deliberate incentive: make the ad more relevant to consumers instead of simply bidding more to grab attention.
• Your actual cost depends on competitors you cannot see. Their bids and quality scores are unknown to you and change over time, so it is impossible to know in advance exactly what a search campaign will cost. The only certainty is that it will not exceed the budget you set.

6 · The metric set, and the profitability formula

Unlike traditional media, digital advertising lets a firm track consumers automatically after exposure, so assessing effectiveness is comparatively simple. These are the definitions to know cold.

Impressions

the number of times the ad was seen by consumers

Click-through rate

CTR = clicks / impressions

Cost per click

CPC = total spend / clicks

Cost per thousand impressions

CPM = (total spend / impressions) * 1,000

Conversion rate

CR = conversions / clicks

Cost per acquisition

CPA = total cost / customers acquired

Impressions show the reach of an ad but nothing about its effect, since simple exposure does not prove anyone was influenced or even noticed - which is exactly why search engines do not charge for them. Click-through rate and conversion rate are the better signals, and the CTR of most search ads sits in the low single digits.

The reading’s illustration is a bank’s 2010 search campaign across five sites: total spend 516,070 dollars for 11,529,237 impressions, 472,433 clicks and 7,107 completed checking-account applications, an overall CTR of 4.10 percent and a conversion rate of 1.50 percent.

Site	Media spend	Impressions	Clicks	Applications	CTR	Conversion rate
Google	288,000	5,575,637	234,963	3,717	4.21%	1.58%
MSN	37,000	897,406	50,242	663	5.60%	1.32%
SuperPages	70	116,922	1,410	16	1.21%	1.13%
Yahoo!	177,000	4,435,709	165,166	2,419	3.72%	1.46%
Unified Marketplace	14,000	503,563	20,652	292	4.10%	1.41%
Total	516,070	11,529,237	472,433	7,107	4.10%	1.50%

Reading the first row the way the text does: CTR is 234,963 divided by 5,575,637, which is 4.21 percent. CPC is 288,000 divided by 234,963, which is 1.23 dollars. Conversion rate is 3,717 divided by 234,963, which is 1.58 percent. Add the margin and you get profitability.

Search ad profit

(impressions * CTR * conversion rate * margin) - search ad cost

Applied at 100 dollars margin per application

(5,575,637 * 4.21% * 1.58% * 100) - 288,000 = 83,700

The same thing said plainly: 3,717 completed applications at 100 dollars each is 371,700 dollars, minus the 288,000 spent, giving 83,700 dollars. Small differences appear if you recompute from the rounded percentages. This calculation is a good first pass at return on investment, but it quietly ignores attribution, which section 9 takes apart.

7 · Display ads: buying context instead of intent

A search ad reaches somebody only while they are hunting a specific keyword. A display ad can be placed on any website the brand considers relevant to its target audience - a card issuer advertising on a financial newspaper’s site, for instance. So display ads carry two possible objectives: they can be built like search ads for immediate action, or like television and print ads with the goal of raising brand awareness.

Format	What it is
Banner ad	One of the oldest and most traditional forms, normally across the top of a page
Interstitial ad	A full page shown before the user reaches the page requested; captures more attention but is more intrusive
Expandable banner	A banner that automatically expands over a large part of the screen, sitting between banner and interstitial in size and impact
Overlay	Like an interstitial but with a transparent background, so the original page stays visible behind it
Rich media	Ads carrying interactive components such as video, audio or clickable links
360-degree ad	Designed for phones, giving a surrounding view as the user moves the handset around

The industry: networks, exchanges and real-time bidding

Search is concentrated in a few engines. Display is the opposite - a fragmented industry with many specialised players along the value chain between the advertiser and the viewer.

Advertisersfrom global consumer goods firms to small startups; large ones hire agencies, small ones use self-serve online tools

→

Ad networksaggregate ad space from many publishers and match it to demand, plus proprietary placement algorithms

→

Ad exchangesautomate the match - programmatic buying via real-time bidding, priced by supply and demand like a stock exchange

→

Content publishersfrom national newspapers to personal blogs; rates depend on traffic, prestige and audience demographics

The spine of the display industry. Large advertisers sometimes buy directly from large publishers, but most transactions pass through an intermediary. By 2017 almost 72 percent of US digital display spend was bought programmatically through exchanges, with each auction settled in a fraction of a second.

Three more roles sit around that spine. Demand-side platforms act as the buyer’s broker, letting an agency manage ad inventory across many exchanges and networks through one interface. Supply-side platforms do the mirror job for publishers, helping them get the best prices for their inventory from various buyers. Data management platforms collect, interpret and sell browsing and other customer information so targeting can be sharper, and specialists in privacy, tag management and analytics fill out the rest of a sprawling landscape.

Pricing, and the click-through problem

Display inventory is bought in one of two ways, depending on the goal. On an impression basis, using CPM, you pay every time the ad is seen whether or not anyone clicks - which suits brand awareness or brand image. On a click basis you pay per click, which suits a direct-response campaign chasing clicks and conversions. Some channels let you choose either.

Display profit, bought on CPM

impressions * [(CTR * conversion rate * margin) - CPM / 1,000]

Display profit, bought on CPC

impressions * CTR * [(conversion rate * margin) - CPC]

The same bank’s display campaign shows how differently these ads behave. Across seven networks it spent 634,000 dollars for 308,779,216 impressions, 139,176 clicks and 7,190 applications, giving a CTR of 0.05 percent and a conversion rate of 5.17 percent. Taking the largest network, at a margin of 100 dollars per applicant:

CPM

(176,000 / 97,466,342) * 1,000 = 1.81

CTR

34,777 / 97,466,342 = 0.04%

Conversion rate

2,934 / 34,777 = 8.44%

CPC

176,000 / 34,777 = 5.06

Display profit

97,466,342 * (0.04% * 8.44% * 100 - 1.81 / 1,000) = 117,400

Same result the direct way

(2,934 * 100) - 176,000 = 117,400

Now compare the two channels side by side, because the pattern is counter-intuitive and very examinable.

Metric	Search, one engine	Display, one network	How to read it
Click-through rate	4.2%	0.04%	Search wins by a factor of about a hundred
Cost per click	1.23	5.06	Search clicks are far cheaper
Conversion rate	1.58%	8.44%	Display clicks are far rarer but much better qualified

That tiny display click-through rate is the format’s structural weakness, and it is the gap that native advertising and better targeting exist to fill. The counterweight is that display genuinely moves brand measures even without clicks. A test-and-control study of 6,400 display campaigns over 2008 to 2010 found a significant effect on brand awareness, favourability and purchase intent, and a 2018 study of 675 campaigns from 2008 to 2017 found that against a control group display raised unaided brand awareness by up to 12 percent, improved brand perceptions by 2 percent and lifted purchase intentions by 3 percent.

Four ways to make display work harder

Contextual ads match the ad to the page topic

• A car advert on a car review website reaches someone already interested in cars and may even be useful information
• The cheapest and most reliable improvement available

High-visibility formats pop-ups and auto-play

• Effective at grabbing attention, and widely found annoying for exactly that reason
• Important nuance: they add nothing incremental once an ad is already contextual

Retargeting behavioural, driven by a cookie

• Shown to people who previously visited your website; that visit created a cookie, and the ad follows them onto unrelated pages later
• Works best while the consumer is still actively searching in the product category
• Overdone, it is perceived as stalking and creates adverse feelings towards the brand

Morphing ads algorithmic personalisation

• Unlike static print, these vary between individuals visiting the same page, selected from a portfolio using known attributes such as demographics and sites visited
• Matching the ad to inferred cognitive style almost doubled click-through in a field test
• On a car website, matching style plus buying stage plus body-type preference more than tripled CTR and lifted brand consideration and purchase intent

The morphing result has a neat explanation: some people are persuaded by facts and numbers while others grasp a concept through graphics, so serving each the version that suits them is not a trick, it is better communication.

8 · Video ads

Where display ads are static images on websites or apps, video ads show moving ones. Many firms simply file video under display, because the advertising objective and the measurement overlap - one platform can target the same audience with display on its partner network and with video on its video site, and the big social platforms accept both formats. But video is growing fast enough to deserve separate handling.

The five basic steps are the same as for any outbound campaign: set the goal (brand awareness, traffic to the website, more purchases), select the target audience (age group, geography, interests), create the ad, set a budget, then measure and monitor the results.

Skippable in-streamthe viewer has an exit

Plays before, during or after the main video and can be skipped after a few secondswhoever stays has chosen to stay

Non-skippable in-streamno exit

A 15 to 20 second ad the user cannot skipguaranteed completion, at the cost of goodwill

Bumpervery short, no exit

A non-skippable video running under six secondsbuilt for reach and repetition rather than a full message

The three common video formats. Skippable versus non-skippable is really a trade between a self-selected, more interested audience and a guaranteed but possibly resentful one.

How video is priced and judged. Because most video ads are designed to build awareness and brand rather than to trigger a click, click-through rate is the wrong yardstick even though it works well for search. Video is normally bought and measured on impressions using CPM, exactly as television campaigns are. Some video does carry a specific call to action that justifies a click, and digital-only measures such as social shares and post-view engagement try to capture effects television never had - but none of these lets a marketer work out the exact return on the campaign, and linking video to sales or purchase intention stays difficult and is sometimes not possible at all.

Video also no longer stops at video websites. Smart televisions and streaming services let brands target television audiences the way they target digital ones. In 2019 that streaming slice was under 1 percent of the 129 billion dollar US online advertising market, but it was expected to rise rapidly, and the broadcasters were adding the capability to make targeted television advertising more effective and more measurable.

9 · Measuring what outbound really did

Everything so far judged ads by impressions or click-through. Those are the common approaches, and they are also the shallow ones. The reading names five deeper problems.

Correlation versus causationAttributionDynamics, the delayed impactOnline and offline interactionCustomer lifetime value

Correlation versus causation. A large field experiment run on a major online marketplace found that measuring clicks and sales does not prove that the advertising caused the increase in sales. Branded keywords showed no measurable short-term benefit, and returns on paid search came out far below conventional estimates. The argument is simple and devastating: many people who click a search ad would have clicked the organic result, or bought anyway. Traditional methods merely correlate clicks with sales, and only an experimental study can uncover causality. The search platform disputed the finding, citing its own work across hundreds of advertisers where more than 89 percent of search ad clicks were incremental, and 50 percent were still incremental even when the advertiser already held the top organic result. Outcomes differ so much between advertisers that the honest exam answer is that the question remains open, and that each advertiser should experiment on its own campaigns.

Attribution. Before buying, a consumer may become aware of a brand through television or print, be influenced by online display ads, and finally purchase by clicking a search ad. Last-click attribution hands the entire credit to that final search ad, even though the earlier channels did the work of moving her along the buying-decision journey. Ignoring the assisting media therefore inflates the apparent effectiveness of search.

Attribution model	How credit is assigned	The catch
Last interaction	The last ad or click takes 100 percent of the sale	Ignores long-term effects of ads earlier in the funnel; over-weights frequently shown ads
First interaction	The first ad or click takes 100 percent	Ignores the later ads that actually converted the customer; over-weights frequent ads
Linear	Every ad or click gets a uniform weight	Ad hoc allocation; over-weights frequent ads
Time decay	Everything gets some credit, more recent touches weighted higher	Ad hoc weights - how much decay? Over-weights frequent ads, and it is unclear why later ads deserve more
Position-based	Credit depends on the position of the touch in the path	Ad hoc weights
Regression or model-based	Weights estimated from data on the whole interaction path	A scientific way to allocate weight, but blind to the fact that some ads are seen only because they sit on a relevant site, even with no impact of their own
Experiment-based	A and B testing decides	The most accurate way to determine effectiveness, but hard to manage across many networks and consumers, and difficult or expensive to run

The first five are commonly used but are ad hoc models with no scientific grounding. Only the last two are rigorous: model-based evaluation uses existing data on the consumer’s interactions along the purchase path to estimate the effect of each one, while experiments randomise ads to treatment and control groups, so the difference in response or conversion between the groups can genuinely be attributed to ad exposure. That is also the practical answer to the causation problem above, and it is why A and B testing appears in the glossary as an experiment that varies a single element, most often ad copy or landing page design, to see which version drives the desired result better.

Attribution also has to handle interaction between formats, because search and display do not work in isolation. A study across eleven advertisers found that combining search and display lifted conversion by 22 percent over search alone, and produced a conversion rate 400 percent better than display alone. Several academic experiments confirmed that ignoring this interaction leads to underestimating display.

Dynamics, the delayed impact. For expensive or complicated purchases such as a car or a brokerage account, nobody buys the moment they see an ad. People spend weeks or months looping through online and offline sources including review sites, a pattern one platform labelled the zero moment of truth, and search heat maps for cars show months of intense information gathering before the purchase. Conventional measurement focuses on short-term effects only. A study of a bank’s online campaign using a multivariate time-series model found that ignoring these long-term effects understated return on investment by up to 40 percent, and that the dynamic effect was much stronger for search than for display - which argued for shifting a greater share of the online budget towards search, even after crediting display for its assist role.

Online and offline interaction. Two problems here: the synergy between the two, and the effect of each on the other’s sales, which is the omnichannel question. Offline interactions turned out to be as important as online ones across many categories, and most influential where personal fit matters (clothing, footwear, personal care and beauty) or where real-life experience is valuable (cars, over-the-counter health products, home goods, groceries). One car company that modelled both channels together concluded its optimal offline spend should be 12 percent lower and its online spend 7 percent higher, so planning the two budgets separately leaves money on the table. Where consumers research online and buy in a shop, plain CTR and conversion rate badly understate the online ads - and if the firm evaluates its e-commerce and store managers on separate channel numbers, it will penalise the very website that is driving people into the stores. Experiments settle it: one furniture retailer found an 11 percent rise in store visits among people shown social ads against an identical control group, and one fashion brand attributed a 31 percent increase in in-store transactions to the same channel.

Customer lifetime value. Click-through rate does not always correlate with return, so the endpoint has to be sales or profitability, not clicks. Mobile gaming is the cautionary tale: firms used to judge digital campaigns on cost per install, until it emerged that most games retain only about 35 percent of players after day one and lose a staggering 94 percent within a month, while fewer than 4 percent of players ever spend money on in-app currency or items. A channel can look excellent on CTR or cost per install and still deliver nobody who pays, so the industry moved to customer lifetime value. The same logic can flip a budget decision: a bank may find online acquisition cheaper than traditional channels, but if online-acquired customers like to shop around they may be more price sensitive, retain worse and be worth less over their lifetime than customers acquired offline.

Worked example

A first search campaign for a small product, run for one month. Start with five inputs and derive everything else.

Clicks

clicks = impressions * CTR

Spend

spend = clicks * CPC

Conversions

conversions = clicks * conversion rate

Cost per acquisition

CPA = spend / conversions

Return on ad spend

ROAS = revenue / spend

Contribution, in the reading’s form

(impressions * CTR * conversion rate * margin) - spend

Step	Working	Result
Impressions bought	given	2,000,000
Click-through rate	given	1.50%
Clicks	2,000,000 × 1.50%	30,000
Cost per click	given	0.80
Spend	30,000 × 0.80	24,000
Conversion rate	given	2.50%
Conversions (orders)	30,000 × 2.50%	750
Value per order	given	60.00
Revenue	750 × 60	45,000
Cost per acquisition	24,000 / 750	32.00
Return on ad spend	45,000 / 24,000	1.88
Contribution	45,000 - 24,000	21,000

Return on ad spend here is simply revenue divided by spend; the reading itself states the same comparison as the subtraction on the last row, value of the conversions minus the ad cost.

Now move one input at a time. Nothing else changes in either case.

Scenario	Clicks	Conversions	Spend	CPA	ROAS	Contribution
Base case	30,000	750	24,000	32.00	1.88	21,000
Conversion rate falls from 2.50% to 2.00%	30,000	600	24,000	40.00	1.50	12,000
Cost per click rises from 0.80 to 1.00	30,000	750	30,000	40.00	1.50	15,000

The sensitivity is the lesson. Losing one fifth of the conversion rate - half a percentage point, the kind of change a worse landing page causes - erases 43 percent of the contribution while the spend does not move at all. A 25 percent rise in the click price, the sort of thing a competitor entering the auction produces, costs 29 percent of the contribution. Break-even is where CPA equals the 60 dollar order value, which needs 400 conversions from those 30,000 clicks, a conversion rate of 1.33 percent. Below that the campaign burns money, so the whole distance between profit and loss is a little over one percentage point of conversion.

Apply it to your project

1. Decide the objective before the format. Immediate action points you at search ads and click-based pricing. Awareness points you at display or video and impression-based pricing. Buying the wrong format for the goal makes the metrics meaningless.

2. Write the keyword list, both branded and generic. Include your own name to defend it, category terms to reach people who have not heard of you, and the obvious variations and phrasings. Use a keyword planner to check the traffic behind each one.

3. Add negative keywords immediately. List the searches that share your words but never your intent, and exclude them before launch. This is the cheapest saving available.

4. Build a real landing page for each ad group. Not the homepage - a page carrying exactly what the search implied. It lifts conversion, and it lifts your quality score through the bounce rate, which in turn lowers your cost per click.

5. Set the bid from the economics, not the ambition. Work backwards: value per conversion times the conversion rate you expect gives the most a click can be worth. Bid your genuine willingness to pay, since the second-price auction rewards truth-telling, and set a hard budget ceiling because you cannot see your competitors.

6. Do not fight for position one automatically. Test lower positions too. Click-through falls but conversion can rise, and the cheaper click may be the profitable one.

7. Instrument the funnel end to end so you can read impressions, clicks, CTR, CPC, conversions, conversion rate, CPA and revenue per campaign and per keyword, not just per account.

8. Run a holdout experiment, not just a report. Randomise a control group that sees no ads and compare conversions. This is the only way to separate the sales you caused from the sales that would have happened anyway.

9. Distrust last-click before you reallocate budget. Check whether display or video assisted the conversions that search is claiming, and allow for the delayed effect if your product has a long decision period.

10. Judge the channel on lifetime value, not on cost per acquisition. A cheap acquisition that churns immediately is more expensive than a costly one that stays.

Key terms

Term	What it means in plain words
Outbound marketing	The firm initiates the conversation and pushes a message at target consumers; online that means search, display and video ads
Inbound marketing	Making sure the firm can be found when the consumer searches, which is where search engine optimisation comes from
Organic link	An unpaid result ranked on relevance or user popularity
Paid link (search ad)	A sponsored result the advertiser pays for, usually above or beside the organic ones
Generalised second-price auction	The bidding rule where the winner takes the top slot but pays the next bidder’s amount, which encourages honest bidding
Quality score	The engine’s secret 1-to-10 rating of an ad, built from likely click-through rate, relevance to the consumer and landing page quality
Ad rank	Where the ad lands on the page, set by bid multiplied by quality score rather than by bid alone
Negative keyword	A word you exclude so your ad never appears on a search that shares your terms but not your intent
Landing page	The purpose-built page a click delivers you to; its bounce rate feeds the quality score
Impression	One instance of an ad being displayed; shows reach, proves nothing about effect, and search engines do not charge for it
Click-through rate (CTR)	Clicks divided by impressions, as a percentage; low single digits for search, a fraction of a percent for banners
Cost per click (CPC)	Total spend divided by clicks, the amount paid to get one click
Cost per thousand impressions (CPM)	Spend divided by impressions, times one thousand; the pricing used when the goal is awareness
Conversion rate	The share of clicking visitors who complete the desired action, such as a purchase or a registration
Cost per acquisition (CPA)	Total cost divided by the number of customers, purchases or buyers acquired
Return on ad spend	Revenue produced by the advertising divided by the money spent on it; the reading states the same idea as value of conversions minus ad cost
Ad network and ad exchange	A network aggregates publishers’ space and matches it to demand; an exchange automates the match through real-time bidding, which is called programmatic buying
Retargeting	Display ads shown to people who already visited your site, triggered by the cookie that visit created
Last-click attribution	Giving all credit for a sale to the final click, which inflates search and hides everything that assisted it
Customer lifetime value (CLV)	The present value of future cash flows from a customer relationship; an upper limit on what you should pay to acquire one

Test yourself

1. Explain in one sentence each what outbound and inbound marketing are, and give the reading’s image for the difference between them.
2. What two factors determine the rank of a search ad, and what three components make up the second of them?
3. Auction calculation. Take the four advertisers from section 5 and suppose advertiser Y raises its quality score from 6 to 10 while everything else stays the same. Who now ranks first, second and third, who misses out, and what does the first-ranked advertiser pay per click?
4. Metric calculation. From the bank’s search table, the Yahoo! row shows 177,000 dollars of spend, 4,435,709 impressions, 165,166 clicks and 2,419 completed applications. Work out the CTR, the CPC, the conversion rate, the cost per acquisition, and the profit if each application is worth 100 dollars of margin.
5. Search ads on one engine had a CTR of 4.2 percent and a conversion rate of 1.58 percent, while display ads on one network had a CTR of 0.04 percent and a conversion rate of 8.44 percent. Explain why that pattern is not a contradiction.
6. Why does last-click attribution overstate search advertising, and which two attribution approaches does the reading treat as rigorous rather than ad hoc?

Answers

1. Outbound is where the firm initiates the conversation and sends a message out to target consumers, using search, display and video ads. Inbound is where the firm makes itself findable so consumers arrive of their own accord, which requires aligning the website to consumer search behaviour and to search engine algorithms. The image: outbound is hunting for a needle in a haystack, inbound is using a magnet to draw it in.
2. The cost-per-click bid and the quality score, multiplied together to give ad rank. The quality score is built from the ad’s potential click-through rate, its relevance to consumers, and the quality of the landing page as judged by bounce rate.
3. The bid times quality products become W = 4, X = 9, Y = 20, Z = 8. So Y ranks first, X second, Z third, and W is not shown, since there are only three slots. Y pays the second-ranked advertiser’s bid times quality divided by its own quality: 3 times 3 divided by 10, which is 0.90 dollars. X would pay 8 divided by 3, which is 2.67, and Z would pay 4 divided by 8, which is 0.50.
4. CTR = 165,166 / 4,435,709 = 3.72 percent. CPC = 177,000 / 165,166 = 1.07 dollars. Conversion rate = 2,419 / 165,166 = 1.46 percent. CPA = 177,000 / 2,419 = 73.17 dollars. Profit = (2,419 times 100) minus 177,000 = 241,900 minus 177,000 = 64,900 dollars.
5. The two formats catch people at different moments. A search ad meets somebody who has just typed the want, so a large share of viewers click, but many are still browsing, so relatively few convert. A banner interrupts somebody doing something else, so almost nobody clicks - and the rare person who does click is unusually self-selected and interested, which lifts the conversion rate. The trade shows in the price too: search clicks cost 1.23 against 5.06 for display in that example.
6. Because the consumer usually met the brand several times before buying, and last-click hands the whole sale to the final touch, which is typically a search ad, while the display, video or television exposure that moved her along the journey gets nothing. The two rigorous approaches are regression or model-based attribution, which estimates each interaction’s weight from data on the whole path, and experiment-based attribution, which randomises consumers into treatment and control groups so the difference in conversion is genuinely caused by the ads.

Revision summary

Tip

The one-minute recap

• Outbound means the firm pushes the message out - search, display and video ads; inbound means being findable when the consumer searches. Needle in a haystack versus a magnet. Social media and mobile complete the four-element framework.
• Search ads buy an intent that already exists. Four steps: pick keywords, design the creative, launch, build landing pages. Being displayed costs nothing; you are billed only on a click, and never above your bid.
• Keyword portfolios explode through variations, languages and misspellings, and need negative keywords to block searches that share your words but not your intent. Bid on branded terms to defend the slot above your organic link, and on generic terms to enter consideration sets you were not in.
• Ad rank = bid × quality score, with the quality score built from potential CTR, relevance, and landing page quality via bounce rate. In a generalised second-price auction you pay the minimum needed to hold your rank: P1 = B2 × Q2 / Q1. A higher quality score lowers your price; a big bid with a poor score can miss the page entirely.
• Core metrics: CTR = clicks / impressions, CPC = spend / clicks, CPM = spend / impressions × 1,000, conversion rate = conversions / clicks, CPA = cost / customers acquired. Profitability: (impressions × CTR × conversion rate × margin) - ad cost.
• Display trades intent for context, is bought through networks and exchanges using real-time bidding (about 72 percent of US display spend by 2017), suffers dismal click-through, and is improved by contextual placement, retargeting from cookies, and morphing ads matched to the individual.
• Video is measured on impressions and bought on CPM, because most of it builds brand rather than clicks. Formats: skippable in-stream, non-skippable of 15 to 20 seconds, and bumpers under six seconds.
• Measurement is the hard part: correlation is not causation and only experiments prove an incremental effect; last-click attribution flatters search; long-term effects can hide up to 40 percent of the return; online and offline budgets planned separately lose money; and the right endpoint is customer lifetime value, not clicks or installs.

Next: Digital Marketing: Inbound, Social & Mobile → - getting found instead of buying attention.