Marketing Analytics: Data-Driven Techniques with Microsoft Excel (2014)

Part IX. Advertising

Chapter 36. Pay per Click (PPC) Online Advertising

Online advertising is big business. In fact, in 2012 Google alone generated $44 billion in revenue from online advertising. The vast majority of online ad revenue comes from Pay per Click (PPC) advertising in which advertisers pay only when an Internet user clicks through to the advertisers' website. Online advertising revenues are still growing quickly (around 20 percent per year), so it is important to understand how analytics can help advertisers optimize their profits from PPC advertising. The chapter begins by showing how an advertiser can determine if PPC advertising is likely to be profitable. Then it explains how Google runs an auction for ad slots based on a keyword search. Finally, this chapter discusses how advertisers can use Google's Bid Simulator tool to optimize their bid on a keyword.

Defining Pay per Click Advertising

Internet users who want to purchase a product often search the Internet for information on the product. In addition to showing search results, search engines also show paid ads relevant to the users' search. Companies such as Yahoo and Google can, of course, charge advertisers to have their ads appear when users conduct a search. For example, Figure 36.1 shows the results when the author searched for swim goggles. You can see ads appear in both the top search results and on the right side of the page. Following are some common questions pertaining to these ads that a marketing analyst would benefit from asking:

·        How does Google determine which ads for goggles appear and who gets each spot on the page?

·        How much should advertisers pay for each spot on the results page?

·        Advertisers can purchase certain keywords, and their ads appear when a user searches for those keywords. On which keywords should advertisers bid?

·        How can Google provide a good experience for both searchers and advertisers and still maximize its profits from search-related ads?

Figure 36-1: Search results for swim goggles

Before 2002, most Internet advertisers paid search engine providers such as Google or Yahoo based on the number of impressions. An impression occurs whenever a company's ad is displayed after a search. When advertisers pay based on the number of impressions, they may find after paying Google that nobody clicked through to their webpage! For obvious reasons, advertisers did not like this method of charging for ads. The website Basketball-reference.com provides a great example of impression advertising. There is a webpage for each NBA player (past and present) containing the player's career statistics. On each webpage is a price for which an advertiser can sponsor the page and post ads on the page. For example, the price of sponsoring the LeBron James page is $1,915, whereas the price of sponsoring the Carmelo Anthony webpage is $415. In this situation the price difference probably reflects the fact that the LeBron James webpage is viewed 1,915 / 415 = 4.6 times as often the Carmelo Anthony webpage. In a different situation the relative price of sponsoring a webpage might depend on the attractiveness to the advertiser of the demographics of the webpage viewers as well as the number of times the webpage is viewed.

When a search engine provider charges based on PPC, the advertisers know that they are charged only when a searcher clicks through to their website. Because click-throughs should lead to sales, advertisers know they are getting some value for their payments. The first instance of PPC advertising was developed in 1996 by a division of Packard Bell NEC Corporation. Beginning in 2002, the AdWords system was used for PPC advertising. From that point on PPC advertising took off.

Profitability Model for PPC Advertising

To estimate monthly profitability of PPC advertising, estimates of the following quantities are needed:

·        Estimated Cost per Click: This is the cost the advertiser must pay for each click. Assume a $1 cost per click.

·        Estimated Clicks per Day: This is simply the daily number of clicks to their site expected by the advertiser. Assume your ads will generate 10 clicks per day. If you sign up for Google AdWords, Google's Bid Simulator feature (discussed in the “Using Bid Simulator to Optimize Your Bid” section of this chapter) gives you an estimate of the number of clicks per day that can be obtained for a given cost per click.

·        Conversion Rate: This is the fraction of clicks that results in a sale. The conversion rate can easily be estimated from historical data. Assume a conversion rate of 5 percent.

·        Average Profit per Sale: This can easily be estimated from historical data. Assume an average profit per sale of $10.00. Assume a 30-day month.

You can see these calculations in action with the following example. The Simple Model worksheet in the Costperclickoptimization.xlsx workbook (see Figure 36.2) shows the calculations needed to determine if PPC advertising can help your company's bottom line.

Figure 36-2: profitability analysis for PPC advertising

In the cell range E15:E18, you can use some simple calculations to compute the estimated profit per month from PPC advertising.

1. In E15 compute conversions per month by multiplying the conversion rate by the number of clicks per month. The exact formula is =Conversion_Rate*Clicks_per_day*Days_per_Month.

NOTE

Multiplying the units of these three quantities yields units of (conversions / click) * (clicks / day) * (days / month) = (conversions / month), as desired.

2. In cell E16 compute the monthly profit by multiplying the expected profit per conversion times the monthly number of conversions. The exact formula is =Conversions_per_Month*Profit_per_sale.

3. In cell E17 compute the monthly click costs by multiplying the cost per click times the number of monthly clicks. The exact formula is =Clicks_per_day*Days_per_Month*Cost_per_click.

4. In cell E18 compute the monthly profit with the formula =Profit-Click_Costs.

Given the assumptions, monthly profit for PPC advertising is -$150.00, so it does not appear that PPC ads would be profitable. A click-through to your webpage might, however, create a new customer who may repeatedly purchase your product. In this case the profit per sale should be replaced by a larger number: the lifetime value of the customer. (See Chapters 19–22 for a discussion of customer value.) Therefore incorporation of the concept of customer value might justify the use of advertising.

An advertiser can use a few simple calculations to determine whether PPC advertising will be profitable. Because you break even if profit per click equals 0, clicks per month do not impact a break-even calculation. Equation 1 provides the key to break-even analysis.

1 

Rearranging Equation 1 you can find you will break even if:

2 

3 

Substituting your assumptions into Equation 2, you can find the break-even conversion rate to be 1 / 10 = 10%. From Equation 3 you can find the break-even Cost Per Click = ($100) * (0.05) = $0.50. Therefore, you can break even by either doubling the conversion rate or cutting the cost per click in half.

Google AdWords Auction

Google uses the bids and quality scores of advertisers to rank advertisers through their popular tool, Google AdWords. Google's AdWords auction matches up online advertising slots with online advertisers.

Considering the discussion of online ad auctions from well-known economics professor turned Google's chief economist, Hal Varian's, elegant paper “Online Ad Auctions,” (American Economic Review, 2009, pp.1-6), suppose that N advertisers bid on the same keyword, and advertiser i bids an amount bi as the maximum per click price he is willing to pay. Also assume there are S ad slots available for the keyword. At first glance you would think that Google would rank bidders simply on their bids with the highest bidder getting the top slot, the second highest bidder the second slot, and so on. The problem with this approach is that the top slot might go to a company bidding a lot, but that company could have a terrible ad that nobody clicks on. This would upset searchers who wasted their time reading the poor ad, and because nobody clicked through, Google would earn no money for the valuable top slot. Therefore, Google assigns each bidder a quality score (qi = quality score for bidder i). A bidder's quality score (a value between 1 and 10, with 10 indicating the highest quality and 1 the lowest quality) is mostly determined by Google's estimate of the ad's click-through rate. Click-through rate = clicks through to webpage / (total times an ad appears). The quality score also includes factors such as Google's evaluation of the quality of the ad, relevance of the ad to the keyword, quality of the advertiser's webpage (often referred to as the landing page), and relevance of the landing page to the keyword. The bidders are now ranked by the product of their bid and quality score (biqi). With this information Google can rank-order bidders, for instance, if there are say five slots, and you rank sixth, your ad does not appear!

Determining what an Advertiser Pays Per Click

You might intuitively believe that an advertiser's Cost per Click would be determined by that same advertiser's bid and quality score. Surprisingly, the amount Google charges an advertiser per click is actually determined by the bid and quality score of the advertiser in the ad position immediately below said advertiser. The intuition behind this result will be discussed later in this section.

The following discussion ignores the possibility of ties. Assume that M advertisers bid on S slots. After ranking the bidders based on biqi, an advertiser pays the minimum amount needed to maintain her position. If there are fewer bidders than slots, the last bidder pays a reserve price or minimum bid r that is set by Google. If you define pi = price paid by the ith ranked bidder, then for i < M ipiqi = bi+1 qi+1. Rearranging this equation, the price paid by the ith ranked bidder is:

4 

If M < S the last advertiser pays the minimum bid, whereas if M = S, the bid for the last advertiser is calculated from the bid of the first omitted advertiser.

NOTE

In Equation 4 the advertiser's Cost per Click does depend on the bid and quality score of the next lower ranked advertiser.

This type of auction is known as a Generalized Second Price Auction (GSP). To see why, note that if all ads have the same quality score, the GSP reduces to each bidder paying the amount bid by the next highest bidder.

From Google's point of view two problems arise with an auction in which all bidders pay the amount they bid for each click (called a first price auction):

·        There is no incentive for a bid to equal the advertiser's view of the true value of a click. For example, if Bidder 1 believes a click is worth $1.00 and Bidder 2 believes a click is worth $0.80, and the bidders know this, then Bidder 1 could bid $0.81 and Bidder 2 could bid the minimum acceptable amount. Then Bidder 1 wins the first position by not bidding her true value.

·        If bids can be adjusted frequently, then in a first price auction over time, bidding will be unstable. For example, Suppose Bidder 1 values a click at $0.80 and Bidder 2 values a click at $1.00. Suppose Bidder 1 bids her true value of $0.80. Then Bidder 2 would bid $0.81 and win the top spot. Then the first bidder would lower his bid to the reservation bid (say $0.05.) This keeps Bidder 1 in the second spot. Now Bidder 2 will lower his bid to $0.06 and maintain the spot. This unstable behavior would not be satisfactory to Google!

In practice the GSP auction resolves these problems for Google, although the resolution of these problems is accomplished at the expense of the bidders.

Auction Examples

To illustrate how AdWords auctions work, look at the two concrete examples of auctions shown in Figure 36.3. In the first example, there are three bidders for three slots. Each bidder placed a $4 bid, so the bidders will be ranked by their Quality scores. Therefore, Bidder 1 gets the first slot, Bidder 2 gets the second slot, and Bidder 3 gets the third slot. Bidder 1 will pay the minimum amount needed to maintain her position. Because the second bidder has quality * bid = 24 and Bidder 1 has a Quality of 8, a bid by Bidder 1 of 24 / 8 = $3 would maintain her position and she is charged $3.00. Similarly for Bidder 2 to maintain her position, she needs 6 * Bidder 2 cost = $12, so Bidder 2 is charged $2.00. Bidder 3 pays the minimum bid.

Figure 36-3: AdWords auction examples

In the second example four bidders compete for three slots. Bidder 3 placed the third highest bid, but Bidder 3's high-quality score enables her to win the top spot. Because the second place in the auction goes to Bidder 2 (with bid * quality = 9), Bidder 3 must pay 9 / 6 = $1.50 per click. In a similar fashion Bidder 2 pays 8 / 3 = $2.67 per click, and Bidder 4 pays 4 / 8 = $0.50 per click. Bidder 1 bid the most per click, but her low-quality score causes her to rank fourth in the auction. Because there are only three slots, Bidder 1 is shut out and wins no ad placement.

Using Bid Simulator to Optimize Your Bid

After you sign up for Google AdWords you gain access to the Bid Simulator feature. If you input a keyword and a given bid per click, the Bid Simulator feature can estimate for bids near (and lower than) your bid how many clicks you will receive for lower bids. This can help the advertiser determine a profit maximizing bid. To illustrate the idea, suppose you sell digital cameras and earn a profit of $100 per camera sale and you expect a 5-percent conversion rate on click-throughs. This implies that you can expect an average profit of 0.05 * (100) = $5 per click, and the most you should ever bid is $5. Suppose you input this information into Google's Bid Simulator tool and receive the information shown in Figure 36.4. (See worksheet BIDSIM of workbook Costperclickoptimization.xlsx; the data comes from Hal Varian's excellent video on AdWords atwww.youtube.com/watch?v=jRx7AMb6rZ0.)

Figure 36-4: Bid Simulator example

You can find, for example, that reducing your max bid from $5 to $4.50 is estimated to cost you 18 clicks, and the expected cost per click would be $3.13. For each bid you can compute the expected profit as:

·        For example, for a $4 bid, expect to earn: ($100) * (0.05) * 154 – ($2.64) * 154 = $363.44.

From Figure 36.4, you can find that for the listed bids, a bid of $4 would maximize your expected profit.

Summary

In this chapter you learned the following:

·        Profit per click may be computed by:

1 

·        You will break even on each click if:

2 

or

3 

·        Essentially, Google AdWords ranks ads based on (Amount Bid) * (Quality Score). A bidder pays just enough to retain his current ranking.

·        Using Google's Bid Simulator tool an advertiser can estimate a profit maximizing bid.

Exercises

1. Suppose you are considering purchasing placement on search results for the keywords swim goggles. You earn a $2 profit per sale of swim goggles. Give several combinations of conversion rates and bid per clicks that enable you to break even.

2. Describe the results of an auction with three slots for the following data. Assume a minimum bid of $0.10.

3. Describe the results of an auction with three slots for the following data. Assume a minimum bid of $0.10.