Efficient Frontier Insights

Managing Events-driven Campaigns

How do you decide what to spend if your business is heavily driven by events? For instance, you may be a concert ticket advertiser planning for a major gig, or perhaps a sporting advertiser planning for the upcoming 6 Nations Championship (a major rugby tournament in Europe). Do you simply try to bid to as high a position as you can on your top keywords and hope that the impressions you gain will bring you sufficient conversions to make the spend worthwhile? But is it really worthwhile? How far will you be from your cost per order (CPO) goal? Will the average order value (AOV) be the same, lower or higher?

The challenge of events-driven campaigns

Events-driven advertising campaigns are challenging because:

(1) you know something is going to happen, but

(2) you are unsure the scale of the impact, therefore

(3) you are uncertain how to prepare for it, hence you implement the “spend and wait and see” strategy.

As a result, you may find your CPO fluctuates wildly indicating your budget had not been allocated in the most efficient manner (assuming constant AOV). The chart below illustrates such a situation. The red circles highlight the instances where events took place and the advertiser bid up to such a level that CPO was much higher than what it was worth. The green circles, on the other hand, highlight the occasions where no event took place and the advertiser pushed down the bids. As a result, the CPO was too low leaving demand to be exploited.




The myth of chasing impressions

The strategy many advertisers employ when managing event-based campaigns may simply be chasing the impression volume. When they can foresee there will be a rise in search interest they plan to spend more, by raising bids, to gain more clicks. However, we will argue that the value of a click is the key factor at play here instead. If the upcoming event only drives impressions but the conversion rate and the average order value (AOV) stay the same, then the revenue per click would remain unchanged and there is no value in raising spend. It is only when an event drives a change in conversion rate and/or AOV, where it is worthwhile to adjust bids to spend more.

So if we know the value of a click is the key, then how do we actually decide the budget on a given event-affected day? The answer is complimenting simulation with analogy adjustments.

As Will Senior, an Account Manager at Efficient Frontier, commented “to ensure you always pay the correct amount for each click, you need to accurately predict how upcoming events will impact click value. The easiest and most reliable method to predict a change in the click value is to use historical data from identical or lookalike events. The click value change could come in many forms, from a conversion rate shift to customer value adjustment.”

Best practice = simulation + analogous data

At Efficient Frontier we are proud of our sophisticated modelling and forecasting algorithms. We build click and revenue models on individual keywords and bid on them at the most efficient levels given the client’s business goal. Our platform’s simulations allow the advertiser to establish the spend level required to achieve the specific CPO or ROI goal based on historical data. When there are forthcoming events, whether a one-day sports event or the week leading up to a major music festival, we will also consider the previous similar (analogous) events and their conversion rates/AOV in relation to the average conversion rate/AOV. The latter ratio will provide us a conversion rate adjustment factor to build into the simulation that is automatically generated by our technology.

This could be illustrated by the graph below. Let us consider only CPO for now and assume constant AOV. The blue line illustrates what the simulation tells us for an average day about the relationship between spend and CPO of a campaign. When an event is to occur and similar past events suggest that there is a higher conversion rate than normal, the curve moves to the right (green line) so that we can increase the spend from just under $3,000 a day to over $4,500 and gain more conversions for the same CPO of $120.

 

This approach of combining simulation with analogous data gives the advertiser control in two ways for managing event-based campaigns.

First, the budget can be determined in advance and the advertiser simply sets the budget for each day of the coming week or month in the bid management platform rather than on the day.

Second, this reduces the daily variation in CPO from the target CPO. For our clients, on average, once this structured approach is implemented (from an intuition-based approach), the variation of CPO from target is reduced by 50%. The more similar historical events there are, the more you can refine your estimate of the adjustment factor used to build into the simulation.

Even if your current bid management platform does not readily provide simulations, using past similar events as reference to your forecasts for the upcoming event would still prepare you better than intuition alone does, as evidenced by academic research (e.g. Green and Armstrong (2007)).

Applying a more structured approach, your “spend and wait and see” strategy can be transformed into “analyze and anticipate”. You’re in control.

Dr Wing Lee, Business Analyst

The value of a Super Bowl ad: A performance marketer’s perspective

Super Bowl ads are a brand marketer’s dream. Last year’s Super Bowl set a new record for the most-watched program in television history, bringing in a whopping 106.5 million viewers across the U.S. Major brands including Google, Coke, GoDaddy, and several automotive and retail companies bought millions of dollars in advertising spots during the game to capitalize on this massive market.  

Yet the value of TV ads is a subject of much debate among performance marketers. While many marketers believe that TV ads improve marketing performance in the long run, others believe that they are an expensive way to build awareness. At Efficient Frontier, we have attempted to answer this question for many advertisers, and since the Super Bowl is only four days away, we thought it would be an opportune time to present some of our research on the connection between TV advertising and Search Engine Marketing (SEM).

First, does TV advertising help build online traffic? The answer is, undeniably, yes.

The chart above shows the brand impression volume for eight weeks during a TV flight. One can clearly see that the impression trend is very closely tied to the trend in TV spend. A closer inspection reveals that the connection between the two is more nuanced. A well established brand with a high degree of awareness will see a smaller jump in online traffic during a TV flight than a small brand. Furthermore, a big TV ad campaign will lead to a bigger impression spike than a small one. As a rule of thumb, a TV flight causes between a 60 and 80% jump in searches for the brand terms and between 40 and 60% for generic terms.

Super Bowl ads provide a huge spike in awareness. Does the spike translate to sustained online traffic? Some of our research attempts to answer this question. Using vector-auto-regressive models, we were able to estimate that a big spike in TV spend increases SEM traffic for four to six weeks.

 

For all Super Bowl advertisers, we advise the following to maximize your ads’ effectiveness online:

1. Anticipate a spike in traffic volume on your site both from paid and organic searches. Ensure that both your brand and non-branded terms are at high positions to capture all the traffic.

2. While you may be worried about spending beyond your allocated SEM budget on Sunday and perhaps for the coming days, you will also see much better than average conversion rates. This is good as it means lower acquisition costs.

3. Ensure that your Facebook campaigns are aligned with your TV efforts and link your Facebook pages to the video of your Super Bowl ads. Measure the effectiveness of your Facebook campaign using upper funnel metrics such as average time spent and engagement, and less to lower funnel metrics such as conversions.

4. Measure the effectiveness of your Super Bowl ad on your online marketing efforts right after the ad runs, then again after a week, a month and beyond. The results may pleasantly surprise you!

While our analysis of TV and SEM data has provided us with valuable insights, it also opens the door to tantalizing media mix questions. What is the right mix of online and offline budget to maximize a marketing campaign’s performance? How should the online budget be spent when TV flights are known in advance? What are the long-term branding effects of a TV campaign and how can they be measured online? What is the effect of TV advertising on Facebook, which is an upper funnel marketing vehicle? We do not have all the answers yet, but we are working on finding them and we’re very excited about the possibilities.

Dr. Siddharth Shah
Director, Business Analytics

Social Search: Facebook + Bing’s partnership and its implications

The announcement of Bing and Facebook joining forces to present socially influenced search is an important yet inevitable evolution of the search industry.  The total time spent on Facebook now exceeds Google and Yahoo!. Further, a Nielsen report stated that the average user spends 23% of her Internet time on social networks, up 77% from a year ago. With such rapid adoption of social networks, it is inevitable that it will (and probably, already has) influenced the way people search and buy products.  If one does the analysis on these changing trends, the following patterns should emerge increasingly on Search.

Virality: Showing friends who like a product will bring the viral nature of social network into the purchasing funnel. Products that would take time to become popular could become popular instantly much like YouTube hits. On the flip side, initial adopters having bad product experiences could make a movie instantly unpopular. Companies will have to be especially careful when launching new products.

Changes in buying behavior:  Collaborative filtering is a new research area that aims to predict one’s preferences based on multi-dimensional user data. The highly publicized “Netflix prize” was a contest for the best collaborative filtering algorithm to predict user ratings for films, based on prior ratings.  Recently, newer algorithms have emerged that use Social Network data (Social Network Collaborative Filtering or SNCF) do a great job in predicting preferences. These have the added advantage of not being computationally intensive and hence can be used for a high volume of data.  From a business perspective we can now confidently assert what marketers always knew – a consumer’s preference for products and services are well influenced by our friends and family. Hence, with the Bing and Facebook integration, we should expect consumers to buy products and services based on what their friends like. Further, the research phase of the purchasing cycle should be shorter. Net-Net: Higher conversion rates for advertisers on Bing and shorter buying cycles.

Greater integration of Facebook and Search: When Bing ads include Facebook data one should expect a greater degree of overlap between Facebook and Bing in the purchasing funnel. All this makes it more important than ever for advertisers to integrate tracking across both channels to maximize their performance.  Needless to say, advertisers have to consider Facebook as part of their overall marketing strategy.

Finally, what about Google ? While Google has made recent forays into the social sphere with products such as Buzz and Orkut, it has been reported that not having Facebook data could be a cause of worry to the company. However, I learned to never underestimate Google a long time ago.

We will keep our eyes open on how these changes are affecting both consumers and advertisers and will report on anything noteworthy as soon as they occur.

 

Dr Siddharth Shah
Director, Business Analytics

Dayparting or Splitting Hairs?

In this month's Search Engine Land "Analyze This" column, I discuss whether dayparting or hour or day bidding is worth the time and effort. For my full analysis, please read http://searchengineland.com/dayparting-profitable-tactic-or-waste-of-time-30727 but here's the bottom line.

For most businesses, dayparting will at best give you a small incremental gain. It is best to look at all of your data, cost and revenue on a per click as well as aggregate basis before making any dayparting decisions.

Dr. Siddharth Shah

The Future of Biddable Media

Every day, Efficient Frontier's technology executes millions of bid changes on behalf of our clients. Dozens of servers contact dozens of other servers at Google, Yahoo, Microsoft and Baidu to adapt campaigns depending on changing market conditions and business goals. A constantly evolving set of APIs makes the communications efficient and seamless.

But for a moment, just imagine a world where this doesn't happen. Where there are no APIs, no automated setting of bids, and not even an interface to set those bids by hand. In fact, every bid needs to negotiated one-by-one, over the phone or in person. Sounds awful, doesn't it? Welcome to the world of online display advertising.
 
The banner is the grandfather of online advertising. It has spawned different sizes, formats, targeting over the years, but like many grandfathers, it still has an old-fashioned way of behaving: trading.

SEM has a dynamic ecosystem of biddable marketplaces that advertisers can value and buy media. Third party technology, such as Efficient Frontier, helps those advertisers evaluate market data to determine when and where to show their ads, and at what price. The time has come for this remarkable system of decision-making and trading to come to the display market.

For too long, display advertising has been hamstrung by an antiquated method of trading that delays decisions, creates inefficiencies and suppresses investment by advertisers. Ad exchanges, such as Doubleclick and Rightmedia, are creating marketplaces that will provide an efficient and dynamic environment for media owners and advertisers to do business.

All the trading benefits that we currently enjoy in SEM are being brought to display, for advertisers and publishers alike. At Efficient Frontier, we are rapidly immersing ourselves into this new world. As biddable market places become a bigger feature of the display world, we will be helping advertisers track, optimise and forecast - just as we do today in search. Performance marketing is about to get a lot more exciting.

Jonathan Beeston
Client Services Director, Europe

Follow the data...not the position

When we take onboard new clients, we find that some clients have a predetermined idea on which position on the Search Engine is the best. Some advertisers believe position 1-3 on Google is a must. Putting aside any branding or defensive PR reason for being placed at number one, one of the first things we try to educate new clients is to let the data dictate your Search Engine Results Page (SERP) position.  The following case study shows the importance of letting the data work its magic and give you a better ROI.
 
Recently, a German travel client used more than 3900 keyword constraints in order to influence its overall performance. A keyword constraint is the maximum/minimum CPC or maximum/minimum position rule for each individual keywords.  Soon, 90% of the daily budget was used to fulfill these constraint targets and only 10% of the budget was actually used to optimize keywords and increase real performance.
 

By taking away the majority of constraints, the complete daily budget was available to have the keyword market develop and to use the data generated from Efficient Frontier's portfolio optimization technology. Monitoring the performance very intensely, the following development was encountered.

When looking at Efficient Frontier's simulations this step became even more obvious. When the constraints were taken away, uplifts can be expected immediately. In this case, we saw an ROI increase of 96% (day to day) and 45% (week to week) just by taking away keyword limitations and letting the invisible hand of the optimizer take over. Since then, the free market environment has allowed the performance to improve continuously.

So if ROI is your key objective,  resist the need to seek the top SERPs positions and liberate your keywords from their constraints to improve performance and ROI.

Julian Stank
SMS Hamburg Germany

A New Challenge with a Familiar Feel

The Sunday NYTimes Business section featured an article by Stephanie Clifford on the evolution of display marketing called "Put Ad on Web. Count Clicks. Revise." It's worth a read. Clifford does a good job of covering how the ad exchanges brought an auction pricing model to display. Importantly, she covers how access to new targeting data from 3rd party sources like Blue Kai significantly alters the value of display inventory.

Until recently, targeting data has been controlled by large publishers and ad networks to drive higher CPMs and margins. As Clifford rightly notes, "this combination — real-time [targeting] data and ad exchanges — has monetized what was once considered throwaway space online." The key point is that access to both real-time targeting data and ad exchanges have shifted price control to the marketer and away from the publishers and networks. The result is that while demand builds in the exchange and data auctions, early-to-move marketers stand to benefit from the low prices and good volume available in-part because of the lack of competition. The early mover opportunity in the exchanges is reminiscent of the early days of search marketing with Overture and Google.

One point omitted from the NY Times story was the link between search marketing and the developments in exchanges. Search marketers and their agency and technology partners have been using data to drive bid and copy decisions for many years. In fact, the data-driven, Wall Street-like approach to marketing now being applied to display is core to any successful search engine marketing effort. Efficient Frontier's founder and CTO, Dr. Anil Kamath, brought the quantitative modeling techniques he built for programmatic trading in the hedge fund industry to search marketing in 2002. To those companies like ours immersed in search engine marketing, ad and data exchanges are an exciting development that fit well with our competencies and technology. This "new era" of display advertising is very search-like at its core.

We're actively working on our display and exchange management capabilities. Early testing has shown clearly the opportunity that the exchanges present in delivering efficient results at significant scale to our performance-oriented clients. The opportunity is still evolving as the exchanges develop and improve their APIs, a necessary step for us to automate optimization. We’re thrilled to be a primary participant as the new era of display unfolds, presenting us with a new challenge with a familiar feel.

by Justin Merickel
Vice President of Marketing and New Product Development

It isn't obvious

Consider the following keywords
You would like to bid each keyword so that you maximize orders subject to the constraint that the overall Cost Per Order (CPO) is less than $20. A lot of us would simply look at each keyword and bid them to a CPO at or higher than the average CPO. So we bid keyword 1 to position 1 (CPO of $20) and keyword 2 to position 2 (CPO of $20). In this scenario we would get 15 orders from keyword 1, 10 orders from keyword 2 for a total of 25 orders at a cost of $500 ($300+$200). The overall CPO is $500/25=$20. So it's good.

But what if we did the exact opposite? Here we would get 10 orders from keyword 1 at a CPO of $10 and 20 orders from keyword 2 at a CPO of $25 for a total of 30 orders at a cost of $600 ($100+$500). What is the CPO now? Its $600/30=$20. We got more orders for the same CPO despite exceeding the CPO target for keyword 2! Welcome to the portfolio theory of SEM. Like we said... It isn't obvious.

Learn more about the portfolio theory of SEM and several other cutting edge SEM concepts in our latest paper: Algorithms and Optimization

Visualization of Data through Motion Charts

One of the key advantages of SEM is that data can be measured at several points in the sales funnel. For example: Impressions measure the degree of user intent, Click Through Rates (CTR) measure the effectiveness of your ads in capturing that intent, Cost Per Click (CPC) measures the price you are paying to capture intent and finally the ROI measures the rate of return on investment in this marketplace. While this multi-dimensional measurability of data is a key asset to the search advertiser, the complex interactions of the metrics can be daunting to analyze without the right mathematical and analytical tools. Visualization techniques such as treemaps, which was discussed in a recent post, can help capture several dimensions of data successfully and make the data meaningful. In this post, we present another very powerful tool called Motion Charts that can capture 5 dimensions of data in an extremely effective manner. Whats more these charts are fun ! Before, delving into these charts, a bit of history. Motion Charts were first presented to a large audience by Dr Hans Rosling in a famous TED talk where he discussed global health in an extremely engaging manner. The tool he used was called the Trendalyzer developed by a non-profit Gapminder.org . Google later acquired the Trendalyzer and developed this technology into the Motion Charts. It is now part of the Google Docs suite of visualization tools . As an example, we will analyze the performance of a hypothetical SEM campaign with this chart.

The two bubbles represent 2 portfolios of campaigns managed by Efficient Frontier. One bubble represents Google campaign and other represents the Yahoo campaign. If one moves her mouse over the right most bubble, 4 dimensions readily become apparent. The horizontal axis captures weekly spend, the vertical axis captures weekly account signups, the size of the bubble is proportional to average conversion rate (free registration to paid subscription rate) for the Google portfolio and the color of the bubble is proportional to the CPA of this portfolio (note the color specturm on the right side of the chart). The paramters to be tracked can be interactively changed by selecting the drop downs on the axes and on the right side of the chart. Finally, note the date on the right side of the chart. This is the fifth dimension. So, to interpret: on Week starting May 1 the Google portfolio generated 800 registrations at $1000 spend with a conversion rate of 9.4% at a Cost Per Registration (CPR) of $1.25. Moving the slider, will change the time and will display the locations of the 3 portfolios on that week. Hitting the play button will play the "movie" of your quarterly performance. Trails are a very powerful feature in the motion chart. To enable trails, check the trails box on the screen (it defaults as already checked), change the size parameter of the bubbles (in the drop down) to "same size", and click on the Google portfolio bubble again. You would now see a name tag pop up next to it. On pressing play again, the Google portfolio is "tracked" over the quarter and one can clearly see the change in performance over time. In this example, the Google portfolio is seen to be 25% more efficient from week 2 i.e. the weekly account signups have increased by about 25% at close to the spend on week 1. Also note that since week 2, the portfolio has clustered at its "sweet spot" i.e. while there are week to week fluctations in performance, i is always better than the performance on week 1. An analysis of several accounts managed by EF reveals a similar pattern i.e. while performance fluctuates on a weekly basis, the portfolios cluster around a region of higher efficiency due to optimization.

Managing Risk in SEM: Using Histograms

Today Siddharth Shah, PhD and Senior Business Analyst at Efficient Frontier will speak on the topic of "Managing Risk in SEM" at SMX Advanced in Seattle. This post will introduce the concept of managing risk in SEM that he will present, and show how you can use histograms to help you assess and manage risk.

Firstly, what does risk mean in relation to your PPC campaign? You have hundreds, thousands, tens of thousands, or in the case of some Efficient Frontier clients, millions of keywords on multiple search engines, with multiple ad copies, landing campaigns, and possibly targeted to various regions. Every day you are expecting a certain amount of clicks to drive traffic and conversions to your site. There can be high levels of variability in impression volume, user behavior, and competitive activity, but the success of your business depends on this traffic. Therefore gaining some level of predictability to what can be a highly volatile marketplace is imperative.

A histogram is a statistical visualization tool that shows the proportion of cases that fall into certain ranges. Unlike a bar chart, it is the area of the bar that defines the values, rather than just the vertical scale. To make a histogram in Microsoft Excel, you need to download the Analysis ToolPak. You then have to define the bins, or ranges, that your data groupings should fall into. For example, the histogram below shows that for this particular head keyword, there were 27 days out of 80 that the keyword received between 60 and 69 conversions. The bins used here were multiples of 10, but as you will see on the Wikipedia entry, the bins do note have to be equal values.

So how can a histogram help you manage risk? Siddharth created histograms of many keywords by number of days that the keywords received certain amounts of revenue (or conversions), and found that head term histograms, like the one above, followed a certain pattern that resembles the normal distribution. In contrast, tail term keywords followed no visible pattern of revenue generation, suggesting that head and tail keywords should be treated differently in your portfolios.

Why is the normal distribution important? The normal distribution, or Gaussian distribution, or bell curve, as some call it, has many important properties, the most important of which in this case is its predictability. The blue line on top of the histogram above represents the normal distribution for the same values in the histogram. The central limit theorem, a fundamental theory of probability, states that a large number of random variables will follow a normal distribution. Thus we can predict, with a reasonable amount of certainty, the range in amount of revenue or number of conversions we should expect from a head keyword.

This is where the mean and the standard deviation (SD) come in, as have been shown on the chart. The mean, or average, number of conversions that occurred from this keyword on any of the 80 days analyzed was 60.55.  But if you expect 61 conversions every day from this keyword, you will be sorely disappointed, as the histogram shows that there were only 34 days (27/80) in which there were between 60 and 69 conversions. Even if you expect 60 or greater conversions, your goal will only be satisfied 79% of the time (63/80), which leaves 17 days that you will be wondering what went wrong.

In a normal distribution, a standard deviation of 1 standard unit above or below the average indicates 68% probability, and a deviation of 2 standard units indicates 95% probability. The standard deviation (a convenient formula in Excel: STDEV) for this group of data is 13.21. Thus there is a 68% chance that on any given day this keyword will bring in between 47 and 74 conversions (60.55 +/- 13.21). There is a 95% chance that the keyword will bring in between 34 and 87 conversions (60.55 +/- 2*13.21).

You can use this method to define a range of expected performance for your head keywords. If you are 95% confident that the keyword will bring in a specified range of conversions, you only have to worry about the days that performance fell outside of that range. Since you don't have to worry about the days where performance exceeded expectations (unless you want to investigate if there is a positive change in the marketplace), you now only need to worry about the days where performance did not meet expectations. In this case, there was only one day where the keyword received fewer than 34 conversions, which is far less taxing than worrying about performance not meeting expectations on say 17 days if you were holding this keyword to a 60 conversions per day goal.

This kind of management of the head keywords allows you to approach the tail with a different mindset. Siddharth did another analysis and found that in one month there were 16,105 keywords in a campaign that generated revenue, while in the following month there were 17,238 revenue generating keywords in the same campaign. Closer analysis showed that there were only 2,937 keywords, or 15% of total, that generated revenue in both months. Tail terms should closely monitored and actively bid on as their behavior is very dynamic and sparse data does not allow the same level of predictability found with head terms. This unpredictable behavior also means that tail terms should not be held to a daily revenue target if you want to extract the most value out of the tail.

Efficient Frontier's algorithmic optimization technology reduces risk by proactively monitoring and acting upon this dynamic keyword behavior in both the head and the tail, in a way that rules-based systems cannot. By modeling every keyword every day at every position for clicks, revenue and CPC, risk is reduced as the bids continually reflect activity in the dynamically changing marketplace.