Beating the Ad-Blocker
Mir Nasir Ali Khan
21st July 2020
Company Profile�Background Information
GoYaFace, Inc. (GYF) is the company for which we are trying to solve a specific issue. It was incorporated in 2000 and today is a large digital search engine, email/messaging, and internet content company. GYF focuses on three core business units: GYF Search, GYF Mail & Chat, and GYF Digital Media. GYF Search is an internet search platform similar to those offered by Google and Microsoft (Bing). GYF Mail & Chat offers email and chat services similar to Gmail and Gchat. GYF Digital Media includes several subject-matter specific content destinations, including GYF News, GYF Sports, GYF Business, GYF Style, and GYF Technology. These sites provide original content and interactive features like fantasy football for example, GYF Sports is comparable to Yahoo Sports. GYF has users all over the world and its three core business are offered free of charge to end users.
A major portion of GYF’s revenue comes from selling advertising to other companies who place ads on GYF’s digital services. Originally, this ad-selling business focused on the use of GYF’s digital services and properties at traditional desktops, but recently it has significantly transitioned to mobile advertising as smartphones and other handheld wireless devices have proliferated. Last year, mobile advertising sales accounted for 70% of GYF’s total digital advertising revenue.
GYF’s advertising sales program for its search engine and mail/messaging platforms resembles Google’s “Adwords” program (more information here https://www.google.com/adwords/), while its advertising program for its Digital Media content such as news, sports, etc. resembles YouTube’s advertising program (more information here https://www.youtube.com/yt/advertise/index.html). GYF users see advertisements as part of their experience on GYF sites and services across devices, as well as some third-party apps and affiliated sites.
GYF competes with Yahoo, Google, Microsoft, and Facebook. You should assume that ad-blocking poses similar issues for GYF as it does for these companies.
Subject to the above, GYF’s business model most closely resembles Yahoo’s business model as it existed in 2015 and early 2016. If you believe you need to know something specific about GYF’s business and it is not explicit or implicit in the above, feel free to assume that that aspect of GYF’s business resembles the corollary at Yahoo.
Problem Statement�The Problem Ad blockers present to GYF
Ad-Blocking has been growing at an exponential rate over the years. Ad-Blockers are inexpensive, simple to install and take a minute or two to magically vanish all the advertisements in a browser. Hence Ad-Blockers have become a nightmare for many companies including giants like Google, Microsoft and Yahoo. Whereas for smaller media creating companies, Ad-Blockers are no less than venom.
Today, almost 30% of web users around the world are using ad-blockers. In 2019, according to Adobe, ad-blocking was responsible for an estimated loss of around $22 billion in online advertising. From a study done by Prof. Kirthi Kalyanam, It was also found that, some of the ad-buyers were facing a loss around 1.8% at their physical stores because many the people see ads online and purchase at a physical store. The ad-blockers also had an impact on brick and mortar sales. This along with many existing issues might have made a lot of ad-buyers back off from making a deal with GYF. Google and Yahoo also went on to accuse Ad-Blocking software Shine of destroying the relationships between advertisers and consumers. According to a Benjamin Faes of Google, “Ad-Blocking is a blunt solution that punishes users and good advertisers. It diminishes the experience of advertising and it is an issue for users themselves”. According to Nick Hugh of Yahoo, “this is a complete destruction of value exchange and ecosystem”. More and more publishers cannot afford to give their content for free and the companies lose a lot of income because neither the user is paying nor viewing the ads for the services they are using.
But for some optimists like Roy Carthy (CMO of Shine technologies), Ad-Blocking is a stellar opportunity to reset the relationship with the customers and new rules of engagement need to come out. He also claims that consumers are being abused by advertising technology. In addition to annoying advertisements our data, battery life, time and privacy are also taken by a blank check with no resources. Ad-Blockers give an opportunity for the customers to customize their own version of the program.
Web is a decentralized platform and there is no such rule to ban Ad-Blockers which have not been designed with malicious intent and do not compromise a user’s privacy or security. Due to this, popular softwares like Ad-Block Plus are unwilling to allow individual sites to block people using their plugin. According to them, the user experience has to come first.
Apple’s support to Ad-Blocking in browsers at the iOS 9 launch has been another surprising event. This has created a havoc by commercialized the Ad-Blockers to a greater extent. Support was supposedly announced to improve the user experience but many accuse Apple to have done this to thwart its competitors and drive users and advertisers to its own apps like Apple News.
Since the advent of this issue, there have been many half-baked solutions like whitelisting , blocking browsers, native advertisements, etc. But none of them have been comprehensive to satisfying both the parties i.e. the users and the advertisers.
From the above it is quite evident that the Ad-Blocking debate is not likely to end soon and advertisers are gloomy about it. A whole golden age of advertising can evolve if we are somehow able to find a solution for this problem. This is a good opportunity for model innovation and to seize the moment by getting it right.
Problem Statement �Application Exercise 1 - Research Methods and Tools
Exploratory Research:
1) Why are people using Ad-Blockers? We can probably create a survey and randomly assign it to a few thousand of our users who have Ad-Blockers enabled in their browsers and know the reason that made them block advertisements. Collecting feedback from time to time about the user experience and improvement opportunities from the Non-Ad-Blocker customers could also prove helpful to predict whether the customer will block advertisements in the future. Since we do not have much control over the users who have already enabled Ad-Blockers, we can make sure that we provide a good experience to the users who haven’t done so to prevent any further future damage. We can use Survey Monkey as our service of choice to do surveys, analysis and convey the results to us. These results would be very vital to form a strategy.
Descriptive Research:
2) What is and would be the rate of increase of Ad-Blocker users for GYF? It would be nice to track the GYF users who use Ad-Blockers yearly and plot them in a graph. Then, we will use a regression model to predict what would be the number of Ad-Block users in the coming year. By doing this, we can also get an idea of whether the strategy that we will be employing in future is causing some positive impact or not. Regression models will also make us ready to face the future circumstances by altering our strategy.
3) What are the kind of pages users are more satisfied with? Sometimes from the 1st question above, people might not be satisfied with advertisement placement or load times of ads in websites or search etc. For this what we can do is set up A/B testing. There are a lot of companies that can do this for us but let us go with Optimizely as a tool of choice.
Causal Research:
4) Is there a causal relationship between ad-blocker users, ad-buyers and GYF’s ad revenue? Well, generally going with the previously done researches, it is known that as the ad-blocker users increase, the amount of ad-buyers decrease which may result in reduction of ad-prices by the company ultimately reducing the revenue. It is quiet clear that there is a causation. But, how much causation is there is the question? Amount of causation can be calculated easily by doing a regression analysis between all the parameters i.e. ad-blocker users, ad-buyers and GYF’s ad-revenue. Coming up with an equation/result of relation between these parameters can hugely help us in setting targets.
Strategy�Description of proposed strategy
After keen assessment of the problem statement, it is identified that the situation we are working with is highly unpredictable and not easily shapeable. The situation is highly unpredictable because it is difficult to predict the effect of different ad placements, ad types, number of ads shown, load times etc. for different consumers. Due to this shaping also becomes an uphill task. The approach to strategy that perfectly fits the situation is what we call the ‘Adaptive’ approach. In this approach, GYF’s advertisements should be adapting to the consumer’s liking. Here a process has been laid out for experimentation that includes measuring effectiveness, speed and return on investment. According to PageFair, at the end of 2019 there was a slight dip from the peak in Ad-Blockers used by desktop users but the case was not the same with mobile. The mobile Ad-Blocking users continued to grow. From this it is evident that GYF and the whole industry has been getting the desktop space right. Hence our major focus should be on mobile. Also there should be a special emphasis on flattening the curve by targeting the potential Ad-Blocker customers. These are the customers to whom we need to prove that we can improve. It is very difficult to prove to the customers who have already installed ad-blockers because we don’t get a chance. However we can try convincing ad-blocking users using a pop up but with full access to the content/service. A lot of companies have tried to get a dip in the curve by restricting the Ad-Blocking users. History is a proof that restricting the customer harms the company itself. Users have a lot of alternatives. Whitelisting our websites in collaboration with major ad-blockers is also a great solution to the above. It at least gives us some opportunity to escape the further loss of revenue. However, it is not a long term solution. If the user experience is bad, today or tomorrow user is definitely going to voluntarily wipe off the ads.
Now let us look at our step by step execution strategy:
Step 1: Identify the reason our consumers are using Ad-Blockers. Also, take a time to time feedback of user experience from our non Ad-Blocking users. This can be done using surveys. Feedback from our non Ad-Blocking users helps us to predict probability of customers that will go for ad-blockers in the near future. It is the most vital step in this strategy. Non ad-block users are our target and they only can help us flatten the curve. We can try retaining or converting dissatisfied users of survey by sending them regret and promise to improve mail immediately. Some users might take a chance on the company and refrain from installing ad-blockers (or) disable ad-blockers.
Step 2: Well, there is a high chance that even after our step 1 customers continue using ad-blocker. There is no instant fix to what a user desires. We can come out with some sort of solution only after step 3. In the mean time we can use collapsible pop-ups without restricting customer’s access to site/service once or twice a day requesting them to disable ad-blockers by telling the importance of ads for the ecosystem and how we are working towards a solution. We need to gain faith of the user.
Step 3: Use A/B testing, feedback from the users and find out what is the ideal page that each individual consumer desires. And serve him/her with only the format relevant and liked by them. Also give them the option to blacklist a particular ad so that we can serve them better. And for better results it is advised to run all the 3 steps in a cyclic fashion continuously in this rapidly changing world to keep up with our ever-changing consumers.
Strategy�Application Exercise 2 - Hiring a Team Leader (Question)
Assume that, as the leader of the DATA Team at GYF, you are looking for a new Senior Associate Director for Digital Advertising Strategy. This person will be your second-in-command as you develop and implement your strategy for addressing the problem of ad-blocking. The hiring process has winnowed the field down to two final candidates: (i) Carrie Candidate and (ii) Peggy Prospect. Below are the metrics upon which the hiring decision will be made:
Using the methods introduced in the People Analytics course, please explain which person you would hire and why?
Strategy�Application Exercise 2 - Hiring a Team Leader (Answer)
Contrary to the belief of most of the organizations, hiring an efficient leader who can actually execute our strategy is very important. If not more, executing a strategy is as important as creating one. According to a popular saying, “Devil is in the details” and most organizations do not attribute this fact and consider execution a job of lower level people.
Let us first compare backgrounds of both the candidates. Carrie graduated from a comparatively prestigious school but I don’t think that there should be a lot of difference in the level of undergraduate education of both candidates. But what makes a difference is that Carrie holds an MBA from a very prestigious school. Carrie also has several years of experience at a large bank dealing with advertising on the ad-buying side but has little experience on the ad-selling side making her experience not very relevant for the current job. On the other hand Peggy has worked with a series of startups and in her most recent role in a larger digital media company, she is handling a similar problem GYF is facing and also been working under the same pressure as GYF team. Hence with her relevant work experience Peggy is a tough competition to Carrie. Comparing the GPA’s wouldn’t be a perfect apples to apples comparison but the GPA’s of both of them seem fine for the job.
Carrie presented great work samples and had an exceptional cognitive ability test score compared to Peggy. On the other hand Peggy was better at structured interview and job knowledge test while both had the same high score on the integrity test. On the personality test, it was found that both had their own unique and relevant personality traits that would help the company.
While references may not be the actual representation of a person, it should be noted that Peggy had a remark from one of her references that sometimes, she may be too focused on short term rather than long term. There were no such remarks for Carrie.
Well, it is too difficult of a task to select a candidate based on the above information. The competition is neck and neck. So, let us first score each of them on individual metrics and then bring in the tried and tested algorithm from People Analytics course to select the best candidate. In the algorithm, the weightage to each parameter is based on correlation to predicting the future performance. I have not assigned any weightage to the background/experience because it is very difficult to do an apples to apples comparison for this metric. Never the less, both of them have an equally impressive background/experience in my opinion.
Scoring of Carry Candidate and Peggy Prospect on various metrics | ||||
S.No | Metric | Carry candidate Score (100) | Peggy Prospect Score (100) | Weightage in % |
1 | Background/Experience | 80 | 80 | -- |
2 | Work Samples | 100 | 75 | 54% |
3 | Cognitive Ability Test | 90 | 75 | 51% |
4 | Structured Interview | 75 | 100 | 51% |
5 | Job Knowledge Test | 75 | 100 | 48% |
6 | Integrity Test | 100 | 100 | 41% |
7 | Personality Test | 100 | 100 | 31% |
8 | Reference Check | 100 | 75 | 26% |
| ||||
Calculation of Carrie Candidate’s Score:
100*0.54+90*0.51+75*0.51+75*0.48+100*0.41+100*0.31+100*0.26 = 272.15
The total algorithmic score of Carrie Candidate turned out to be 272.15
Calculation of Peggy Prospect’s Score:
75*0.54+75*0.51+100*0.51+100*0.48+100*0.41+100*0.31+75*0.26 = 269.25
The total algorithmic score of Peggy Prospect turned out to be 269.25
From above calculations, it is clear that the competition was fierce and Carrie Candidate came out on top. It seems that Carrie’s higher
quality undergraduate education and an additional qualification i.e. MBA along with her several years of experience in large firms had
instilled in her enough knowledge and skills to beat the very relevant experience of Peggy.
Final Selection:
Carrie Candidate.
Effects�Description of anticipated effects of the strategy
Effect on consumers and ad-buyers: There should be a lot of positive effects of this strategy for the consumers. It could result in performance improvement i.e. faster load times of search results and websites, improved user experience, minimal data usage, enhanced privacy & security and most importantly customer satisfaction with the ads that are delightful, informative, useful, meaningful and inspirational. This strategy is meant to create an impact through a beautiful and seamless web experience for users in an ecosystem where the customers directly express all their problems to the company before resorting to ad-blockers. As a result, the ad-buyers are also expected to increase.
Effect on revenue: If the strategy works as planned, we could see a huge improvement in customer satisfaction. And it is assumed that satisfied customers would never ever think about the ad-blockers. This could be beneficial in terms of revenue to all the parties i.e. ad sellers (GYF, Google, Yahoo, etc.) ,ad buyers (who want to display their ads on search results, websites etc.) as well as content creators. In this era of ad-blockers, online ad buyers don’t get their potential customer reach. If our strategy gets things under control, there could be huge shift in the market. Faith of the ad buyers would be regained and GYF would get more clients increasing its revenue. Ad- buyers get more visibility on websites increasing their conversions. And ads on websites means more income to the content creators like GYF Sports, GYF business etc. More income to the content creators could also mean even more content in the future and more content means more ads. In this way, just solving this one issue can bring the whole ecosystem back on track.
Effect on internal organization: Here more revenue to GYF doesn’t mean everything is sorted. If everything works out according to plan, we would have even more ad buyers. And to keep those ad buyers happy, we would need to keep the customers happy. Keeping the customers happy is a huge responsibility for the company. We had outsourced our work for this strategy. It might not be economical to always outsource. So, we might need to hire talent who could handle the company’s responsibilities. We might also need to make investments to track the hired talent. Also there are a bunch of other complications we need to look after when we increase our workforce.
Effects �Application Exercise 3 – Designing a Deterministic Optimization Model (Question)
Assume that an important part of executing your strategy in addressing ad-blocking will be efficiently allocating the limited resources of the data team. Your task in this Application Exercise is to build a resource allocation and optimization model. Your model should be based on the scenario below.
Optimization Model Scenario
GYF has a robust training program. Two of its training programs are known as the “Hard Skills Program” and the “Soft Skills Program.” Each of these trainings has two sub-programs: “External” (focusing on employee tasks mainly dealing with clients and customers outside GYF) and “Internal” (focusing on employee tasks mainly dealing with fellow GYF employees). As the leader of the data team, you can enroll your employees in these training programs, but you must decide how to best allocate your training budget of $65,000 among these four training options (Hard Skills/External, Hard Skills/Internal, Soft Skills/External, and Soft Skills/Internal). You decide to base this decision on productivity return rates. GYF’s management has calculated that the productivity return (i.e., the expected extra productivity in the next period, in the equivalent of U.S. dollars, that is achieved for each U.S. dollar spent on training, net of training cost) is proportional to the amount of money spent on training and can be expressed as follows:
This means that $10,000 spent on the Internal Hard Skills Program results in $2,000 worth of increased productivity in the next period. This return is cumulative, meaning that if you spent $10,000 on the Internal Hard Skills Program and $10,000 on the External Soft Skills Program, your total dollar return in productivity increases would be 10000*.2+10000*.4. You need to decide how much of your budget to spend on each type of training to maximize the total productivity return. But you can’t simply spend all your budget on the type of training with the highest return, because there are three requirements (in addition to staying within your budget) that management is requiring you to satisfy in allocating your training budget. They are:
The Hard Skills training program must achieve at least $20,000 in the total net productivity increase;
The Soft Skills training program must achieve at least $12,000 in the total net productivity increase; and
The Internal program should achieve at least 60% of the net productivity increase realized for the External program.
Assume all these figures are “deterministic” – known, and non-random.
Effects �Application Exercise 3 – Designing a Deterministic Optimization Model (Answer)
The explanation to the table presented above is in the next page.
The table in the blue i.e. net productivity increase table is the same as given in the exercise question. This tells us about the net productivity increase for a dollar spent on each type of training. For example, hard skills internal training productivity increase for $1 spent is $0.2
The table in the purple color i.e. the second table from the top, represents the spending amount for various types of trainings. And the values in these cells represent the decision variables. The values in the decision variable cells are selected by the solver, in excel, in order to maximize our objective function considering the constraints.
The table with grey color i.e. the third table from top as named represents the four constraints that were given to us. They are: 1) Total Spending Budget must be at most $65000 i.e. the sum of the values in the purple box should be less than $65000. Algebraically, hard skills internal training spending + hard skills external training spending + soft skills internal training spending + soft skills external training spending <= $65000 2) Total productivity increase from hard skills training must be at least $20000. Here, the productivity increase in hard skills is measured in excel by doing sum product of hard skills columns in the first two tables. Algebraically, (hard skills internal training spending * hard skill internal training productivity increase) + (hard skills external training spending * hard skills external training productivity increase) >= $20000 3) Total productivity increase from soft skills training must be at least $12000. Here, the productivity increase in soft skills is measured in excel by doing sum product of soft skills columns in the first two tables. Algebraically, (soft skills internal training spending * soft skill internal training productivity increase) + (soft skills external training spending * soft skills external training productivity increase) >= $12000 4) Total productivity increase from internal related training must be at least 60% of the total productivity increase from the external training. The text in white, in the constraints box represents 60% of the total productivity increase from the external training. The productivity increase from internal related training is calculated in excel by doing sum product of internal training rows in the first two tables. Similar procedure can be applied for calculation of productivity increase from external training. But note that, in the constraints table, only 60% of the productivity increase from external training is represented in order to facilitate the comparison with 100% of internal productivity increase. Algebraically, (hard skills internal training spending * hard skills internal training productivity increase) + (soft skills internal training spending * soft skills internal training productivity increase) >= 0.6 ((hard skills external training spending * hard skills external training productivity increase) + (soft skills external training spending * soft skills external training productivity increase))
The total net productivity increase is represented in red color cell in the last row. This can be calculated by doing the sum of total productivity increase from hard skills and total productivity increase from soft skills. Another way of calculating this is by doing the sum of total productivity increase from the internal training and the total productivity increase from external training. Algebraically, Total net productivity increase = total productivity increase from hard skills + total productivity increase from soft skills
Here we get a maximum total productivity increase of $42823.53. As you can see in the previous page, all the constraints have been satisfied. Solver (the excel add-in) has chosen for us, the optimal investment into each type of training to get the maximum possible net productivity increase considering all the constraints.
Measurement�Measurement of anticipated effects of the Strategy
Our strategy is meaningless if we cannot measure the outcomes we desire. We need to put some real hard numbers to support our results. Otherwise we would not be able to convince the company higher authorities, ad buyers, users or audience about our progress in reducing the ad-blocker users. This diminishing the sole purpose of the strategy. We need to prove that our strategy is working to attract all the parties i.e. management, ad buyers, creators and users. These are the people who make the ecosystem.
Measurement of effect on customers and ad-buyers:
We will be plotting a graph of yearly total Ad-Blocker users from history. Then we will use this data for regression i.e. to predict the amount of Ad-Blocking users we would see in the current year. Any reduction from the regression prediction of ad-blocking users would be termed as success . These are consumers saved from installing ad-blockers/ consumers who disabled ad-blockers. Increase/decrease in the ad-blocker users with respect to regression prediction year over year would be an effective measure which would tell whether our strategy is working or not. Similar process can be adopted to measure the effect on ad-buyers.
Measurement of effect on revenue:
With proper execution of the strategy, our revenue is expected to increase. More will be the number of new users refraining from ad-blockers and existing ad-blocking users uninstalling ad-blockers i.e. non ad-blocking users , more will be the revenue. We can calculate the return on investment of the strategy by comparing the cost incurred for the strategy with the amount gained from consumers saved of installing ad-blockers/ consumers who disabled ad blockers. Most of the companies have separate dashboards for ad-buyers and creators. From those dashboards, we can easily measure the increase/decrease in revenue for each creator and ad-buyer.
Measurement of effect on internal organization:
This is the area which will be majorly impacted due to the success of strategy. We would need to hire a team of quality people. Their performance might not be as intended initially. Well, one of the ways to measure the effect of new department/hires in the internal organization is to compare the key performances indicators of existing departments with their past performances over a period of 1 year after the hiring has been done. Make sure that the comparison with the past performance adheres apples to apples rule. If there is any drastic change in an existing department performance, find out whether there is any correlation or causation between the performance of new hires/department and that particular existing department of our company.
Conclusion�
Our causal business model consists of a series of steps that are expected to occur one after the other. For this, I had performed literature surveys to equip myself with more insightful information on this subject. In addition to literature survey, this causal model is also based of my intuition as a web user since the past 10 years and an ad-block user for 6 years who has successfully transitioned from being an ad-block user to a non ad-block user. In this model, I have also incorporate some small things that made me shift.
Our key drivers are intimately related to the business model and organizational strategy. Here let us discuss more about our causal business model to understand about the key drivers. Our main aim in this business model is to prevent people from installing ad-blockers so that we could show people some relevant ads thereby creating win-win situation for everyone associated with the ecosystem. So, according to me, the key driver of preventing people from using ad-blockers is good user experience. As, discussed earlier we can do surveys and A/B testing to find the optimum load times, ad types, number of ads per page and various other auxiliary drivers which would actually help us to abstain people from using ad-blockers. More is not always better. Same is the case with these auxiliary drivers. We do not need 1 nano second load time. We need to weigh these auxiliary drivers and find the optimum values for each of them to fit our purpose.
Here our intuition is that, if we (GYF) as a company focus on the strategy given and continuously improve our user experience by optimizing the auxiliary drivers there is a high chance that most of our existing non-ad blocking users would never install ad-blockers. There would also be a slight conversion of ad-blocker users if we play our pop-up game right and give customers some flexibility in terms of ads by allowing them to blacklist ads that annoy them. This in-turn would grab the attention of the ad-buyers. Our company could be one of the lucrative ones to the ad-buyers because we had found our way in this war with ad-blockers and we could provide more customer impressions to them. More the impressions of a relevant ad, more the probability of clicking. Hence, by just improving the user experience, we could improve our company revenue, creator revenue (for creators associated with our company for ads) and ad-buyer satisfaction. It is kind of a ‘one solution for all’. Doing this may also cause some creators to push more content than usual.
There is another very important thing to note. As every business process is highly iterative, I would suggest everyone reading this to take this strategy and methodology with a pinch of salt. We need to consistently verify the linkages using methods mentioned in previous page. All the views expressed in this project are purely subjective and by no means it is a full proof strategy to beat the ad-blockers.
References�