AN EXPERIMENTAL STUDY ON DECEPTIVE ONLINE POST-TRANSACTION MARKETING OFFERS

  • : Ms Word Format
  • : 70 Pages
  • : ₦3,000 | $25 | ₵60 | Ksh 2720
  • : 1-5 Chapters
  •  
  • Click to DOWNLOAD Materials

AN EXPERIMENTAL STUDY ON DECEPTIVE ONLINE POST-TRANSACTION MARKETING OFFERS

ABSTRACT

Post-transaction marketing offers are often designed to mislead consumers into purchasing products that are only tangentially related to their primary purchase. These products are presented directly following the completion of legitimate first-party transactions. To sell these offers, marketers use deceptive tactics that include illegally acquiring data from merchants and exploiting cognitive psychology. These practices are not only unpopular with consumers, but are possibly illegal. A US Senate report released in 2009 estimated that between 1999 and 2009 over 35 million people were misled into purchasing offers that they did not want and did not use. Due to the lack of regulations, and the global nature of the Internet, these deceptive practices are far from being extinct. In order to address this problem, this thesis presents a scenario, based on actual cases, where a small music shopping site has partnered with a third-party marketer to show post-transaction marketing offers to consumers. After a successful primary transaction in the music store, customers are automatically redirected to a post-transaction offer.  The goal of this thesis is to examine the effectiveness of two ways of intervening to mitigate the harm of the deceptive post-transaction offer. This problem can be addressed from the perspective of the firstparty e-commerce site or the perspective of the third-party marketers. This thesis presents the results of two experiments, each with over 450 participants, which test solutions from these two perspectives.  The first experiment highlights the difficulty in reducing the harm of posttransaction offers from the first-party merchant perspective – only the strongest intervention significantly reduced the likelihood of participants purchasing a post-transaction offer. The second experiment empirically tests how different presentations of post-transaction offers play a role in how many participants purchase the offers. By gathering data on the effectiveness of intervening from the first and third-party perspectives, this thesis aims to inform regulators about these issues for use in future policy decisions.

 

Chapter 1 

Introduction

Deceptive post-transaction marketing offers can be harmful. These offers are typically presented to customers on e-commerce sites after making a primary purchase. These posttransaction offers are usually of little to no value to consumers. Therefore, third-party marketers often rely on deceptive tactics to mislead customers into purchasing these offers. These deceptive tactics include collaborating with merchants to acquire consumer credit card numbers and exploiting cognitive psychology.

Deceptive post-transaction offers are a significant problem. As of 2008, the Federal Trade Commission (FTC) had brought cases against over 300 companies and individuals (Ballaré and Von Bergen 2008) for practices that are “likely to mislead the…reasonable consumer” (Cliffdale Associates 1983). To investigate the extent of these practices, the US senate compiled a report in 2009 that detailed how nearly 4 million customers were currently enrolled in deceptive posttransaction offers and over 35 million offers had been purchased since 1999. More shockingly, 99% of the people surveyed in this report that purchased offers never used the product and were not satisfied with their purchase (US Senate 2009).

In order to address this problem, this thesis presents a scenario, based on actual cases (e.g. Meyer 2009), where a small music shopping site has partnered with a third-party marketer to show post-transaction marketing offers to consumers. After a successful primary transaction in the music store, customers are automatically redirected to a post-transaction offer.  The goal of this thesis is to examine the effectiveness of two ways of intervening to mitigate the harm of the deceptive post-transaction offer. This problem can be addressed from the perspective of the firstparty e-commerce site or the perspective of the third-party marketers. This thesis presents the results of two experiments, each with over 450 participants, which test solutions from both of these perspectives.

The first experiment attempts to intervene at the level of first-party merchants to reduce the harmful effects of these post-transaction offers. This experiment considers interventions that use prominent interstitial (full page) warning screens and those that do not. On these interstitial screens, the effect of interventions that require participant action to advance the study (active interventions) as well as those that do not require input from participants (passive interventions) are tested. Through experimentation, we find that intervention from the first-party perspective is a difficult task – only the most effortful of the interventions have a noticeable effect on reducing the number of people that purchase the post-transaction offer. From this, we conclude that intervention from this perspective may be helpful, but that intervention from the third-party perspective is also needed.

The second experiment intervenes at the level of the third-party marketers. It empirically tests how effective regulations required by the United States government (through the “Restore Online Shoppers’ Confidence Act” (ROSCA; U.S. House 2010)) actually are at reducing the amount of people that purchase deceptive post-transaction offers. ROSCA prescribes that marketers who use these post-transaction offers clearly disclose the terms of their offer and require an “additional affirmative action” to complete a sale (U.S. House 2010). This experiment tests a variety of interface differences and ways of presenting offers to customers (e.g. opt-in vs. opt-out) to see their effects on conversion of customers. The following section further discusses the aims and contributions of this thesis as a whole and of the experiments conducted within.

Contributions

General contributions

The contributions of this thesis can be divided into general contributions of the study and contributions made by the individual experiments conducted within. The general contributions of the study are:

  • The experiments conducted in this thesis are the first experiments on deceptive posttransaction marketing. Therefore, the results of these experiments are potentially informative to policy makers to form future regulatory policies.
  • These experiments are part of a small body of sophisticated studies conducted on Amazon Mechanical Turk.

First experiments on deceptive post-transaction marketing

Researchers and government bodies have previously examined the ecosystem of retailers and marketers to highlight the extent of post-transaction marketing and the strategies that are currently being used to deceive people (e.g. US Senate 2009; Meyer 2009). There has not, at this point, been a study that isolates the process that customers encounter when faced with deceptive post-transaction offers. By conducting a controlled experiment (in an naturalistic environment), this thesis is able to test the effects of various presentations of deceptive post-transaction offers on consumers. This thesis also tests the effect of warning messages before deceptive offers on consumers’ willingness to purchase these offers. This thesis is the first to gather experimental data on these different phenomena.

This thesis presents Mechanical Turk experiments

The experiments conducted in this thesis contribute to a small body of sophisticated studies that have been conducted on Amazon Mechanical Turk (e.g. Downs et al. 2010; Garg et al. 2012; Nochenson and Grossklags 2013; Reitter et al. 2013). It has become commonplace to use Amazon Mechanical Turk for research that consists solely on surveys. Research that uses Mechanical Turk for a study that has a self-contained environment with experimental treatments is rare. Therefore, this thesis is noticeable since it adds to the growing literature of such experiments. Participants in both experiments conducted in this thesis visit a website specifically designed to experiment with deceptive post-transaction marketing offers.

Experiment 1

The first experiment conducted in this thesis investigates mitigation of the problem of deceptive post-transaction marketing from the perspective of first-party retailers. There are three research questions addressed by this experiment:

  • Does warning participants before they see a deceptive post-transaction offer affect the ability of the marketer to use the offer to deceive them?
  • Are full-page interstitial warnings more effective than warnings that are part of the preceding first-party transaction page?
  • Are active interventions more effective than passive ones?

Effectiveness of warnings

The main purpose of the first experiment is to test if better informed consumers are more likely to be able to resist deceptive post-transaction offers. This experiment offers warning messages to participants before they see the post-transaction offer to better inform them about the nature of the page that follows. Presumably, consumers that see these warnings will be better informed and will purchase fewer low-value offers than less-informed consumers.

To examine the effectiveness of these warnings, Experiment 1 includes a single control treatment (T0) that does not offer participants any warning message before the post-transaction offer. It also includes four experimental treatments that present different warning messages in different modalities. One of these treatments that presents a warning (T4) had a significantly lower conversion rate (rate of purchase of the post-transaction offer) than the control treatment. All other experimental treatments had a lower conversion rate than the control treatment, though those effects were not significant. Therefore, warning participants before they encounter deceptive post-transaction offers does appear to be effective, though the warning needs to be designed in the thoughtful way.

Full page warnings versus on-page warnings

To investigate the best way to frame warnings, the effects of both full-page and on-page warnings are tested in Experiment 1. Presumably, full-page warnings will be more effective than on-page warnings2 at reducing the number of people that purchase post-transaction offers. However, it is possible that alerting participants to the possible deception to follow (though a fullpage or on-page warning) will have the same effectiveness, regardless of the form of that warning. From a policy standpoint, it is likely easier to get compliance with a policy that requires on-page warnings than one that requires full-page interstitial warnings (since interstitial warnings are more likely to disrupt the flow of a customer). Therefore, the relative effectiveness of on-page and full-page warnings is of interest to policy makers.

To study this, a single treatment (T1) in Experiment 1 shows participants an on-page warning, while three treatments show participants a warning on a full interstitial page. The treatment that showed an on-page warning had a higher conversion rate than the treatments that used a full-page interstitial warning. However, the reduction in conversion rate from the on-page warning treatment to the full-page warning treatment was only marginally significant for one of the full-page treatments (T4). Thus, there is support for the hypothesis that full-page interstitial warnings are a more effective intervention than on-page interventions, though the nature of the interstitial warning does matter.

Active versus passive interventions

A third phenomenon tested in Experiment 1 is the role of active versus passive interventions. Both types of interventions are a subset of the interventions that use a full-page interstitial screen to warn participants about the post-transaction offer to follow. Active interventions are those that require participants to perform an action to dismiss the warning, while in passive interventions, warning messages are automatically dismissed after a set period of time. It is likely that active interventions reduce the number of conversions more than passive ones, since the requirement to perform a task is more likely to lead to participants reading the text on the page.

                                                                                                                                                             

2  On-page warnings refer to warnings that are part of the preceding first-party transaction page that precedes the post-transaction offer page. These are contrasted with full-page interstitial warnings that are presented on a separate page between the first-party transaction page and the post-transaction offer.

Active and passive interventions were realized in this experiment by a page containing a warning message that participants dismiss by pressing a button and a page containing a warning message that automatically closes itself after a set period of time, respectively. In this experiment, there was a single active intervention that required participants to press a button, and there were two passive interventions. One passive intervention closed itself after being shown to participants for 5 seconds and the other closed itself after being shown for 10 seconds. There was no significant difference in conversion rates between the two passive interventions. The active intervention resulted in a marginally significant lower rate of conversion than the passive interventions. Thus, there is support for the hypothesis that active interventions are better than passive ones at informing participants prior to the post-transaction offer.

Experiment 2

The second experiment conducted in this thesis investigates mitigation of the problem of deceptive post-transaction marketing from the perspective of third-party retailers. This experiment addresses the following three questions:

  • Are post-transaction offers that require consumers to opt out more harmful than those that, by default, do not opt in consumers?
  • Does the use of data sharing agreements between merchants and marketers (i.e. data pass) help them sell more post-transaction offers?
  • Does consumer knowledge of the presence of data-pass arrangements decrease the number of conversions?

 

Opt-in versus opt-out

Post-transaction offers can be classified as either opt-in or opt-out. Opt-in offers are the norm for product purchases on the web. An opt-in offer proceeds as follow: a consumer purchases one item and is then asked if they would like to purchase another. If they wish to purchase the second item, the total amount of their transaction will reflect the purchase of two items, not just one. Opt-out offers assume that consumers want the second product, even before asking them. Then the consumer only needs to be asked if they wish to not purchase the second product. For example, a consumer goes to a travel website where they purchase their ticket, and are then told they have purchased travel insurance along with their ticket. If the consumer instructs the website that they do not want travel insurance, the amount of the travel insurance will not be added onto the final transaction. It is likely that opt-out offers will have higher conversion rates than opt-in offers since ignorant consumers will follow the default (which in opt-out cases leads to an extra purchase).

In Experiment 2, three treatments are opt-in and three are opt-out. In the opt-in treatments, participants were required to hit a large button to consent to the sale of the posttransaction offer presented after they purchased a song. In the opt-out treatments, if participants did not hit the button, they were charged for their purchase of the post-transaction offer. Each treatment also had other characteristics, which are discussed below. Therefore, when determining the effectiveness of opt-in and opt-out treatments, it is only reasonable to compare pairs of treatments that have all the same properties except for opt-in or opt-out status. After running the experiment, the opt-out version in one pair of treatments was significantly higher than its opt-in counterpart[1] . The conversion rate for that opt-out treatment was three times that of the opt-in one.

This gives weight to the hypothesis that opt-out offers are more harmful than opt-in offers.

Effectiveness of data pass

Data pass is the practice of merchants sharing personal information with other parties. In this context, merchants pass credit card data, which they acquire from the primary transaction, to third-party marketers who use the data for post-transaction marketing offers. These types of arrangements are frowned upon by credit card companies and the United States government (US Senate 2009). Visa’s rules explicitly forbid sharing credit card numbers or sensitive information with parties not included in a transaction (Metzger 2010). However, even though data pass between merchants and third-parties is explicitly forbidden, merchants can partner with third parties to subvert these regulations. Through these partnerships, they can, for example, delay charging a consumer’s credit card until after a post-transaction marketing offer is presented. Though this behavior is certainly not in the spirit of the law, the legality of this practice is nonobvious.

In this experiment, participants do not actually enter any payment information (e.g. credit card numbers). Instead, they are given an endowment that they use to conduct the primary transaction (the purchase of a song) and to purchase the post-transaction offer. To simulate the transfer of payment information from first-party merchants to third-party marketers, this experiment asked for some personal, non-financial, information from participants (including their email addresses). Some of this personal data (email addresses) were passed to the post-transaction offer from the first-party merchant, thus simulating data pass[2] .

This experiment studies whether or not the use of data pass arrangements increases the number of consumers who purchase post-transaction offers. Four of the treatments in this experiment used data pass, while two did not. It is expected that for opt-in treatments, using data pass will increase the number of conversions, since it reduces the amount of effort participants need to expend to purchase the post-transaction offer. The opposite effect is expected for opt-out treatments, since data pass eases the process of opting out. The experimental results show that, in opt-in treatments, data pass arrangements resulted in significantly higher conversion rates than counterpart treatments that did not use data pass. Similarly, the opt-out treatment that had the highest conversion rate is the one that did not use data pass. Thus, the experimental results support the reasoning that data pass is effective at reducing barriers to purchasing post-transaction offers.

Consumer knowledge of data-pass arrangements

When it comes to using the tactic of data pass, merchants and marketers need to decide whether or not to disclose this practice to consumers. Using data pass raises the likelihood of participants purchasing a post-transaction offer. Therefore, this strategy is attractive for merchants and marketers to use. However, the effect of informing participants that their information is being transferred is still unclear. Consumers who know data pass is taking place are presumably more likely to purchase post-transaction offers. Data passed from first to third parties gives consumers an indication that the two sites are partners, or even that they are not two distinct sites. Therefore, consumers are likely to trust the offer and to not read it as closely as they would have if they were not given this positive indicator.

To test the effect of informing consumers about data-pass arrangements, the posttransaction offer in Experiment 2 either displays or hides the data that is passed from the firstparty music store. In two treatments, the data is shown on the post-transaction page in a box that is pre-filled. In another two treatments, that same information is passed, but the box that would be filled is hidden from the participants. The experiment is concerned with the effect of hiding the presence of data-pass arrangements. The experimental results indicate that both treatments that hide the presence of data-pass arrangements had significantly lower conversion rates than their counterpart treatments that disclosed the data-pass arrangement. This gives reason to believe that hiding the presence of data pass does indeed reduce the likelihood of consumer purchase of posttransaction offers.

Structure of Thesis

The structure of the thesis is as follows: Chapter 2 discusses background information that was used to motivate this study and is useful in understanding it. Chapter 3 discusses the methodology employed in this study. It details reasons for conducting this type of study and the specific design of both of the experiments conducted here, including motivations, procedure and treatments. Chapter 4 presents general results captured across both experiments. Chapter 5 and Chapter 6 present the results of Experiment 1 and Experiment 2, respectively. Chapter 7 distills the results from the proceeding sections into a discussion about the implications of the results found in the context of the contributions presented earlier. Chapter 8 concludes the thesis with high-level lessons learned and Chapter 9 details next steps for this strand of research

[1] In all other pairs of opt-in and opt-out treatments, there was no significant difference in conversion rate.

[2] Thus, the data pass in this study involves the transfer of email addresses instead of credit card information. And, since this study involves a music store and a marketer that were invented solely for it, the email addresses are not actually being passed from one company to another, though it appears that way.

AN EXPERIMENTAL STUDY ON DECEPTIVE ONLINE POST-TRANSACTION MARKETING OFFERS

Sharing is caring!

Leave a Reply