Experimentation / CRO

Experimentation and Testing Programs acknowledge that the future is uncertain. These programs focus on getting better data to product and marketing teams to make better decisions.

Research & Strategy

We believe that research is an integral part of experimentation. Our research projects aim to identify optimization opportunities by uncovering what really matters to your website users and customers.

Data and Analytics

90% of the analytics setups we’ve seen are critically flawed. Our data analytics audit services give you the confidence to make better decisions with data you can trust.

Does the Presenter’s Paradox Actually Work in Digital Marketing? [Original Research]

When people weigh choices, the Presenter’s Paradox says they do so by averaging (not adding) the value of each item in a package.

This means if you add more items to a list or more products to a bundle, it could reduce the overall value perception (if the added items are deemed less valuable.

Research on this phenomenon is fairly scarce, though, so we decided to conduct a study through CXL Institute.

We provide 3 perspectives:

  1. We outline what products and lists two academic studies have tested,
  2. We duplicate a product and list test with a larger sample size to try and replicate the findings, and
  3. We then apply the test to six new products, three experiential products (travel package, hotel night, massage) and three physical products (camera, printer, kitchen mixer).

Results summary

  • Generally, we were able to reproduce results from the two studies:
  • IPOD DEAL - After increasing the sample size, there was still a pattern, though weak, of difference in mean price valuations for the two iPod variations, indicating that the addition of the free song download slightly lowered price perceptions of the offer.
  • SMOKING LIST - The short list of a few persuasive anti-smoking arguments implied more consequences, however, a longer list of ten arguments of varying persuasion communicated more risks.
  • Our own tests likewise didn't get strong results, but, with the exception of the camera, these results generally, though weakly, support the idea of the presenter's paradox.
  • There were no patterns among experience goods (hotel stay, vacation package, massage) vs. the search goods (camera, printer, mixer).
  • These results together suggest that the averaging effect in the presenter's paradox is not strong, is contextual, and likely is influenced by the value gap between the items presented and the presentation of them.

How do I apply this research?

This idea of the presenter's paradox influences how you might bundle packages or offer lists of information when trying to persuade. Lists of persuasive arguments are diluted in their effectiveness if weak arguments are included, though the results of our smoking study tells us that people are affected by the 'quantity' of arguments as well as the 'seriousness' of them, so if you're specifically trying to impress with quantity, a long list might be preferred. Package bundles are tricky, and as Rodger Dooley put it in our post on the subject - “If you are thinking about bundling multiple products, do it right:

  • Avoid mixing a cheap item and an expensive item and simply promoting the package.
  • If you are mixing products with different values, establish the value of the individual items first, particularly the most expensive one.
  • Take a lesson from infomercial producers and emphasize the additive nature of bonus items.
  • Focus on non-price attributes of the product (e.g., durability or comfort) – the researchers say this will reduce the devaluation effect from mixed-value items.”

Background

The basic idea behind the presenter’s paradox:

When constructing a persuasive argument, points of low value (weak aspects of the argument) bring down the strength of the overall argument

Multiple research studies have been conducted that have identified the existence of this phenomenon.  However, we questioned some of the published research methodology, especially the small sample sizes. So here we present:

  1. A review of two papers that first presented the concept
  2. Replication of one study from each paper in hopes of replicated the results at higher sample sizes, and
  3. Six additional studies of our own that test 3 ‘search’ products (printer, camera, and kitchen mixer) and 3 ‘experience’ products (a hotel vacation package, one night in a hotel, and a massage).

1. Review of two ‘Presenter’s Paradox’ papers

In the 2012 research study “The Presenter’s Paradox”, researchers applied the paradox to four studies:

Participants were broken down into two categories:

  1. Evaluators of the variable (iPod, hotel with five-star pool, $750 fine, and $1,750 scholarship).
  2. Evaluators of the variable plus the much less valuable add-on (song download, three-star restaurant, two hours of community service, and $15 towards textbooks).

Here are the results for each test:

In the 2014 study titled “Top 10 Reasons: When adding persuasive arguments reduces persuasion,” four studies were once again tested.  This time, however, researchers applied the presenter’s paradox to persuasive arguments or reasons to do something:

  1. Reasons to quit smoking (all participants were current smokers)
  2. Encouraging young people to vote
  3. Engaging in fitness
  4. Attending a university

There were two versions of a persuasive list for each study: A version with 10 reasons including weak and strong reasons, and a version with two very strong reasons.

Here are the results for the “Top 10 Reasons…” study:

Based on past research, it seems as though the presenter’s paradox applies to a variety of situations:

  • Products (iPod, hotel)
  • Penalties (littering)
  • Awards (scholarship)
  • PSA-type persuasive arguments (all four “Top 10 Reasons…” studies)

2. Replication of two studies with higher sample sizes

Before jumping into our own tests, we first wanted to know for sure that these findings could be replicated, as the sample sizes were very low for most of the paper’s individual studies. So, we recreated a study from each research paper.

iPod Touch

In the 2012 study “Presenter’s Paradox”, participants who saw an iPod alone were willing to pay $65.48 more when presented with an iPod alone than participants who saw an ad for the same iPod plus one free song download.  We re-tested this study, increasing the sample size to 596 (~300 participants per treatment), the original only had N = 41.

Data Collection and Methods and Operations

We asked the same questions as the original study:

iPod only (N=596):  “Imagine you are looking to buy a gift for a friend and you are considering purchasing and iPod Touch. In the store you see the following iPod package for sale:”

Participants viewed a photo of either the iPod alone (n=297) or the iPod with a graphic including one free song download (n=299).

iPod Touch w freebie
iPod Touch media used in survey

They were then asked, “Please estimate how much you’d be willing to pay for this iPod”.

Findings

There was a $11.07 difference (though not significant) in survey participants’ valuation of the iPod Touch without the song download (M=$183.42, SD=$102.06) and the iPod with the song download (M=172.35, SD=$94.44) variations; p=0.17. There is a pattern of higher valuation for the iPod without the free song download, but it wasn’t significant at a high confidence level.

Reasons to Quit Smoking

In the 2014 “Top 10 Reasons” study, a sample size of just 34 participants revealed that two, strong reasons not to smoke are significantly more convincing than ten reasons varying in strength.  We re-tested this study and increased the sample size to ensure the validity behind our findings.

Data Collection and Methods and Operations

First, a survey was administered via Amazon Mechanical Turk with a list of reasons against smoking tobacco products. We asked participants,

Please select the three STRONGEST, MOST PERSUASIVE arguments against smoking cigarettes:

Arguments included:

list of arguments against smoking
List of arguments against smoking

From the results of this survey, we constructed two argument lists:

Two lists of smoking arguments used in study
Two lists of smoking arguments used in study

To match the original study, we asked participants two questions:

  1. After reading this list, how serious are these consequences of smoking to you?
  2. In light of these consequences, how risky is it to smoke?

Both questions were answered on a scale of 1-7 (1=not at all serious, 7=very serious).

NOTE: We did not ask participants the third question, “If you currently smoke, to what extent are you inclined to quit smoking?”

Findings

For question one, participants who read the three reason list rated the consequences of smoking at 6.59/7 (n=300, Mconsequences=6.59, SD=0.777).  Participants who read the ten reason list rated the consequences of smoking at 6.4/7 (n=298, Mconsequences=6.37, SD=1.028).  Those who read the three reason list against smoking perceived significantly more consequences than those who read the ten reason list (p=.004).

For question two, participants who read the three reason list rated the risks of smoking at 6.06/7 (N=300, Mrisk=6.06, SD=0.749).  Participants who read the ten reason list rated the consequences of smoking at 6.46/7 (N=287, Mrisk=6.46, SD=0.876).  Those who read the ten reason list against smoking perceived significantly more consequences than those who read the three reason list (p=0.0298).

Takeaway, we mostly reproduced findings from the two studies: IPOD DEAL – There was a weak pattern of difference in mean price valuations for the two iPod variations, indicating that the addition of the free song download slightly lowered price perceptions of the offer. SMOKING LIST – The short list of a few persuasive anti-smoking arguments implied more consequences, however a longer list of ten arguments of varying persuasion communicated more risks.

3. Testing 6 additional product deals

For this portion of our research, we selected 3 search goods and 3 experience goods, here they are along with the number of people surveyed:

  1. Stand mixer (N = 617)
  2. InkJet printer (N = 578)
  3. DSLR Camera (N = 617)
  4. 60-minute massage (N = 617)
  5. One night at the Westin Cancun Hotel (N = 616)
  6. All-inclusive trip to Italy (N = 603)

Here are the resulting 2 sample t-tests to test differences in the average value across the two groups in each study:

Stand mixer

Confidence intervals and estimated difference between value estimations of the mixer when it was offered by itself (top) and with a cheap 'extra' (bottom)
Confidence intervals and estimated difference between value estimations of the mixer when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case, a free cook book.

InkJet printer

printer
Confidence intervals and estimated difference between value estimations of the printer when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case a free gift card and redeemable harddrive.

DSLR

Confidence intervals and estimated difference between value estimations of the camera when it was offered by itself (top) and with a cheap 'extra' (bottom)
Confidence intervals and estimated difference between value estimations of the camera when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case, a free additional battery.

60-minute massage

Confidence intervals and estimated difference between value estimations of the massage when it was offered by itself (top) and with a cheap 'extra' (bottom)
Confidence intervals and estimated difference between value estimations of the massage when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case, a bottle of water.

One night at the Westin Cancun Hotel

Confidence intervals and estimated difference between value estimations of the mixer when it was offered by itself (top) and with a cheap 'extra' (bottom)
Confidence intervals and estimated difference between value estimations of the hotel night when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case, free valet parking.

All-inclusive trip to Italy

Confidence intervals and estimated difference between value estimations of the mixer when it was offered by itself (top) and with a cheap 'extra' (bottom)
Confidence intervals and estimated difference between value estimations of the Italy trip when it was offered by itself (top) and with a cheap ‘extra’ (bottom), in this case, boat cruises, cooking class, and concierge services.

Takaways – With the exception of the camera, these results generally, though weakly, support the idea of the presenter’s paradox. There were no patterns among experience goods (hotel stay, vacation package, massage) vs. the search goods (camera, printer, mixer).

Note the camera bundle was perceived much more valuable (~$50) with the addition of a spare battery. This was the strongest pattern seen of any comparison. See our note below in ‘Limitations’ regarding a theory on this.

Limitations

In the “Reasons to quit smoking” replication portion of our study, we did not ask participants the third question which centered around likeliness to quit smoking after reading the persuasive list.  We weren’t confident that the results of this question would stem from the list that participants read, but instead would stem from extraneous outside factors.  Extraneous factors may have very well played a role in the results of this study anyway, considering that we do not know whether our participants were smokers, had lost a loved one to lung cancer, etc.

Additionally, we replicated just one test for each study.  While our results reflect those of the original studies, we cannot speak for the studies we didn’t replicate.

We’re curious about the variable of product or package design, and if the aesthetics of the image affects perceptions in any way. Presenting the offer via a text description would have eliminated that variable and potentially provided a more valid test of the presenter’s paradox.

Side note on the results of the inkjet printer test: There was a large difference in variation between the two surveys (with and without the freebie) that wasn’t seen in any other study (see the shape of the blue distributions within that study and among studies generally). We couldn’t explain this strange pattern after inspecting the study setup, but suspect something was screwed up in that test. With no real reason otherwise to suspect an error, we present it as is and added this note.

For the DSLR camera study, participants estimated that the camera plus freebie would be worth more (contrary to the presenter’s paradox).  There is a possibility that, because the camera had a picture of the freebie, participants felt like they were getting more and so estimated a higher cost.  Notice that the other variations, which all follow the presenter’s paradox trend, do not include a picture of the freebie.  This is an interesting nuance that may be worth further research in the future.

Conclusion

While we could reproduce the original studies, our own research on products showed weak support that the Presenter’s Paradox exists. There weren’t any patterns between the different product categories (experience goods vs. search goods).

These results together suggest that the averaging effect in the presenter’s paradox is not strong, is contextual, and likely is influenced by the value gap between the items presented and the presentation of them.

Related Posts

Who's currently reading The Experimental Revolution?