We have seen garbage posts on Publish0x receiving hundreds of tips within one day worth of several dollars, while we have seen great informative posts receiving just a few tips worth of a few cents in months. And we wondered: 'Why is that?' Is it about the author? Is is about the timing? Is it about the subject? So we decided to run an experiment to confirm the suspicions we have - that you readers, are not actually reading posts published on the platform, but rather quickly scroll down to the tipping bar to get your tip. Here are the results.

Research Hypothesis, Question, and Methodology

Our hypothesis looking to research was that most readers in the Publish0x platform are not really reading each article before they reach the tipping banner, but rather scroll down to it, without even scanning what's in each post, adjusting the tipping bar to their liking, and hitting the 'Tip' button. We have done that in some boring posts, ourselves - so at least for some readers the hypothesis should be valid.

So the obvious question was not about the hypothesis was happening, but in what extent. Is it true for the majority of the readers, or a small minority?

So, we had to come with a methodology that would be taking metrics from actual reader interactions, rather than a questionnaire - the latter would only provide biased results as most respondents would answer they are actually reading each post. Thankfully, we had several user metrics available as publishers: post views, and user reactions (likes/dislikes). What we needed to do, was to create an article provoking the reader to hit any of the two buttons - in the third wave article we actually revealed our research hypothesis to the readers, further provoking them to prove our research hypothesis wrong. If readers were actually reading each article, they would try to defend their integrity (especially in the third wave post), by clicking on the like or dislike button.

The three waves would be posted throughout different times of the day, to reach users in different time zones.

Igor Tomić from the Publish0x team pointed that some or even many readers might not care to defend their integrity to the research. Although this is a valid hypothesis, it could only be confirmed if in the third - and most provocative - wave we had a spike (significant increase) in reactions.

Data for validation are available to the Publish0x team on their own servers. In fact, the Publish0x team has more data available, as they can filter-out duplicate views, in order to get unique views per post. It it always happens, unique user data provide more positive results to unfiltered data. However, this further research would be useful if our results were inconclusive.

The Results

All three posts received 882 views, and generated 81 positive or negative reactions (that's 9.2% overall reaction rate).

The first wave was the most viewed one with 406 views and 31 reactions (7.6% reaction rate, with 1.6% deviation from the overall results - which is within the statistical error threshold for this sample size).

The second wave received 220 views and 24 reactions (10.9% reaction rate, with 1.7% deviation from overall results - which is within the statistical error threshold for this sample size).

The third wave, with the most provoking content, received 256 views and 26 reactions (10.2% reaction rate, with 1% deviation from the overall results - again within the statistical error threshold).

Waves 2 and 3 combined to provide similar group size to wave 1, received 476 views and 50 interactions (10.5% reaction rate, with 1.3% deviation from overall results - again within the statistical error threshold for this sample size).

With three waves providing very similar reaction rates, it's safe to conclude that no matter how many times the experiment is repeated, the results will be always within the 2% statistical error threshold (7.2% to 11.2% reaction rate). It should be underlined that even the third wave, with the most provoking content, addressing the Publish0x team concerns, did not receive more positive results than the second wave, thus debunking the intended non-reaction attitude. In addition to the above results, we know that the three posts used in the experiment, received significantly higher reaction rate than all our 15 posts published in the Publish0x platform until now, ranging from 1.3% to 6.6%, meaning that real readers were indeed participating in the experiment.

Side Results

For all three posts in the experiment, we earned $0.78, bringing the EPM (earnings per 1000 views) rate to $0.88 for the authors' side - which is somewhat higher than the Google EPM for display ads. With over 500,000 impressions per month, the platform could be earning some revenue from display ads.

Conclusion

As with all incentivised content platforms, Publish0x is not immune to fraudulent behaviour from its users - both authors and readers. For the reader side, it is evident that the vast majority of the readers are not reading the content published by the authors, but rather scroll down directly to the tipping banner, thus blindly tipping authors, not for the quality of their posts, but to get their own piece of the incentive. If readers are incentivised to simply visit a specific part in the article and perform a specific action, that is not connected to the quality of the article, they will adjust their behaviour to earn the most from the platform - they already have.

If the Publisher0x team wishes to promote quality content and quality interactions from its readers, the incentive system should be adjusted accordingly.

If you really liked our research, and you have a BAT or two spare in your Brave browser, we would welcome a tip on our Twitter account, here.