Send
The business platform for digital media

Setting Up a Metered Paywall I: Estimation

Oct 07, 2015
Analytics

This series of three articles that Piano’s Lead Data Scientist Roman Gavuliak wrote in 2014 and is being reposted for the edification of Piano’s current clients and readership.

Setting up a metered paywall II: Data Sources
Setting up a metered paywall III: Time Period
To read all of Roman’s articles please click here

A metered paywall is one of the most popular options to monetize media website content. The notion seems simple: users can read a predetermined number of articles for free before being asked to pay for more access. Usually only regular users will be impacted; those who visit occasionally will probably never reach their pageview limit.

Where to set the meter limit?

Setting the meter limit is literally the million dollar question. Further, any kind of limit needs to be tied to a particular time period, so instead of one question, there are two:

1. How many free articles?

2. How often should the meter refresh? Daily? Weekly? Monthly?

Choosing the limit

Setting the limit too low means users will reduce their engagement, search for information elsewhere or simply leave. On the other hand, setting the limit too high results in little to no conversion. Naturally limits can be adjusted, but doing so might both confuse users as well as discourage others from implementing a metered paywall.

One of the easiest approaches to setting the meter is to base it on average pageviews per user per time period. This however carries all the risks of aggregation and approximation. If my webpage had two users and in a month one of them would access 1 page only while the other 100, my average pages per user per month would be 50.5, same as the median (no mode though). In basing a limit on such a metric, there are three possible outcomes. My meter might affect 0%, 50% or 100% of users. While such a case is unreal for any relevant webpage, the implications of this mental exercise are applicable to any possible real life example. An average value can be skewed in both directions and barely begins to describe the real user content consumption. Besides, average pageviews per user per time period is not a ready-made metric in all website traffic tracking systems.

That is a lot about how to not estimate metered limits, the question remains: is there a right way to estimate limits for the meter? Ideally what is needed is an extraction that statistically biased people might relate to an inverse cumulative distribution function for how much content users consume. Without talking much about theory, here is an example shown in the following chart:

Content comsumption

The x-axis represents the amount of articles read – 1+ means more than one article, 2+ more than two articles etc. The y-axis maps the corresponding share of users belonging into such a category. If the meter is set at a limit of 6 articles, it would affect all of the users in the 5+ category – over 30% users. This chart is purely illustrational, but in vast majority of cases the curve drops steeper in the beginning and the decline slows as it progresses further.

Keep in mind that this curve should only include article consumption because neither home or section front pages are counted in metered limits. For many web pages, the ratio of non-article pages (home page, section fronts) to article pages is 1:1. If all the necessary calculations needed to construct such a curve were applied, the data needed to calculate the basic monetization potential of the metered paywall using different conversion scenarios is now available.

The concepts outlined here are not the only ones we take into account when analysing and creating recommendations. Besides data relevance and the correct time period for a metered paywall we also look into site consistency as well as content consumption dynamics over time.