What subscriptions cost over time

One row per plan, not per service: the index averages a service's plans together, so a service whose plans moved differently appears more than once. A plan appears only when we have a price at both ends of the window — one reading is not a trend, and showing it at 0% would imply we had checked twice. The price it started from also has to have been read within about a year of the window opening: between readings a price is carried forward, and a figure carried a long way tells you what a plan cost but not when it changed, so a rise we never watched happen would otherwise be dated to a year in which we did not see it. Prices are per month; an annual plan is shown at a twelfth of its price. A ~ marks a figure taken from a report at the time rather than read off the vendor's own page.

100 is not a price. It is where the line starts, in May 2012. Starting it somewhere else would move the whole line up or down and change none of the percentages between any two months.

A service joining cannot move the index. On the month a service or a plan first appears there is no previous month to compare it against, so it contributes nothing. That is what lets the catalog we started scraping in 2026 join a history that runs back years, without putting a step in the line that no price change caused. A service leaving is the same in reverse.

It is not a fixed shopping basket. This does not say that some particular set of subscriptions costs 67.90% more than it used to. It is month-on-month comparisons multiplied together over a membership that changes as we are able to price more services, so two distant readings are not the same thing as comparing those two months directly.

Between readings, a price is carried forward — a subscription nobody told you about is still costing what it cost last month. So a rise shows up in the month we read it rather than the month it happened, unless the reading is a historical one that carries a real date. After 12 months with no reading a service drops out of the basket instead of being assumed unchanged, and a service we have only ever priced once never enters it: we would be turning "we have not looked" into "the price held".

The dashed stretch rests on fewer services. Everything up to Apr 2018 is computed over a much narrower basket than what follows it, which is a change in what the line is about rather than in what it measures.

Some of the history is approximate. Figures marked ~ are right but not certain — taken from a report at the time rather than the vendor's own page, or dated only to the month. The first one enters the comparison in Apr 2014, and every level from there on carries that uncertainty forward. Only the months where an approximate figure entered the comparison are marked — not the ones where a price we already had was simply carried along, and not the month a new plan appears, which has nothing to be compared against. A mark on everything would tell you nothing.

Before Aug 2026 the line is reconstructed from published sources, each one quoted and linked; from Aug 2026 it is read from each service's own pricing page.

This index revises, and a real one would not. The whole line is recomputed from the table of readings every time this page is drawn, so adding a historical price we did not have before changes the past as well as the present. National statistics offices publish a number and never touch it again. We are not doing that, and you should know it.

Readings come from each vendor's own pricing page and are stored with the date and the source URL; where a historical price predates our tracking it is entered by hand from a cited source and marked as such. The whole series is derived from that table and nothing else. This page used to show an average of a basket fixed once and never changed, which could not begin until the newest service we tracked had appeared and so threw away every year of history before that. Chaining replaced it, and the published figure changed with it.

Multiply a month's index by the next month's move and you get the next month's index. Fresh is how many services had a new reading that month, carried how many kept the price we last saw.