The Subscription Price Index.

The named stack.

Six apps that between them cover notes, budgeting, tasks, archiving, writing and household finance. Chosen once and held fixed, so the series measures price rather than our changing opinion about which apps matter.

Prices are in USD before tax. Where a vendor publishes both an annual and a monthly rate we list both; the "per month, annually" column is the annual price divided by twelve, which is how most vendors present it themselves. Mixing the two modes inside one basket is the commonest error in this genre and it inflates the total by half.

Ten years of it.

A subscription is not a price, it is a price per year, for as long as you keep the thing. Two forces act on that and they pull against each other. Prices tend to rise, so the nominal total is larger than a flat multiplication. Money also loses value, so a payment made in 2036 costs less in real terms than the same payment made today. The honest comparison against a single payment made now is the one in today's money.

Escalation of 3.5% a year in the central case, banded 2.4% to 6.0%. Deflated at 2.4%, the intermediate long-range CPI assumption in the 2026 Social Security Trustees Report. Note how little the escalation actually matters: over ten years it adds about five percent to the total in the central case. We would rather publish the small honest number than the large borrowed one.

The list price is not the price.

This is the finding we did not expect. Most of these apps have not raised a published price in years.

• Ulysses (unchanged since 2017)
• Trello (unchanged since 2021)
• Obsidian Sync (unchanged since 2024)

What has risen is the cost of the same experience, delivered by removing things from the free tier, unbundling a feature that used to be included, or deleting the cheap plan so the cheapest option becomes the expensive one. None of that appears in a conventional price series. All of it appears on your bank statement.

Changes to a published price.

A row appears here only when a source states both the old figure and the new one. Where we know a price went up but can only source the new number, we leave it out rather than infer the old one, which is why this table is shorter than it could be. It is the beginning of a series rather than a history: archived pricing pages were not available to us. One row is marked press, meaning the change is documented in reporting rather than by the vendor, because Strava's own announcement of its increase states no figures at all.

The whole index.

Every product we priced, including the ones outside the named stack, and including the ones we could not price at all.

Downloads.

Published under CC BY 4.0. Quote it, chart it, put it in your own spreadsheet. Attribution to Offline.Ltd with a link is all we ask, and we would rather you checked a figure and corrected us than took it on trust.

What we could not verify.

A price we could not read at source is left blank. A blank is honest; a guess dressed as a measurement is not. Seven entries in this index carry no number, and the reasons are interesting in themselves.

Things 3, Scrivener and Streaks show no figure either, but for an unrelated reason: they are bought once, so they have no annual or monthly price to record. Their prices are in their notes.

Method.

We price each product at the plan a single person would actually buy, at the cheapest published billing cadence, before tax, in US dollars. Every figure is read on a page the vendor controls: a pricing page where possible, a help centre or vendor blog where the pricing page serves no figures to anything but a browser, and the App Store where the vendor publishes no price on its own site. We do not take prices from comparison sites, review sites or "alternatives to" pages, which is where most wrong numbers in this genre come from.

Each row carries a confidence. Confirmed means we read it on the vendor's own page on the date shown. Unverified means we could not, and the row carries no number. 38 of the 45 rows in this index are confirmed.

Per-seat products are marked, because a per-seat price is not comparable to a personal one without saying how many seats. Products with meaningful free tiers are noted, because a free tier that shrinks is a price rise, and this index treats it as one.

Two caveats we would rather state than have pointed out. Prices here are not quality-adjusted: the 2026 version of most of these apps does more than the 2019 version did, and a strict price index would account for that. And real price changes arrive as steps, not as a smooth annual rate, so the escalation figure is a modelling convention rather than something anyone experiences. Todoist did not raise its price by two percent a year; it raised it by twenty five percent on one Tuesday in December.

On the numbers people quote.

The figures in circulation for software price inflation are mostly not measuring what they are used to measure, so it is worth being explicit about what exists.

• No statistics agency publishes a consumer software subscription index. primary

  The US Bureau of Labor Statistics excludes streaming, online subscriptions and app fees from its internet services category outright. Its "computer software and accessories" series bundles software with flash drives and blank media, and Federal Reserve Board staff published a note in May 2026 attributing a recent surge in that series to flash-memory prices during an AI datacentre shortage rather than to software at all. Eurostat has no equivalent index at any level.

• The widely quoted 12 to 16 percent "SaaS inflation" figures come from one vendor. vendor

  They are published by a company that sells software for negotiating SaaS bills downward, drawn from enterprise procurement contracts rather than consumer apps, with no published methodology and figures that differ between the company's own pages. We mention them only because they are quoted everywhere, and we do not use them.

• The closest official proxy runs at about 3.6 percent a year. primary

  The BLS index for cable, satellite and live streaming television rose from 475.3 to 609.8 between June 2019 and June 2026. It is dominated by legacy pay television, so it is a weak stand-in for app subscriptions, but it is the nearest thing an agency publishes.

• The long-run inflation assumption is 2.4 percent. primary

  The intermediate assumption in the 2026 Social Security Trustees Report, which is what US federal actuaries use for long-horizon projections. The Cleveland Fed's ten-year expected inflation model agreed to within three hundredths of a point when we checked.

• Which is why we measured a basket instead.

  The number on this page is small, self-measured, and fully auditable, and we think that beats a large number borrowed from someone selling something. If you find a better source, we will use it and say where it came from.

Why we publish this.

We sell knives you pay for once, so a page about the cost of subscriptions is not a neutral document and we will not pretend otherwise. What we can do is make every figure checkable, publish the ones that weaken our case alongside the ones that help it, and put the data in a file you can open. Three of these apps have not raised a price in years. One has held its annual price since 2017, which is a real-terms cut. Those facts are in the table because leaving them out would have been the easier and worse choice.

The arithmetic still lands where it lands. Ten years of this basket is $5,542 in today's money. An Every Blade Ever knife is $199, once, and then never again.

See the full comparison against our own prices, or do your own math in the Forge.