Revel Roots

what apps actually cost

What They Actually Cost

There is no single true figure for what households spend on subscriptions, and the published numbers differ by more than fifty dollars a month. That spread is informative rather than confusing, once you know what causes it.

Reviewed August 9, 2026. All figures are consumer surveys, not measurements of actual billing data. For a workplace cost comparison, this page provides another example of how recurring software turns time or labor into a billed metric.

The published numbers

$273 a month. West Monroe, State of Subscription Services Spending — a nationally representative consumer poll, up from $237 when first measured in 2018. Covers streaming, software, fitness, food delivery, cloud storage and recurring physical goods.

$219 a month. Widely cited from a different survey and methodology, and generally covering a narrower set of categories. For broader context on recurring costs, billing, or household software spending, Microsoft account and billing support provides an outside reference point.

11.2 active subscriptions per household. Deloitte's Consumer Tracker. A count rather than a total, and easier to check against your own situation. (Nearly half of US consumers cancelled at least one last year.)

Around $66 a month on AI subscriptions specifically, across an average of four services, per Bango research from November 2025 — which also found 53% of people cancel and restart AI tools as needed rather than holding them continuously.

Why they disagree

Different category boundaries. Does a meal kit count? A gym membership? Cloud storage bundled with a phone plan? Each survey draws the line somewhere and rarely states where.

Self-report against a known perception gap. 89% of people underestimate their own spending, which means every one of these surveys is asking people a question they answer wrongly. Surveys that prompt category by category get higher figures than those asking for a total.

Different populations. Households with children, households in metropolitan areas, and households under forty all subscribe differently.

And different years. Prices in this category rise annually, so a 2024 figure is not a 2026 figure.

The honest reading: somewhere in the low-to-mid hundreds per month for a typical US household, higher than the household believes, and rising.

What the surveys cannot tell you

Your number. No survey describes your subscriptions, and the average is composed of households nothing like each other.

Which of yours are unused. The interesting figure is not the total — it is the portion you would not miss.

And what you would actually cancel. Some subscriptions are worth well above their price and some are worth nothing, and no aggregate distinguishes them.

Finding your own figure

Twelve months of statements, not memory. Card, bank, and PayPal.

Plus the app store purchase histories, which are separate systems and where the forgotten charges live. Both Apple and Google keep a full subscription list in account settings, and most people have never opened it.

Plus anything billed annually, which will appear once in twelve months and is the easiest category to miss entirely.

Write each one down with its annual cost. Not monthly. The annual figure is the one that changes decisions.

The three columns worth adding

Once the list exists, three marks per line take five minutes and do most of the work.

Used in the last month? A tick.

Would I notice if it stopped? A second tick. (And check whether something else already does the same job.)

And is there a free tier that would do? More often than people expect, and in some categories a one-off purchase still exists.

Anything without either tick is the answer to the question you started with. In most lists that is two or three items, and occasionally it is considerably more.

What to expect

Your total will be higher than your estimate. Two thirds of people are out by more than $200 a month, and there is no reason to assume otherwise about yourself.

And two or three charges will be unidentifiable without looking them up. That is the normal result and it is the strongest argument for doing this at all.

The short version