How Software Is Priced in 2026: The $18 Median and the $178 Mean

Ask a founder what to charge and you will get an anecdote. We wanted the actual distribution, so we read the pricing pages of 5,194 software tools that publish at least two plans and measured how the market really prices itself. The headline number is deceptively simple: the median tool starts at $18 a month. The mean is $178. That ten-fold gap is the whole story of how software is priced in 2026.

The striking part is not that the two numbers differ. Averages are always dragged around by outliers. The striking part is how completely the mean misses the market it claims to describe. At $178, the mean lands inside the most expensive band we measured, the one that holds 14% of tools. In other words, the "average price of software" is a price that roughly one tool in seven charges, and that no ordinary buyer will ever pay. The median, $18, lands inside the single largest band in the catalog. One of those numbers describes the market. The other describes its tail.

How we built this report

This report covers 5,194 tools in the Toolradar catalog that publish at least two pricing plans, measured against our August 2026 catalog snapshot. For each tool we recorded four things: whether it has a free tier, whether it offers a free trial, its starting paid price (the cheapest tier above zero), and how many tiers it lists. Starting prices are normalized to a monthly figure where possible and capped to plausible values. Median and mean are computed over tools with a parseable paid starting price. This is descriptive, first-party analysis of public pricing pages, released under CC BY 4.0.

Three limits are worth stating up front, because they shape how you should read every number below.

Every count here is a lower bound. The catalog is a living dataset that grows every week, and 5,194 is the population of tools that met the inclusion rule at snapshot time, not the population of priced software in the world. Treat the shares as the shape of the market and the counts as a floor.

The inclusion rule biases the sample downward in price. We only measured tools that publish at least two plans on a public page. Products that hide everything behind "contact sales" are, by construction, absent. Those are overwhelmingly the expensive ones. So if anything, the real market is more expensive than our distribution suggests, and the $178 mean is a conservative estimate of the tail rather than an inflated one. The same rule excludes single-plan flat-rate products at the bottom, but that group is far smaller than the sales-gated group at the top.

Rounding is visible. Band shares are published to the nearest whole percent, which is why the five price bands sum to 101 rather than 100. We have not silently adjusted them to hide it.

The shape of the curve: a $18 median against a $178 mean

The single most useful fact about software pricing is that the average is a lie. Here is where starting prices actually land.

Where software starting prices actually land

Share of 5,194 tools by cheapest paid tier, normalized to monthly

Under $10/mo

27%

$10 to $25/mo

33%

$25 to $50/mo

17%

$50 to $100/mo

10%

$100/mo or more

14%

Read the bars left to right and the distribution does something unusual: it falls, then it stops falling. Share drops from 33% to 17% to 10% as prices climb, exactly what you expect from a normal long tail, and then it jumps back up to 14% at the top. That final bar is not a tail. It is a second population.

This is why the mean and the median disagree so violently. Means are the right summary for symmetric distributions with a bounded range. Software starting prices are neither. The bottom of the range is pinned at zero, the top is effectively unbounded (infrastructure and enterprise platforms can open at hundreds or thousands a month), and the mass sits near the floor. Under those conditions the mean is not a description of the center, it is a weighted vote in which the most expensive tools cast the most ballots.

Notice where each summary statistic lands. The median, $18, falls inside the $10 to $25 band, which at 33% is the single largest bucket in the dataset. The median is not just the middle value, it is also sitting on top of the mode. That is a strong signal: the center of the distribution and the most common outcome agree with each other, and both say the same thing about the market.

The mean, $178, falls inside the $100 or more band, which holds 14% of tools. So the "average" price is a price that at most one tool in seven charges to start, and that the other 86% undercut. When someone quotes you an "average SaaS price," ask whether they mean the median a normal buyer actually pays or the mean a spreadsheet produced.

Sixty percent of software is an expense-report decision

The distribution is easier to act on when you read it cumulatively, as a set of price ceilings rather than a set of bands.

Cumulative share of tools at or below each price ceiling

Six in ten tools start under $25 a month

Under $10/mo

27%

Under $25/mo

60%

Under $50/mo

77%

Under $100/mo

87%

Any price

100%

Six tools in ten start under $25 a month. That is the real center of gravity of the software market: cheap enough to expense without asking, priced for a single user or a small team to self-serve. Push the ceiling to $50 and you have captured 77% of everything we measured. Push it to $100 and you have 87%.

Those ceilings are not arbitrary. They map onto how buying decisions actually get made inside companies. Under roughly $25 a month, the purchase is an individual act: one person, one credit card, no approval, no procurement, no security questionnaire. Between $25 and $100 the decision usually acquires a manager. Above $100 it acquires a process. The distribution shows vendors clustering hard on the side of the line where the buyer can say yes alone, which is a statement about go-to-market as much as about value.

The steepest step in the cumulative curve is the one from 27% to 60%, the $10 to $25 band absorbing a third of the entire market in a fifteen-dollar-wide window. Nothing else in the distribution is that concentrated. A band that narrow holding that much mass implies vendors are copying each other's price points rather than deriving them from cost or willingness to pay, which is exactly what you would expect in a market where every competitor's pricing page is one click away.

We can go one level deeper, because the band is not evenly filled either. Among the 1,493 tools in our catalog whose cheapest paid tier lands between $10 and $25, a single price point takes 517 of them. Not a range, one number: $10. That is more than the next three price points put together ($15 with 177, $20 with 162, $19 with 148), and roughly a third of the entire band.

Inside the $10 to $25 band, one price point dominates

Tools starting at each exact price, among the 1,493 catalog tools whose cheapest paid tier falls in the band.

$10

517 tools

$15

177

$20

162

$19

148

$25

100

$12

98

The rest of the band clusters on the same handful of psychologically round or just-under-round figures: $25 (100 tools), $12 (98). Almost nobody prices at $13, $17 or $21. This is the clearest evidence in the dataset that pricing here is imitative rather than derived. A cost model or a willingness-to-pay study would produce a smooth spread of numbers. Copying the tool your buyer already pays for produces exactly this: a spike at $10 and a scattering of familiar anchors around it.

The free tier is a distribution channel, not a giveaway

The recurring worry that free tiers are disappearing does not hold up. Across the tools we measured, 51% offer a free tier and 54% offer a free trial. The majority of software now lets you use it, or at least test it, before any money changes hands.

How software lets you in the door

Free tier and free trial overlap, so these shares do not sum to 100

Offers a free trial

54%

Offers a free tier

51%

No free tier

49%

Two notes on reading that chart. The free tier and the free trial are not alternatives, they are independent attributes, and plenty of tools ship both, so the bars overlap and do not sum to 100. And the third bar is simply the complement of the second: 49% of tools have no permanently free plan, though many of those still offer a trial.

The interesting thing is how close 51% and 54% are. The free trial is marginally more common than the free tier, and the near-tie tells you the market has not settled the underlying question, which is whether it is worth carrying non-paying users forever. A trial is the cheaper answer. It caps the vendor's serving cost, it creates a deadline that does conversion work on its own, and it can be switched off without anyone feeling robbed. A permanent free tier is the more expensive answer, and it only pays when free users generate something beyond their own cost: collaborators invited into a shared workspace, content that ranks, an integration that makes the paid product stickier, or usage data that improves the product. Roughly half the market has decided that bet is worth making. The other half has not.

The economics behind both choices are the same, and they follow directly from the $18 median. At that price, a customer is worth so little in the first year that a human sales conversation cannot pay for itself. There is no version of a demo, a follow-up, and a negotiation that a product at the median can afford. Self-serve, free-to-start funnels are not generosity, they are the only distribution model the price point supports. Paywalling the front door only starts to make sense when a product is expensive enough to justify a sales motion, which is exactly the top 14%.

That reframes the free tier as a segmentation device. Vendors are not deciding whether to be generous. They are deciding which side of the price distribution they are on, and the entry path follows mechanically from that decision.

Three tiers won

Pricing-page design has converged harder than pricing itself. The distribution of tier counts is tight and unambiguous.

How many plans software actually publishes

Share of tools by number of published pricing tiers

2 tiers

19%

3 tiers

43%

4 tiers

25%

5 or more tiers

13%

The good-better-best structure won. 43% of tools use exactly three plans, and adding 4-tier tools, more than two-thirds land at three or four. Three tiers is enough to anchor a middle option, offer a cheap entry, and reserve an enterprise ceiling, without the paralysis of a five-column comparison table. If you are choosing how many plans to publish, the market has already run the experiment.

Each cluster tells you something different about the vendor. The 19% at two tiers are usually running the minimal shape: one self-serve plan and one "talk to us" plan, or a free plan plus a single paid one. It is the structure of a young product that has not yet found a reason to segment, or of a deliberately simple one that treats a short pricing page as a feature. The 43% at three tiers are running the canonical design, where the point of the third column is not to be bought but to make the second column look reasonable. The 25% at four tiers have usually bolted an enterprise column onto a working good-better-best, which is what growing upmarket looks like on a pricing page. The 13% at five or more are a mix of genuine complexity (usage-based products with several consumption shapes) and accumulated plan sprawl that nobody has been given permission to delete.

One honest caveat: our inclusion rule requires at least two published plans, so single-plan flat-rate products are excluded by construction. The 19% floor at two tiers is therefore a property of the sample, not a measurement of how many products publish exactly one price. Everything above two tiers is unaffected.

The market map

The clearest way to see the two-population structure is to plot the price bands against how they get customers. The map below places each starting-price segment on two axes. Both axes are relative positions we derived from our own catalog signals, not published scores and not measured per-segment rates. The horizontal axis orders the segments by starting price level, taken straight from the bands above. The vertical axis expresses free-to-start posture, derived from the report's central finding that free-to-start funnels are what cheap self-serve pricing requires and that the expensive top of the market is where the paywalled front door concentrates. Read positions as ordering, not as coordinates.

The market map: price level against free-to-start posture

Relative positions derived from our catalog signals, not published scores

Momentum

LeadersEmergingNicheEstablished

Free-first entry (under $10)

Cheap self-serve ($10 to $25)

Mid-market ($25 to $50)

Upper mid ($50 to $100)

Enterprise entry ($100+)

Market presence

The diagonal is the point. Price level and free-to-start posture move in opposite directions, because they are two expressions of a single decision about who does the selling. In the top left quadrant, the product sells itself and the price has to be low enough for that to work. In the bottom right, a person sells it and the price has to be high enough to pay for that person. The two corners are coherent businesses.

The interesting failure mode is the empty corner. A product priced like enterprise software with a self-serve, free-to-start front door has to convert an enormous volume of free users to cover its cost base, and a cheap product gated behind a sales call spends more acquiring each customer than the customer is worth. Vendors drift into those corners by accident, usually by raising prices without changing distribution, or by adding a sales team to a product whose price never moved.

We can show the direction of this relationship from the catalog's own structure, but not yet its magnitude.

We can now put a measured rate on that vertical axis rather than a derived position. Splitting the catalog by starting-price band and asking how many tools in each are classified freemium produces a curve that is not monotonic:

Freemium peaks in the middle of the price range, not at the bottom

Share of tools classified freemium, by the band their cheapest paid tier falls into.

Under $10

59.5%

$10 to $25

72.1%

$25 to $50

52.6%

$50 to $100

40.6%

$100+

38.9%

Freemium does not simply decline as price rises. It peaks at 72.1% in the $10 to $25 band, the same band that holds the median and the mode, then falls away on both sides: 59.5% below $10 and down to 38.9% above $100. The read is that free-to-start is strongest exactly where self-serve competition is fiercest. Below $10 a vendor has less to gain from giving away a tier, because the paid step is already trivial. Above $100 the sales motion pays for itself and the free tier stops being the cheapest way to acquire a customer. In the middle, where the market is thickest and every competitor's pricing page is one click away, giving the product away first is close to mandatory.

The full breakdown

Every distribution in this report, in one place. Cumulative shares are computed by us from the band shares and apply only to the mutually exclusive dimensions.

* Published price-band shares are rounded to whole percents and sum to 101, so the final cumulative figure is shown as 100.0% rather than the arithmetic total. Entry paths overlap (a tool can offer both a free tier and a free trial), so they do not accumulate.

What this means

For buyers: the $18 median is your reference point, not the vendor's enterprise quote. If a tool in a normal category is asking far more than $25 to start, there should be a clear reason (heavy infrastructure, deep compliance, real services), otherwise it is priced above the market. And you can almost always try before you buy, so a vendor that refuses to let you touch the product before a call is telling you which side of the distribution it is on. Use the cumulative curve as a negotiating frame: knowing that 77% of software starts under $50 changes how a $99 quote reads.

For anyone pricing a product: three tiers, a starting price in the $10 to $25 band, and a free tier or trial is not playing it safe, it is matching where the market already is. Deviating upward without a strong story means competing against a median you are ignoring. The more useful instruction is to pick a corner of the market map and stay in it. If you price in the top band, build the sales motion that price implies and stop apologizing for the absence of a free tier. If you price at the median, accept that free-to-start is not optional, because no other distribution model survives at $18 a month. The expensive mistakes happen in between.

For the "SaaS is getting expensive" narrative: at the entry level it is not. Most tools start cheap and let you try them free. The expense lives at the top, in the enterprise tier, which is a different market with different rules, and it is where the SSO tax and the biggest price hikes concentrate. The two populations in this dataset do not just have different prices, they have different physics, and conflating them is how the narrative gets written.

Frequently asked questions

What is the average price of software?

There is no single useful average. The median tool starts at $18 a month, but the mean is $178 because expensive enterprise tools skew it. For a normal buyer, the median is the honest number: most software starts under $25 a month.

Do most software tools have a free tier?

Yes, just. 51% of the tools we measured offer a free tier and 54% offer a free trial, so a slim majority let you use or test the product before paying. Free-to-start is now the default distribution model for affordable software, not a rarity.

How many pricing tiers should a SaaS have?

Three. 43% of tools use exactly three plans, and more than two-thirds use three or four. The good-better-best structure is the industry standard because it anchors a middle option without overwhelming the buyer.

Why is the gap between median and mean price so large?

Because software pricing is bimodal. Most tools are cheap self-serve products under $25 a month, while a smaller group of enterprise and infrastructure tools start at hundreds or thousands. The mean sits between two clusters and describes neither, which is why the median is the more honest figure.

Where this goes next

The structure we measured is stable in a way that pricing commentary rarely admits. Three tiers, a low-double-digit entry price, and a free way in are not a trend, they are the equilibrium a self-serve market settles into once every competitor's pricing page is public and one click away. Convergence that tight is a sign that vendors are pricing against each other, and that is unlikely to reverse.

The pressure will come from the top and from the unit of measurement. The 14% that sets the mean is the part of the market with the most room to move, and usage-based and consumption pricing sit uneasily inside a tidy three-column table, which is the most plausible source of change to the tier-count distribution we measured. What we would want next is per-category resolution, because a market-wide median hides the fact that a design tool and a data platform are not competing on the same curve.

We can answer part of that now. Restricting to categories with at least 60 priced tools and taking the median cheapest paid tier inside each, the market-wide figure clearly does hide a lot:

Two things stand out. First, the categories a business buyer actually shops in start well above the market-wide median: the software that touches revenue (CRM at $35, Lead Generation at $34, Sales at $30) is roughly twice the price of the market's midpoint, because the buyer has a revenue number to justify it against. Second, the freemium rate moves independently of price. App Builders sit near the bottom on price and top the table on freemium at 68.5%, while Inventory Management and Practice Management are expensive and largely paywalled at around 27% to 28%. Those are two different distribution strategies, not two points on one line: tools sold to a developer or a small team give the product away and let it spread, while tools sold into an operational workflow are bought, configured, and never touched by a casual user.

We will re-run this snapshot as the catalog grows and report what actually moved, rather than what everyone assumed had.