I priced fifteen project management tools. The winner changes four times.

Ask which project management tool is cheapest and you will get an answer.

Ask it again with one requirement attached and you get a different one, at eighteen times the price.

I spent a week reading pricing pages: sixty-two paid tiers across fifteen vendors, every price on both billing cadences, every seat cap, every seat minimum, and thirteen capabilities recorded tier by tier.

Then I built the thing that does the arithmetic, which is an Excel workbook rather than a dashboard, for reasons I will get to.

What came out of it is not a ranking. It is a claim about what actually decides which tier you end up on, and it is not the size of your team.

There is no cheapest vendor

Below six seats the cheapest paid tier in the whole set is Zoho Projects Premium at $4 a seat. From six seats up it is Basecamp, and Basecamp never loses again, because Basecamp does not charge per seat at all. Its $590 a year Studio plan covers unlimited people, which at a hundred of them works out to 49 cents each.

Basecamp's jump at twenty seats is real: that is where the Freelancer plan's seat cap runs out and the next plan up costs more per head than the one you just outgrew.

That is the whole shape of the problem. "Cheapest" is a function of team size, and the function has a crossing point in it. Any article that answers the question with a single vendor name has picked a team size without telling you.

Then you require one thing

Here is the part I did not expect. Take a hundred seats and start adding

requirements, one at a time.

what you require	cheapest at 100 seats	per seat	vendors left
nothing	Basecamp Studio	$0.49	14 of 15
a Gantt or timeline view	Zoho Projects Premium	$4.00	10
SAML single sign-on	Zoho Projects Enterprise	$9.00	5
single sign-on and an audit log	Runrun.it Enterprise	$26.00	1

The winner changes at every step. Basecamp cannot do a timeline, Zoho Projects can do single sign-on but not with an audit log, and Runrun.it is the only vendor left that prints a price for both.

Counting the crossover at six seats, that is four changes of winner in one market: Zoho Projects below six seats, Basecamp above it, Zoho Projects again the moment you need a timeline view, its Enterprise tier when you need single sign-on, and Runrun.it when you need an audit log alongside it. Same fifteen vendors, same day, five different answers.

Across all sixty-two tiers, exactly three carry both single sign-on and an audit log. Two of those are contact-sales. So the whole published-price market for "we need SSO and we need to know who did what" is one tier, from one vendor, at $26 a seat.

The thing forcing you onto a tier is usually not headcount. It is a security checkbox. And crossing it does not just raise your bill, it ejects you from the part of the market that will tell you its prices.

The sharpest version of this is a floor, not a price

A five-person startup that needs single sign-on, and wants to know what it costs before talking to anyone, pays Trello $875 a month.

Trello Enterprise is $17.50 a seat with a fifty-seat minimum. Five people buy fifty seats. The first forty-five are free in the sense that they cost nothing extra, and $17.50 each in the sense that you are paying for them.

Trello Enterprise is the cheapest priced tier in the set that carries SSO at any scale. It also has a fifty-seat minimum. So the question "what does single sign-on cost us" has no per-seat answer for a small team. It has a floor, and the floor is most of a developer's monthly salary in some markets.

This is the measurement I would not have found by reading the headline price on fifteen pricing pages, which is the entire reason the seat minimum is its own column in the dataset.

Where the prices are, and where they are not

I widened the sample to thirty in-category vendors to check whether hidden pricing is a quirk of the fifteen I looked at closely. It is not.

Only Adobe Workfront and Lytho publish nothing. The modal vendor publishes most of its ladder and goes quiet at the top.

Total blackout is rare. Selective opacity is the norm, and it is selective in a consistent direction: the hidden tier is almost always the one where single sign-on lives.

Four vendors go further and never name single sign-on anywhere on their pricing page, at any tier: monday.com, Jira, Basecamp and TeamGantt. A buyer reading those pages cannot tell whether the feature exists at any price.

I want to be careful here, because there is an innocent reading. Enterprise pricing genuinely is negotiated, and a published number would be fiction. But the pattern is not "we hide complicated things." It is "we hide the tier containing the feature that a certain kind of buyer cannot do without."

The unit you want to compare in does not exist

Fifty-four of the sixty-two tiers price per seat. Five charge a flat fee regardless of headcount. Three charge by project: TeamGantt is $120 a month for five projects and $24 for each one after that.

So a per-seat comparison table, which is what every roundup article publishes, requires inventing a seat price for eight tiers that do not have one. You can divide a flat fee by your team size, which is what I did for Basecamp above, but that number is not a price. It changes when you hire someone, without the vendor changing anything.

The workbook reports a blank rather than a number wherever the comparison would be meaningless. TeamGantt therefore does not appear in any of the per-seat results, which is why "14 of 15" is the vendor count with no requirements at all rather than fifteen.

Why this is a spreadsheet

The output of this project is not a chart. It is a question a specific person asks once: we are this many people, we need these things, what does it cost and who is left.

A dashboard answers that if I guessed the right filters. A spreadsheet answers it if I got the model right, and then the reader can disagree with the model. The workbook has 1,264 live formulas and four input cells. Set your seat count, your growth rate, your billing cadence and your requirements, and every tier is costed, ranked, and given a written reason when it is ruled out: outside the seat cap, no published price, does not meet a requirement, not priced on that cadence.

That last column is the one I would keep if I had to throw the rest away. "Nothing qualifies" is a bad answer. "Nine tiers were excluded because you asked for an audit log" is a finding.

Three ways this data can be wrong, all of which happened

Reading a pricing page with code fails in three distinct ways, and I hit all three before I caught any of them.

Concatenated prices. Pricing pages render both the monthly and the annual price into the page and hide one with a toggle. Text extraction returns them joined. Smartsheet Pro came back as $129 per Member/month. It is $12 monthly and $9 annual. What gave it away was the other currencies on the same page: $1713 AUD and $1612 CAD. Nobody prices like that. Every price in the final dataset was read from the rendered page instead.

Invented opacity. The same extraction reports no price at all when the prices are drawn by client-side script. Todoist came back as publishing nothing. Todoist publishes everything.

That one is worse than it looks, because the error runs in exactly the direction of the thing I was measuring. A method that silently converts "rendered in JavaScript" into "contact sales" will find a pricing transparency crisis in any market you point it at. Every vendor that looked like contact-sales got re-checked in a browser before it was recorded that way.

Silence read as absence. A capability missing from a pricing page is not necessarily missing from the product. This dataset treats an unclaimed capability as absent, which is the right default for a buyer and unfair to vendors with vague pages. It is also why not_stated is a distinct value from no in the capability data. Of sixty-two tiers, exactly one vendor, Teamwork.com, ever explicitly says a feature is not included.

The bug a clean recalculation did not catch

One more, because it is the kind that does not announce itself.

The workbook finds the cheapest qualifying tier per vendor with MINIFS. When no tier qualifies, MINIFS does not return an error. It returns zero. So two vendors showed a cost of $0.00 at team sizes where the honest answer was "nothing here fits," and the file recalculated with zero formula errors, because zero is a perfectly valid number.

I only found it by checking the whole grid against an independent Python calculation, cell for cell. A green recalculation proves your formulas evaluate. It says nothing about whether they are right.

What this is and is not

It is a snapshot. Collected October 2nd, 2026 from vendors' own pricing pages. Software pricing moves quarterly. Every row in the dataset carries the date it was read and the URL it was read from, so any single figure can be checked against the live page.

Ten of the thirteen capabilities are incomplete. Single sign-on and audit log were collected to completion because they decide the result. The rest are recorded where a pricing page happened to mention them, which is not a sample, it is whatever the vendors chose to write down.

Volume discounts are recorded but not applied. Trello Enterprise runs from $17.50 a seat at fifty users down to $7.70 at 4,405. Jira is flat to a hundred users and then drops. The model uses list price, so it overstates both at large team sizes.

No negotiated pricing, no nonprofit or education rates, US dollars only. Microsoft Planner is missing because its price lives inside Microsoft 365 licensing and has no standalone page, which is its own small finding about comparability.

Have a look

• Dataset: https://www.kaggle.com/datasets/chadwambles/project-management-software-pricing-october-2026

• Workbook, build script and method: https://github.com/chadwambles/project-management-pricing

The workbook is the part worth your time. If you think I have weighted the wrong things, or your team needs something I did not gate on, the inputs are four cells and the logic is visible in every formula. I would rather you argue with the model than take the number.