What 30 SaaS Pricing Pages Actually Say (2026): The Seat, Usage, and Free-Tier Patterns Across the Tools Small Teams Buy

As of 28 September 2026, the thirty pricing pages in this dataset sell thirty different units — and only about a dozen of them sell what the buyer thinks they are buying.

That is the whole article in one line, dated. What follows is the document behind it: 30 SaaS pricing pages across the five tool categories this site covers (CRM, developer, marketing, finance, HR), checked between 13 November 2025 and 28 September 2026, read as one dataset. Four patterns survive the cross-category pass: how the seat model actually locks, what the usage unit actually is, what the free tier actually caps, and which lines are billed separately from the sticker. Every number carries its check date, because the pricing page is the spec sheet, and a spec sheet without a date is a rumor.

The dataset, and how to read it

The set: 30 pricing pages, five categories — 7 CRM, 7 developer, 6 marketing, 6 finance, 4 HR. The category split follows the site’s sections; the head-to-head section is where this article lives, because it is the comparison that does not name two vendors.

The pages, with check dates:

Three rules I applied, so the tables below mean what they say:

1. The date is the check date, not the publication date. “Checked 28 September 2026” means I looked at that page on that date and recorded what it said. If the page I could reach was an archived capture, the table says so.
2. Quote-gated pages are marked, not guessed. Rippling and Personio publish no rates on their pricing pages — they sell the quote. Those rows carry structure, not dollars, and they are labeled unverified.
3. The locale is part of the data. The Pipedrive page I captured showed Canadian pricing (27 August 2026), the Zoho page showed INR (5 August 2026). Where a US list price was separately verified, both appear; where it was not, the currency is named.

What “the fine print” means here: every line on or near the pricing page that is not the sticker — the seat rule, the unit definition, the free-tier cap, the storage line, the integration line, the overage rate, the billing-cycle spread. The sticker is one line of a longer document. This article is the longer document, and it re-renders when the vendors change it.

Pattern 1 — The seat model, and what the entry tier locks

First, the seat. Price per seat is a rumor until you count the seats — and in this dataset, “seat” is a word with four different meanings, by category:

Four observations the cross-category view makes visible:

The seat is never the whole bill; it is the key. In developer tools the seat unlocks the allowance (GitLab Premium’s $29 buys the 10,000-minute meter; without it you are on 400). In HR the seat is the unit and the feature gate at once — BambooHR’s $10 Core line includes time-off tracking and excludes 360-degree reviews; the feature move between $10 and $17 is the product. In CRM the seat is paired with a data cap: monday’s Basic tier (US list $12 annual, Oct 2025) carries 1,000 contacts and 5-column boards; the cap, not the seat, is what the buyer hits first.

“Per employee” is stricter than “per user.” BambooHR prices per employee with volume discounts; Deputy prices per user and counts every administrator and supervisor in the number (3 September 2026); Personio’s license rule is that all active users count toward the license total (17 April 2026). Rippling’s FAQ states the structure — per employee per month plus, on some products, a monthly base fee, with the core platform required (24 September 2026) — and leaves the rates to the quote, unverified here. The buyer who counts only the people who type in the tool has undercounted on all four of these pages.

The finance split is structural, not cosmetic. QuickBooks prices the tier with the seats inside it (1/3/5/25 per tier, as of 26 September 2026) and the accountant rides along as included seats. Xero prices the organization and does not price the user at all. Same category, two unit systems, and the QuickBooks vs Xero piece in this wave prices the difference over 36 months — $5,040 against $1,980 at the comparable tier, per the 26-27 September 2026 check. The pattern is the vendor’s choice of unit; the delta is what the unit does to a bill.

The entry tier locks behavior, not features. Across the seven CRM pages, the entry paid tier is where the caps live: 2 pipelines (HubSpot Starter, Sep 2026), 1,000 contacts (monday Basic, 5 Aug 2026), 1,000 contacts and companies (Agile Starter, 16 May 2026). The Best CRM matrix runs the same five-seat arithmetic across eight tools; the numbers there are the cited examples, and this table inherits them rather than repeating them.

Pattern 2 — The usage model, and what the unit actually is

Second, the unit. The usage-based pages in this dataset sell at least seven different things, and several use the same word for different things. The unit-definition map:

Read the map twice, because it is doing two jobs.

Job one: the same word is not the same unit. “Minutes” on GitHub is a wall-clock job minute on a named machine; “compute minutes” on GitLab is a formula (duration × cost factor, up to 12× for the largest class); “credits” on CircleCI is a prepaid abstraction that the vendor converts to minutes at 10:1 on one class and to GB at 420:1 for storage. A 30-minute job on a GitLab medium-class runner is 60 compute minutes; the same job on GitHub is 30 minutes; the same job on CircleCI is 300 credits. The CI piece in this wave reconciles those three at the 2 vCPU point and finds the per-minute prices converged at $0.006 as of 27 September 2026 — the unit definitions are the story, and the convergence is the finding.

Job two: the meter is chosen to be hard to predict. Sentry’s overage is tiered — $0.0003625 per error at 50K–100K falling to $0.00015 per error past 20 million (28 September 2026) — so the marginal price drops as the bill grows, which is the opposite of the usual surprise. Datadog prices the same host three ways by billing type ($18.75 annual, $22.50 monthly, $22.50 on-demand) and prices retention windows as separate rates on indexed events ($1.325 per million at 3-day retention, $4.6875 at 30-day, per the 28 September 2026 rate list). Grafana splits one GB of logs into three billable steps — process $0.05, write $0.40, retain $0.10 — and its own page currently carries the banner “Send more, charge more is broken. We are changing it.” (28 September 2026). A vendor putting the churn of its own unit on the page is a dated fact, and it is the first entry on the watchlist below.

The marketing split: contacts versus subscribers. Mailchimp and ActiveCampaign price the contact and meter the sends as a multiple of it (10× to 15× by tier); Kit and BeeHiiv price the subscriber and do not meter sends at all — unlimited on every tier (28 September and 23 September 2026 respectively). Two unit systems in one category, and the per-recipient meter has retreated to the transactional side (SendGrid, Brevo). If your bill is “marketing email,” know which of the two systems your vendor is in, because they compound differently: the multiple model punishes a list that outgrows the send rate; the subscriber model punishes growth itself, with no send line at all.

Pattern 3 — The free tier: what it actually caps

Third, the free tier. The cross-category cap table, as checked:

Three shapes cover every row.

Shape one: the tier sized to be outgrown. Mailchimp’s 250 contacts, QuickBooks’ 2 invoices, GitLab’s 400 compute minutes, GitHub’s 2,000 minutes. These caps are not generosity; they are the distance between the free tier and the buyer’s real workload, expressed in units. A five-person sales team does not run on 2,000 CI minutes; a real bookkeeping workload does not run on 2 invoices a month. The cap is the product — the free tier is the meter set to the buyer’s growth.

Shape two: the trial wearing a free costume. The CRM pages (14- and 21-day windows), SendGrid’s 60 days at 100 emails a day, FreshBooks’ 30 days. These are demonstrations of the paid tier with an expiry date, and the expiry is the fine print: when the trial ends, the data moves and the pricing changes in one motion. The CRM matrix notes that Freshsales’ 3-user free plan, present on a 29 October 2025 table, is gone from the 3 September 2026 page — a free tier can be retired inside a quarter, which is why the check date matters.

Shape three: free software, metered elsewhere. Wave’s free accounting with the payment-transaction meter (2.9% + $0.60), Ramp’s free card platform with the team-size platform fee on the paid tier, Deputy’s per-user plans with the SMS meter and the $30 minimum. The free line is the hook; the usage line is the bill. And the outlier that breaks the pattern: Agile’s free tier holding 10,000 contacts against a paid Starter tier capped at 1,000 (16 May 2026) — the free tier holding ten times the data of the paid one is an inversion, and the inversion is what to re-check at renewal, because it is the kind of line vendors fix quietly.

The free tier is a marketing tier, not a discount. It is priced at $0 and engineered at the cap; the cap is where the buyer’s workload is, six months from now.

Pattern 4 — The hidden line: what the sticker does not say

Fourth, the lines that are not on the sticker. Counted across the 30 pages, three “included vs billed separately” lines recur, and they account for most of the gap between a pricing page and an invoice.

1. Storage and retention — 5 of 30 pages carry an explicit line. Grafana prices retention as a distinct step from ingestion ($0.10/GB retain, separate from $0.40/GB write, 28 September 2026), and the free tier’s 14-day retention against the Pro tier’s 13-month metrics retention is the feature the retention price unlocks. Sentry includes 5 GB of logs and charges $0.50/GB beyond it (28 September 2026), and prices the lookback window (30 days on Developer, 90 on paid) as a plan line. Vercel includes 1 GB of blob storage on Pro then bills $0.023/GB (28 September 2026), and runtime logs are 1 hour on Hobby, 1 day on Pro. CircleCI meters network/storage overage in credits (420 per GB, 27 September 2026). Datadog’s indexed-event prices are retention-graded across the whole $1.325–$4.6875 per-million range (28 September 2026). The pattern: retention is the second meter, and it is priced per window, not per byte, so the bill grows with how long you keep things, not only with how much you send.

2. Integrations and connections — 5 of 30. QuickBooks’ Free plan connects 1 bank account and unlimited connections is a paid-plan feature on the same page (26 September 2026) — the connection is a paid-tier line. Xero includes feeds but keeps a usage line for chargeable direct bank feeds, and the UK page passes bank fees through (27 September 2026 and 25 September 2026 respectively — the QBO/Xero piece runs those lines line by line). Ramp gates the accounting integration by tier — QuickBooks/Xero on Free, NetSuite/Sage Intacct on Plus, Workday on Enterprise (28 September 2026). Rippling’s structure is the integration as the subscription: a required platform base fee plus per-product lines (24 September 2026, unverified rates). Vercel includes 2 microfrontend projects then bills $250 per additional one (28 September 2026). The pattern: the connection is the product the buyer already has; the vendor sells it back as a line.

3. Overage beyond the cap — 11 of 30. The CI trio (GitHub $0.006/min, GitLab $0.01/min, CircleCI $15 per 25,000 credits — 27 September 2026), Sentry’s tiered per-error overage, Vercel’s invocation and build-minute rates, Mailchimp’s send pause at the cap (overage as suspension, 28 September 2026), Brevo’s pay-as-you-go credits, ActiveCampaign’s send multiple as a built-in overage, Wave’s 2.9% + $0.60 per transaction, Deputy’s SMS line and $30 minimum, Gusto’s per-payroll-run lines (instant pay $100, same-day pay $90 per payroll, 13 November 2025). The pattern: the overage rate is published, but it is published on a different page from the sticker — the billing docs, the rate list, the FAQ — and it is the line that appears on the invoice but not on the page the buyer screenshotted.

Two meta-lines sit above all three:

The price behind the slider. Brevo’s and ActiveCampaign’s pages define plans by volume sliders — the static page of 28 September 2026 (Brevo) and of 1 September 2026 (ActiveCampaign) shows the tier structure, the feature deltas, and a “per month” with no number next to it. The number is there if you drag the slider; it is not there in the page. For a dated dataset, that is a finding, not a gap: two of six marketing pages keep the price in the interaction.

The billing-cycle spread, inside the sticker. HubSpot Starter is $7 a seat on annual commitment and $20 on monthly — the same plan, nearly a 3× spread (September 2026). Datadog’s rate list carries three prices for the same host SKU by billing type ($18.75 / $22.50 / $22.50, 28 September 2026). Kit states “pay yearly — save 17%” on both paid tiers (28 September 2026); BeeHiiv lists the annual-equivalent price — $43/mo billed as $517/yr, “save $71” — with the standalone monthly rate implied rather than stated on the captured page (23 September 2026); FreshBooks’ Lite runs $1 a month for the first year against $23 list (28 September 2026). The cycle is a multiplier, and it is the line that re-prices a renewal with no visible change to the page.

And the sticker itself can move under the reader: QuickBooks’ mid-tier list prices sat at $75/$115 in a May 2026 review and $85/$140 on the 26 September 2026 page — $10-$25 in four months, as the QBO/Xero piece documents. The hidden line is sometimes the sticker.

The buyer’s read: four patterns, five minutes

So: how does SaaS pricing work, as a method? The dataset compresses to a five-step read that works on any page in the set — and, because the patterns are cross-category, on any page outside it:

1. Name the unit before the price. What is a “credit” here? What is a “minute”? What is a “seat” — an employee, a user, an admin who also counts? Get the definition from the page or the billing docs. If the page keeps the price behind a volume slider (Brevo, ActiveCampaign), write down the slider and the bracket; that is the unit’s container.
2. Find the entry tier’s lock. Not the feature list — the cap table: contacts, users, minutes, GB, pipelines, invoices. In this dataset the entry tier’s lock is a pipeline count, a contact count, a minute allowance, or a client count. The cap is the product; the seat is just the key.
3. Hunt the three hidden lines. Storage/retention, integrations/connections, overage. If the pricing page does not name them, they live in the billing docs or the FAQ — the invoice is where they show up, so the invoice is the better page to read. On the 30 pages here, all three lines appear on at least 5 pages each; the typical bill carries two of them.
4. Check the billing cycle. Annual versus monthly versus on-demand — the same SKU at different prices (Datadog’s three columns, HubSpot’s $7/$20), and the promo that expires (FreshBooks’ first-year $1, Gusto’s six-month $0 base). The renewal invoice is the real pricing page.
5. Date it, then diff it. Every number above carries a check date because the page is a moving document. At renewal, re-run steps 1-4 on the page you already bought: the diff between the page you signed and the page in front of you is the number nobody announced. Check the diff, not the ad.

That read answers the three questions this section gets asked. “How does SaaS pricing work” — it works as four patterns: the seat lock, the unit definition, the free-tier cap, the hidden line. “Per seat vs usage-based” — the categories are split: CRM and HR price the seat (with HR counting every user), developer and monitoring price the unit (with the unit defined per vendor), finance is split by design (per-user tiers versus per-organization), and marketing is split between contacts and subscribers. “Why is my bill higher than expected” — the overage and the billing cycle: the two lines that are published, on a different page, in a different column, and they are the ones the invoice collects.

The dated watchlist: what shifts in 2026-27

What to watch before your next renewal — each with the date it was last true:

1. Grafana’s usage-pricing rework. The banner on its pricing page, 28 September 2026: “Send more, charge more is broken. We are changing it.” A vendor stating the churn of its own unit in its own page is the loudest signal in the dataset; when the new model lands, the process/write/retain split and the 10,000-series free allowance are the lines to re-check first.
2. CI repricing mid-cycle. GitHub cut hosted rates up to 39% effective 1 January 2026 and postponed the $0.002/min self-hosted platform charge on 17 December 2025 — still free in the 27 September 2026 billing docs, with no replacement date published. The CI piece carries the dated diff; re-run it when the platform charge lands.
3. Increases announced without figures. Xero US states its subscription prices increase from 1 October 2026, with the new numbers not yet on the 27 September 2026 page; QuickBooks’ mid-tiers moved $10-$25 inside four months of 2026. An unfigured increase is an open line on every renewal that touches October 2026.
4. Pricing pages that stopped pricing. SendGrid’s page is now a Twilio-migration notice (28 September 2026) with the last dated table archived 30 May 2025; Gusto’s page moved from /pricing to /product/pricing (a 308 redirect, captured 17 September 2026), with the last dated rates from 13 November 2025. A repurposed or relocated pricing page is a re-pricing in progress — treat its last dated numbers as a floor, not a forecast.
5. The AI line, metered before it is tiered. Sentry’s Seer debugger is “subscription required” on top of a $26 plan (28 September 2026); BeeHiiv steps AI credits 10/25/50 per day by tier (23 September 2026); Mailchimp lists generative AI as a no-additional-cost add-on on Standard and Premium but not on Essentials or Free (28 September 2026); Datadog sells AI credits at $500 per 500 credits per month (28 September 2026). The pattern forming across categories: the AI feature arrives as a metered add-on, then migrates into the tiers — the migration is the next repricing, and it has not happened yet.

The pattern is the product; the vendor is one instance of it. These four patterns — the seat lock, the unit definition, the free-tier cap, the hidden line — are what the 30 pages actually say, and they will outlive any single table in this article. When the pages change, the dataset re-renders; the patterns are the part that stays.

Read the page as a spec sheet: unit, cap, hidden line, date. And at renewal, run the read on the page you already have, because the check is the diff, not the ad.

Related on this site

• GitHub Actions vs GitLab CI vs CircleCI — the per-minute and per-runner fine print dated to the December 2025 change; the developer-tools rows of this dataset, priced on one workload.
• QuickBooks vs Xero — the per-user-versus-per-organization split of Pattern 1, priced over 36 months on a named workload.
• Best CRM for Small Business — the seat model of Pattern 1 across eight CRMs, with the entry-tier caps and the counted five-seat cost.