The fear has no number attached to it
The case for tearing out your SaaS stack and having an AI build the replacement usually rests on two claims. The monthly fees add up. And they will keep going up.
The first is arithmetic anyone can do. The second is a feeling. Ask how much prices actually rise and the conversation goes quiet, because the current pricing page only shows today — there is nowhere to look it up.
Unless someone kept the older pages.
Nine repricings
Each was checked against the vendor's own pricing page as archived before and after the change, and each has its own write-up. Prices are only compared where currency, billing cycle and quoted period match.
| Product | What happened | Effective change |
|---|---|---|
| Oh Dear | Six plans became eight, all renamed | +78-98% monthly at the same site count |
| Plausible | Price held; the $9 tier went from 10 sites to 1 | +111% to keep 10 sites |
| Slite | Middle plan retired in favour of a $20 bundle | +60% for new customers |
| Postmark | Sticker price fell 73-87%; included volume fell further | +36-63% per 1,000 messages |
| Netlify | Free tier moved from units to credits | 100GB → ≈30GB of bandwidth |
| Notion | Business seat price | +33% |
| Linear | Basic seat, over four years | +25% |
| Healthchecks.io | Every plan's limit disappeared from the page | Not determinable |
| Obsidian | Sync | −50% |
Increases range from 25% to 111%. One product halved. One cannot be judged at all.
What the spread actually says
"SaaS prices go up" is true. Seven of the nine cost more in real terms than they did. But the range is wide enough that the sentence stops being useful. Linear's Basic seat rose 25% over four years — about 6% a year, which is roughly what everything else did over the same period. Oh Dear came close to doubling in a single month. Treating those as the same phenomenon, under the same word, is how the argument gets made without ever touching a number. And the price tag is unreliable. Postmark's headline fell by three quarters while the cost per message rose. Plausible never changed a digit and doubled what ten sites cost. Netlify swapped the units. Healthchecks.io simply stopped printing its limits, so nobody outside the company can say whether they still exist.Four of the nine move in a direction you cannot read off the price.
The other side of the ledger is missing, and that is the point
This table has one column too few. There is no equivalent for what building it yourself costs, because that side has no list price — which is precisely why it tends to get entered as zero.
It is not zero. Somebody watches whether it is up. Somebody is on call when it is not. Somebody patches the dependency with the advisory against it, follows the API when the vendor changes it, and inherits the whole thing when the person who wrote it leaves. Paying for SaaS is, among other things, outsourcing that list by the month.
Whether that is worth $20 a seat depends on your size and on who has the spare afternoon. This site cannot answer it.
What it can do is supply the half of the comparison that is knowable.
What to check before you sign
On the evidence above, the monthly figure is the least informative number on the page.
The three that mattered across all nine cases were whether the free or entry tier has picked up a limit it did not have before, whether the plan names have changed since you last looked, and whether the unit of measurement has changed.
None of those appear in the price column.
The table gets longer
This site currently saves 115 official pricing pages. The nine above are the first findings to come out of that.
A year from now there will be more, and the reason is dull: a price history can only be built forward from the day you start keeping it. There is no way to buy the missing years later.
Anyone deciding whether to buy or build needs the next version of this table more than they need another opinion.
Sources
Each row has its own article with the archived URLs and the verification behind it.
Every figure was read from the vendor's own pricing page as saved by the Internet Archive. Each article also states what could not be established from the archive.
Only published pricing and plan terms are used. No user reviews or forum posts.
Correction (2026-08-24): the Oh Dear row first read +95-123%, which compared two different billing cycles. Corrected to +78-98% billed monthly (+62-80% billed annually).