Skip to content

Build1 publisher3 min readPublished

Five hours of monthly maintenance ate a 40% cut in an eleven-tool SaaS bill

A five-person team swapped eleven paid subscriptions for self-hosted open source and came out close to level, because the cost that mattered was the maintainer's own hours.

The Engineer · Build desk

Illustration accompanying Five hours of monthly maintenance ate a 40% cut in an eleven-tool SaaS bill

What happened

  • A five-person team was paying $647 a month across eleven SaaS subscriptions, and every one of them had a free open-source equivalent available.
  • Replacing them cut the cash bill by 40%, while servers, backups, email delivery and AI API usage added $86 a month of new running cost.
  • About five hours of maintenance a month, priced at $50 an hour, absorbed most of what was left, leaving a reported net saving of $8 a month.
  • The migration ran three months, and one of the replaced tools was switched back to its paid version after five weeks.

Compiled by The EngineerSomething wrong?How this is made

Why it matters

  • cost The maintenance hours cost more than the servers do, so any team that prices its own time at zero will report a saving it never banked.
  • decision On-call coverage decides which categories stay bought: logins, passwords, payments and email delivery keep going to a vendor.
  • capability Metering AI by token only suits the users who do not want voice or memory, which splits one five-seat purchase into two decisions.
  • constraint One hungry service forces a second server, so the hosting line in these migrations is set by the worst-behaved app in the stack.

The published figures do not close. Forty per cent of $647 is $258.80 a month of cash saved [1][3][1]. Against that the post lists $86 of servers, backups, email delivery and AI API usage [4], plus about five hours of maintenance a month at $50 an hour, which is $250 [5][2]. So $336 goes out against $258.80 coming in, a loss of about $77 a month [3][4]. To land on the $8 the post reports [6], the cash bill would have had to fall by $344, or 53% [5].

The setup time is logged per swap, and it is the more useful column. The Notion replacement took 9 hours, most of it rebuilding table relations by hand after the export carried pages across but not linked databases [12]. Chatwoot took 7 hours and a server of its own, because it wanted more memory than everything else combined, and getting email replies to land in the shared inbox took an evening of settings [14]. n8n took 8 hours, a full day of that rebuilding 23 Zaps [13]. The scheduling swap took 3 [15], Open WebUI 3 [16]. Five of the eleven tools, 30 hours [7]. Priced at the author's own $50, that is $1,500 of one-time labour: about 5.8 months against the gross cash saving, and roughly 187 months against the $8 net [8][9].

The AI swap changes how the bill is metered. Open WebUI charges through an API key per use instead of per seat, and three light users came to about $30 a month between them [16]. That $30 comes out of the $86 of new running costs, leaving about $56 for servers, backups and email delivery [10]. The two heavy users missed voice mode, memory and the polish of the paid app [16].

The post's rule is about sleep. Its author wrote: "if a tool breaking at 3 am would wake you up, pay someone else to be awake" [8]. Earlier the same post puts the cost side plainly: "The software was free. The servers were cheap. I was the expensive part." [10]

For the $8 to transfer, two inputs have to hold: an hour of your own time worth $50, and eleven tools for five people, which is $129.40 per person per month [1][12]. Move the first input and the sign changes. At $20 an hour the same five hours cost $100, and the same 40% cut clears about $73 a month [11]. A team with someone who would run the upgrades on a Sunday anyway banks a real saving.

One check in the post is worth copying whatever your total comes to. Sort candidate projects by last commit: the author used alternatives pages that show each project's last update date [18], and found that two of the five Loom alternatives listed had no code update since November 2024 [17].

What to watch

  • A second year of figures, since the $8 rests on maintenance holding at five hours a month.
  • The identity of the tool reverted to paid after five weeks; the excerpt stops before naming it.
  • Whether the Loom alternatives dormant since November 2024 get commits, or drop off the recommendation lists still carrying them.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories