Should I self-host ?
One question, asked of 100 paid apps: is running it yourself actually cheaper once your own hours are on the invoice?
Three grades, defined once and applied identically to all hundred. The grade answers a single question: at your cost of time, does running it yourself come out ahead?
Self-host it. The alternative is credible, the migration is survivable, and the maths is not close.
Plausible · Uptime Kuma · Stirling PDF · Grist
Only if the box already exists. On a machine you run anyway the case is fine; standing one up for this is not.
Notion · Dropbox · Figma · Trello
Keep paying. The alternative exists and still loses once your hours and the cost of it being down are priced.
1Password · Copilot · Lightroom · Zoom
Vaultwarden is the strongest self-hosted alternative we have tested in any category — a faithful, Bitwarden-compatible server that boots in under a minute on hardware you already own. It is not the server that fails this rating.
It is the stakes. A password manager is the one service where downtime is not an inconvenience but a lockout, and where a botched restore is unrecoverable. Priced honestly — 20 minutes a month of patching, backups and restore drills at the $20/h reference rate — the all-in run rate is $7.17 against a $3.99 subscription. The math never breaks even, and that is before anything goes wrong at 3am.
1Password raising the individual plan above ~$7/mo. At today's premium the maintenance minutes alone keep this NOT REALLY; past the all-in run rate the rating flips on price.
The maintenance burden collapsing — a turnkey Vaultwarden appliance with tested, automatic off-site restore would remove most of the 20 min/mo that decides this rating.
Per-seat pricing pressure. The subscription multiplies with seats; the self-host cost does not. A per-seat move on the family tier reopens the math.
OUTLOOK — the direction we expect this rating to move within twelve months. Positive means the case for self-hosting is strengthening.
Every rating carries a review date. A rating older than twelve months is marked stale and excluded from the headline counts.
CannyfiderYESNOT ASSIGNED18.0×0.0 mo10 AUG—
LokaliseweblateYESNOT ASSIGNED10.0×0.1 mo10 AUG—
PandaDocdocumensoYESNOT ASSIGNED6.6×0.1 mo10 AUG—
ToolJettooljetYESNOT ASSIGNED6.6×0.2 mo10 AUG—
Adobe Acrobatstirling-pdfYESNOT ASSIGNED6.5×0.2 mo10 AUG—
JotformheyformYESNOT ASSIGNED5.4×0.4 mo10 AUG—
TypeformformbricksYESNOT ASSIGNED5.4×0.4 mo10 AUG—
PingdomgatusYESNOT ASSIGNED5.0×0.3 mo10 AUG—
ChatbasetypebotKINDANOT ASSIGNED4.5×0.2 mo10 AUG—
AmplitudecountlyYESNOT ASSIGNED4.5×0.3 mo10 AUG—
TolgeetolgeeYESNOT ASSIGNED4.5×0.2 mo10 AUG—
SurveyMonkeylimesurveyYESNOT ASSIGNED4.2×0.6 mo10 AUG—Our verdict is one opinion. What readers actually cancelled is another, and we publish both.
Comes online with privacy-preserving, self-hosted analytics — a YES verdict we intend to dogfood. Until then this column stays empty, because we don't simulate readers.
Same story — real numbers or none.
No reader votes yet — the ✋ on every rating page feeds this list.
Votes are self-reported, one per browser, and never touch the verdict. They are the only number on this site we do not control.
Raise labour to $60/hr — a realistic senior rate — and 31 of the 44 YES ratings fall outside the break-even window. Set your time to $0/hr — the homelab position — and 54 of the ratings we scored against you clear the bar. Every rating on this site is a function of a number we picked on your behalf, so we publish it, the data, and the code that computes it. Disagree with us with your own inputs — every page has the slider.
New rating actions, price hikes turned into break-even math, and the occasional honest “keep paying”. We email when something is worth your weekend — which is rarer than newsletters like to admit. Unsubscribe is one click.