Self-Hosting Fatigue: When SaaS Is Actually the Better Choice

The self-hosting community rarely talks about the cost that doesn't appear on a bill. Every NAS and home server you run has a maintenance burden: updates that need testing, services that need monitoring, disks that need replacing, and configurations that need updating when software changes break your setup. This cost is real but invisible until it accumulates into what people in the self-hosting community recognise as 'self-hosting fatigue'. The point where maintaining the infrastructure feels like more work than the services are worth. This article does not argue against self-hosting. It argues for self-hosting selectively, which means being honest about which categories are worth the overhead and which are better left to SaaS providers.

What Self-Hosting Actually Costs You

The visible costs of self-hosting are easy to calculate: NAS hardware ($400-1,500), drives ($100-400), electricity ($40-100/year), and your internet connection. These compare favourably against cloud subscriptions for storage and software.

The invisible costs are harder to quantify but worth naming:

• Attention and decision overhead: Every service you run requires monitoring, even if that monitoring is just occasionally checking it still works. As the stack grows, the number of things that can fail grows with it. A stack of 15 services means 15 possible failure points, each with its own update cycle, dependency chain, and configuration surface.
• Update risk: Every update is a potential regression. Testing updates before applying them requires time; not testing them is a silent risk. Self-hosters who fall behind on updates create security exposure; those who stay current accept periodic breakage risk.
• Knowledge debt: Configurations built without documentation are a tax on your future self. The Nginx reverse proxy config you spent a weekend tuning in 2023 will need debugging in 2026 when it stops working, and you will have forgotten why it was configured that way.
• Single point of failure: When the NAS has a drive failure, the power supply dies, or DSM updates badly, everything goes down at once. A SaaS provider's outage affects their SLA, not your personal availability record.

None of these are arguments against self-hosting. They are arguments for being deliberate about what you self-host and realistic about the ongoing cost.

Categories Where Self-Hosting Clearly Wins

Some categories are good candidates for self-hosting because the SaaS alternative is expensive, the self-hosted option is mature and stable, and downtime is an inconvenience rather than a crisis:

• File storage and sync: Nextcloud, Synology Drive, and Seafile replace Google Drive or Dropbox storage (which cost $15-30 AUD/month for 2TB). The self-hosted option is better privacy, comparable functionality, and pays back hardware cost quickly. Synology Drive in particular requires very little maintenance for Synology NAS owners.
• Password management: Vaultwarden (Bitwarden-compatible) is one of the best self-hosting decisions available. 1Password costs $5 AUD/month per user. Vaultwarden runs on virtually any hardware, draws under 1W, and is extremely stable. Downtime means temporarily using cached passwords. Not a crisis.
• Photo management: Immich has become a strong Google Photos replacement with face recognition, album sharing, and a polished mobile app for uploads. Google Photos at 2TB costs $15 AUD/month. The self-hosted option handles large libraries well on a NAS with sufficient RAM.
• Ad blocking: Pi-hole or AdGuard Home run on almost any hardware. The SaaS alternative does not exist in a meaningful form. Network-level ad blocking is a self-hosting-only feature.
• Home automation: Home Assistant is the definitive home automation platform and has no credible commercial equivalent. Cloud-based home automation products are fragmented, have spotty longevity, and often sunset without warning. Self-hosting HA is nearly always the right choice for serious home automation.
• Note sync: Joplin, Obsidian, and Standard Notes with self-hosted sync replace Evernote ($130/year) or Notion ($10-20 USD/month) with privacy-first alternatives that work well for personal use.

Categories Where It Gets Complicated

Some categories are technically self-hostable but come with friction that makes the comparison less clear-cut:

• Email: Self-hosted email (Mailcow, Mail-in-a-Box, Mailu) is technically achievable but involves managing IP reputation, deliverability, spam filtering, and staying off blacklists. Most residential IP addresses are on dynamic IP ranges that are pre-blocked by major providers. A dedicated VPS helps but adds ongoing cost. For personal use where deliverability matters, Fastmail or Proton Mail is a better value proposition than self-hosted email for most people.
• Collaborative document editing: Nextcloud Office (Collabora Online) and OnlyOffice are functional alternatives to Google Docs but have notable performance and compatibility gaps. For occasional use, they are fine. For a household where someone regularly collaborates with external colleagues who use Google Docs or Microsoft 365, the format compatibility and real-time collaboration performance make SaaS the more practical choice.
• Analytics: Plausible, Umami, and Matomo self-hosted are good options for website analytics with privacy-by-default. The self-hosted version saves $10-20 USD/month versus hosted plans. But they require a server with good uptime. If the analytics container is down when traffic spikes, you lose that data. Borderline: worth it if you already have a VPS or reliable NAS with good uptime.
• Video conferencing: Jitsi Meet is self-hostable but WebRTC performance is highly sensitive to network configuration, NAT traversal, and TURN server availability. Unless you have a static IP, a properly configured TURN server, and good upstream bandwidth, self-hosted video conferencing is unreliable compared to Zoom or Google Meet. The convenience gap is large enough that most households should not bother.

Where SaaS Is Genuinely the Better Choice

Some categories are better left to SaaS, not because self-hosting is impossible but because the cost-benefit genuinely favours the commercial service:

• Calendar and contacts sync (CalDAV/CardDAV): The protocols work, but the maintenance overhead of running a CalDAV/CardDAV server (Radicale, Baikal, Nextcloud) for the privacy benefit over iCloud, Google Calendar, or Fastmail is rarely worth it. These services cost $0-5/month and are deeply integrated with mobile operating systems. A calendar sync outage affects daily life immediately. Proton Mail with its calendar provides a reasonable privacy-first alternative without self-hosting if that is the concern.
• Transactional email delivery: If you run a small business or personal project that sends email (form submissions, notifications, invoices), self-hosted SMTP delivery is unreliable from residential or even standard VPS IPs. Postmark, SendGrid, or Resend have free tiers sufficient for most small-scale needs and handle deliverability, DKIM, and DMARC automatically.
• Database backups for critical business data: Off-site backup of critical business data using a verified, compliant cloud provider gives you audit trail, geographic redundancy, and SLA guarantees that a home NAS setup cannot match. Self-hosting is appropriate for the backup destination of personal data; it is a liability for business-critical compliance requirements.
• DNS: Self-hosted recursive resolvers (Unbound, NextDNS) make sense for privacy. Self-hosting your authoritative DNS for public domains does not. The reliability requirements are too high, and Cloudflare's free DNS is more reliable than anything a home user can operate.

A Framework for Deciding What to Self-Host

When evaluating whether to self-host a new service, four questions cut through most of the noise:

Managing Fatigue: The Sustainable Self-Hosting Stack

Self-hosting fatigue typically comes not from any single service but from stack sprawl. Adding services faster than the maintenance overhead can be absorbed. A sustainable self-hosted stack has a few characteristics:

• Everything is in Docker Compose: Services deployed as Docker Compose stacks with docker compose pull && docker compose up -d can be updated in minutes. Services deployed manually, via package managers, or with custom init scripts require custom update procedures that you will forget by the time the next update is needed.
• One reverse proxy handles all TLS: Traefik or Nginx Proxy Manager terminates HTTPS for all services through one configuration. Maintaining separate TLS certificates per service does not scale.
• A limit on the number of services: Experienced self-hosters often converge on 8-12 core services and stop. Each additional service beyond that adds diminishing returns. The ten services you actually use every day deliver more value than twenty services that seemed interesting when you deployed them.
• Monitoring that alerts you: Uptime Kuma or Healthchecks.io watching your services means you find out about failures before the family does. Without monitoring, outages become support incidents rather than infrastructure events.
• Documented configuration: A brief README or comments in your Docker Compose files explaining non-obvious configuration choices costs 10 minutes when you set something up and saves hours when you need to debug it six months later.

Related reading: our NAS buyer's guide.