Five SaaS Products Built Solo: What I'd Do Again and What I Wouldn't

Four of these five products I built alone. On the fifth I was a collaborator on a product somebody else owns, and rounding that up to "solo" would poison every other number on this page.

If you're weighing whether to build your own SaaS as a solo developer, here is the shortest useful answer from my records: scope and a named buyer decided the outcome far more than build speed did. Below are the figures read back out of my project records rather than remembered: BookBed, Callidus OS, Pizzeria Bestek, an eight-template FlutterFlow Marketplace portfolio, and Relocate. Pick the one closest to what you're building and read its row first.

How long does it take to build a SaaS MVP alone?

About ten weeks for a multi-tenant SaaS with billing and compliance, if the scope is settled before you start and nobody moves it. Callidus is my datapoint; BookBed took six months.

Callidus had unusually clean conditions: one decision-maker, and a spec the failed FlutterFlow attempt had already produced. It inherited about two hundred unresolved errors from that attempt and was rebuilt from scratch. The record lists roughly 1,887 commits, about 426 KB gzipped across 37 lazy routes, 100+ Cloud Functions and around 1,200+ tests across Vitest and Playwright. BookBed had no external deadline, and over six months it grew an embeddable widget, an admin surface and six platforms. My inference: the variable is not typing speed, it's whether anyone is allowed to add to the list once the build starts. I cover the scoping side in the SaaS MVP development guide.

What does a solo founder SaaS case study usually leave out?

Revenue. Build timelines are easy to publish and recurring revenue is not, so the honest version of a solo developer SaaS story is a distribution, not a headline.

Put mine next to the market. Freemius's 2025 State of Micro-SaaS report says 45.7% of the SaaS makers it surveyed are solo founders, that about 70% of micro-SaaS businesses earn under $1,000 MRR, and that the median profitable one makes about $4.2K MRR. It measures its own platform, so treat it as one dataset rather than the whole market.

My five products show the same shape at a small scale, with one caveat: only one of them reports recurring revenue, and it's the client's. BookBed has no paid tier active. Pizzeria Bestek was a fixed-price build. Both paid FlutterFlow templates display “10+ sales” as of October 1, 2026; their exact combined total is not public. Relocate belongs to its client.

What the record leaves out matters as much. It has no monthly active user counts for BookBed, no Callidus revenue figure beyond the first paying clinic going live on 18 May 2026, and no conversion data for the templates beyond downloads and sales. The record doesn't show that I've run paid traffic through a funnel either, so I can't tell you what a trial-to-paid rate looks like for any of these. If a retrospective hands you those numbers without a source, ask where they came from. Mine come from the case studies, and where they're missing I've said so.

This suggests a lesson I'd defend but can't prove from five samples: the products with a named person waiting for them are the ones that reached revenue. Callidus had a founder waiting. Pizzeria Bestek had an owner who wanted to leave an ordering platform that took fees. BookBed was built for myself.

Which tech decisions would I unmake, and which paid off?

Each regret below is documented in a case study, which is the only reason I trust it. Where a product has no documented regret I say so, because a short honest list beats five invented ones.

Regrets on record

• BookBed: the date normalizer. An early overbooking detector normalized dates with setUTCHours(0,0,0,0) plus toISOString(). Bookings stored late in the UTC evening came back as the wrong day, which triggered false-positive overbooking warnings on bookings that didn't actually overlap. The rewrite uses toLocaleDateString('en-CA', { timeZone: 'Europe/Zagreb' }), iterating at noon UTC to stay DST-safe, with a save_trimmed action and eight tests covering real Adriagate imports.
• Callidus: none documented on my side. The costly decision, attempting the platform in FlutterFlow first, was made before I joined.
• Pizzeria Bestek: none documented. I won't invent one to fill the list.
• FlutterFlow: designing for one widget at a time. After the first three templates it became clear that buyers tend to need a small cluster at once, such as a chart plus a chip input. The April 2026 batch was designed around thematic clusters. In hindsight I'd have started that way, though that's my judgment, not a measured result.
• Relocate: a Unicode codepoint. The RevenueCat entitlement ID was Relocate․app Pro, copied from a designed mockup. That dot is U+2024 ONE DOT LEADER, not an ASCII period, so the SDK returned active: false in production. It was renamed to plain pro.

What paid for itself

Flutter for BookBed is the clearest one. One codebase reached six platforms, and the timeline calendar keeps fixed cell dimensions so drag math doesn't change per platform. The bet also got safer without any work from me: at Google I/O 2026, Canonical became lead maintainer and strategic steward of Flutter desktop for Windows, macOS and Linux, per OMG! Ubuntu's report.

For Callidus, the case study documents Stripe Connect Standard with direct charges: each clinic is the merchant of record for the medical service it performs. Stripe's direct charges documentation says the payment appears as a charge on the connected account, not the platform's, and the platform's balance grows through application fees, which for Callidus is 0.1%. The multi-tenant design pays off the same way, and I wrote up the rules side in how Firestore rules held up under six roles and fourteen tenants.

Pizzeria Bestek paid off through a workflow choice. Three Supabase Edge Functions run the admin loop: an order arrives, two emails go out, the owner clicks CONFIRM or DECLINE in the inbox, and the customer gets the result in one of four languages. It has run at five to ten orders a day since September 2025 on infrastructure the restaurant owns, and each order is 100% revenue to them instead of about 70% after aggregator commission.

Checked October 1, 2026, the public FlutterFlow profile displays “1.8K” downloads (rounded) and four positive reviews for PDF Viewer. Dream Home at $75 and Schedule / Advanced Calendar at $50 each display “10+ sales”; the public labels do not disclose exact paid totals. The case study describes the free listing as the credibility anchor the paid templates point back to.

On Relocate, Riverpod 3 with code generation carries a 38,711-line codebase across 27 screens, with 30 providers and 40 Freezed models, and the client continues development on it.

Does AI tooling make a solo developer as fast as a team?

It removes the typing bottleneck and leaves the verification bottleneck where it was, which is on you. The 2025 Stack Overflow Developer Survey reports 84% of respondents using or planning to use AI tools, up from 76% in 2024, and 51% of professional developers using them daily. It also reports that more developers distrust the accuracy of AI tools (46%) than trust it (33%).

That gap is the working condition. On Callidus it shows up as 100% Firestore rules coverage and 1,200+ tests, because unverified generated code is a liability in a GDPR-regulated system. Faster drafts, same review burden.

Attention is the input that doesn't scale. Every hour on a template batch was an hour not spent on BookBed's billing, which is wired, tested in Stripe LIVE and priced at €9 per month for up to twenty units, but not yet activated for paid traffic. The engineering is done. What's left is the part of shipping that isn't engineering.

So here is the question for you: which of your products is 90% built with the billing switch off? Pick the one with a named person waiting for it, and write down today the date you'll switch billing on.