Not Every Startup Should Be a SaaS Business

"Of course all our revenue is SaaS. That's the only way to build a scalable startup."

These words are music to an investor's ears. We've all internalised the idea that without recurring revenue and 80%+ gross margins, a business isn't venture-backable. Add triple-digit YoY growth, and you're golden, right?

I'll admit, I've been there. It's easy to get excited about a startup, assume its revenue is SaaS, and then feel a pang of disappointment when you realise it’s actually services revenue instead. Higher margins, the ability to land sticky, land-and-expand contracts and what appears to be a tech moat are appealing.

However, we live in a world where choosing the right business model for your buyer matters, one where the cost of supplying AI insights requires usage-based rather than per-seat pricing. The unloved world of “services”, now enhanced with AI-driven efficiencies, shows an ability to scale and enable models not previously possible. These are attractive business models, and we need to embrace them rather than force-fit a SaaS model. 

Let's dive into three reasons why you shouldn’t try to make non-SaaS startups fit the SaaS mold and why it isn’t always as compelling or necessary as we think.

REASON 1: Non-Recurring Revenue Can Scale Faster

The best part of SaaS is the recurring revenue that you can depend on. You sell a contract, then collect a monthly subscription, based on a tier of service or a number of seats. Consumers and businesses alike have been trained to this model, paying for music with a monthly $9.99 payment or covering a business's cybersecurity with an annual payment. It is shockingly sticky, and as you add new features, seats, locations, or team members, every renewal offers an opportunity to increase revenue over time. It's the reason why we have the layered cake chart, and when it works, we see the segments and cohorts grow so well over time.

The concept of a 100%+ retention number, built off the fact that a cohort of customers will spend more over time through upsell to offset the cost of those churning – look at the bright blue above – feels an incredible feat. This is why such revenue commands high multiples. It’s like planting a seed and seeing fruit delivered for years to come. We can easily sum up these recurring payments and create a view of monthly recurring revenue (MRR) or annual recurring revenue (ARR). Due to this, many startups feel a temptation or investor pressure to call revenue as "recurring". This revenue may be repeating, but it isn’t recurring unless there is a clear subscription with a contract. 

Repeating revenue isn't recurring revenue, but it can scale

If you're selling a service to a business—like a tech-enabled cleaning offering that combines robots and human staff—you may want to stick with familiar pricing models: by FTEs, square footage, or overall service volume. While the idea of selling a subscription (perhaps "Robots as a Service") might be appealing, many buyers aren’t used to that model, and it could slow down your sales cycle.

Similarly, if you're an adtech solution selling tokens that give martech platforms, agencies, or their clients access to proprietary AI models for customer targeting, you'll likely charge based on usage—specifically the number of model calls and possibly the task complexity. Demand is often predictable, allowing you to forecast revenue, but it's repeatable, not recurring. In this case, it makes more sense to think in terms of predictable revenue run rate rather than forcing an ARR-style label. The key is to understand your actual business mechanics, not mimic SaaS language. Perhaps “annualised revenue run rate” (ARRR) is a better way to distinguish it from subscription-based annual recurring revenue (ARR).

It’s important not to label revenue as ARR unless it’s tied to a committed, fixed subscription—implying dependable, sticky revenue that typically grows only at renewal points, whether annually or every few years. Non-subscription revenue, by contrast, can be just as attractive: it may repeat and often scales faster when usage spikes. If customers suddenly use more of your service, revenue could double overnight—making even top-tier SaaS net dollar retention (NDR) rates of 130% look modest in comparison.

Take a fictional AI insights platform. It can charge using a seat- or headcount-based subscription model (as shown in the above graphic on the left chart), or adopt usage-based pricing tied to API calls (as shown on the right). With the traditional SaaS approach, they sign enterprise contracts with defined ACV per client and rely on upsell opportunities in year two. New clients typically sign around budget cycles due to the upfront cost, creating a steady, predictable growth path.

On the usage-based side, early revenue is lower due to ramp-up, demand fluctuates with seasonality, and revenue varies with experimentation. In the early years, it's hard to claim revenue is repeating. But by year five, a baseline emerges—even with seasonal swings. While this model may make it harder to raise early capital, it can reach the same revenue run rate by year six. Valuations may be lower and gross margins less stable, but the model supports faster revenue growth without waiting for contract renewals.

REASON 2: Gross Profit Matters More Than Margin Percentage 

Yes, margins for SaaS are a game changer at 80-90% GMs. Much of this comes from being self-serve, with minimal need for customer support or service providers. That’s why SaaS companies focus on customer success rather than customer support—the goal is to drive adoption and growth, not just resolve issues.

That said, two things are worth noting. First, usage-based models don’t necessarily have much lower margins. While they may require more oversight—monitoring pricing and usage—they can still land in the 50–70% range. Second, we typically see services businesses at around 30% gross margin, and tech-enabled ones at 40–50%. But with AI-enhanced delivery, even these models can push into the 50–70% range.

Home services offer a good example. A traditional pest control SMB might struggle with inefficiencies, as technicians manage scheduling, customer issues, and updates themselves. Platforms like ServiceTitan helped these businesses operate more like tech-enabled services by streamlining back-office tasks. Add AI, and you can go further—automating review responses, generating calls with summaries delivered to technicians between jobs, and effectively putting parts of the business on autopilot. Just as SaaS companies can now run lean with a few engineers, we’ll see tech-enabled services operate with minimal support staff.

Much higher revenue capture, even with a lower margin, means the same or greater gross profit

As a business scales, it makes more sense to focus on annualised gross profit to understand its ability to generate like-for-like dollars. Selling vertical SaaS to an SMB might only capture 3–5% of the potential industry spend. Enterprise buyers often have long sales cycles, are highly budget-conscious, and can be tough in renegotiations. The rest of the market may consist of small regional groups or sole proprietors who resist price increases—even for new features. The SaaS revenue may be high quality and sticky, but it’s limited in size.

In contrast, delivering the service yourself means capturing the full value—even at a lower margin. Take home services again—window cleaning, for example. If a customer pays $200 for a visit, a tech-enabled provider with AI can retain the full amount at a 40–60% gross margin. A vertical SaaS platform serving window cleaners might only capture the equivalent of $10 at 80% margin through a subscription. The revenue potential is significantly greater for a lower-margin, tech-enabled service, and in some cases, vertical SaaS may not make sense as a venture-backed business.

REASON 3: Insight & Access Moats Are The New Tech Moats

SaaS businesses have a reputation for building technology moats against their competition. They build proprietary UI and own access to customer data. That allows the solution to become deeply integrated into an industry and feel like a natural fit. Ask a user of ServiceTitan or Salesforce what it would take to switch, and you may find yourself hearing a long list.

However, in a world of no-code platforms and AI-generated code, a technical moat alone is no longer enough. Assistants and co-pilots now automate support and workflows—more accurately, faster, and cheaper than human teams—freeing up staff for higher-value work. As a result, companies need to be more deeply integrated into the task itself, leading to what we call an insights and access moat.

This more layered moat combines the traditional SaaS strengths—technical capability, a strong UI, and integration into workflows and systems—with advantages often seen in non-SaaS businesses: operational know-how, embedded distribution, and deep industry context. On top of this, companies that own proprietary data or models can build a separate data moat—one not always present in SaaS businesses but increasingly valuable over time. Together, these elements create a defensible position that’s harder to replicate and increasingly critical to compete.

A great example of the insights moat is HarveyAI, a company that uses generative AI to streamline time-consuming tasks for lawyers. They’ve developed not just proprietary technology but a deep awareness of sector-specific pain points that resemble decades of specialised expertise. They directly take on lawyers’ workloads, aiming to deliver high-quality output that can be trusted by both lawyers and their clients (with oversight, of course). Their strength lies in combining operational knowledge with technology to build an insights and access moat.

Another compelling example is Instacart, which built a tech-enabled service rather than selling its technology as SaaS to grocery retailers. By operating the service itself, Instacart captured a much larger portion of the grocery delivery value chain than a pure SaaS play would have allowed. They know how to work with grocers, understand what consumers are looking for, and can dominate far more than a niche SaaS solution could in this context. Similarly, Airbnb created a tech-enabled marketplace rather than selling property management software to turn homes into rentals. They understood that the real constraint was access to new supply, and built a distribution moat by solving that problem—creating a new category with a much larger addressable market. 

Let’s Remove The SaaS Bias

For founders, digressing from the traditional SaaS model often means losing investor interest. It's an immediate turnoff for many VCs, who then suggest bootstrapping or raising PE or growth capital instead. This is shortsighted.

Yes, the valuation of a non-SaaS business may be lower—public comps often trade at 3x forward revenue rather than 8x. However, the potential scale of these companies is significant. Look at Toast ($20B market cap), Instacart ($10B), Airbnb ($80B), or Uber ($150B).

Even if a startup could plausibly be structured as a traditional SaaS product, subscription pricing may simply not make sense. Take OpenAI and Anthropic. Last year, OpenAI reported $4B in sales—around 75% from customer subscriptions—yet those subscriptions are likely losing billions. That segment acts as a loss leader to support R&D and, increasingly, lead generation for its more profitable, fast-growing API business with 50% gross margins. Similarly, Anthropic's revenue is heavily weighted toward API-based usage pricing, with approximately 70% coming from its third-party API (largely via Amazon), 20% from direct API sales, and only 15% from chatbot subscriptions—again, likely at a loss.

The bottom line is simple: focus on building and investing in the right business model for your product and market, rather than forcing a SaaS approach. A well-executed tech-enabled service or usage-based model with strong growth and healthy margins can be just as attractive—and may ultimately create more value—than following the traditional SaaS playbook.