Why We Still Like SaaS: Regulated Vertical Software as an Acquisition Target
TL;DR: The software category got repriced in 2026 and regulated vertical software got caught in the downdraft it least deserved. The SEG public SaaS index median fell from 5.7x EV/TTM revenue in 2Q25 to 3.2x in 2Q26, and Aventis had the median bottoming at 3.2x in June before recovering to 4.6x in August, against an 18.6x peak in 2021. The trigger was February's repricing, when roughly $285 billion came off public software valuations on the view that AI coding tools would let customers build internally what SaaS sells. That logic holds for thin horizontal tools. It does not hold for software that produces a record a regulator audits, because the customer is not buying features, they are buying an audit trail and somebody else's liability. Meanwhile private deal multiples barely moved: SEG's private median went from 4.1x to 4.0x over the same year, and Acquire.com reported a median 3.9x profit multiple on SaaS sales in both 2024 and 2025. So the opportunity is not a crashed price. It is a thinner buyer pool, more motivated founders, and a cost structure where AI removes 30 to 50 percent of the largest expense line in the business. Below is the full thesis, the economics, the verticals we like, the risks, and what we would not buy.
Why do we still like SaaS when everyone is calling it dead?
Because "SaaS" stopped being one category the moment AI got good, and most of the commentary has not caught up.
The February repricing was rational about one thing and sloppy about everything else. It was rational that a form builder, a generic dashboard, a single-feature utility, or a thin wrapper around a workflow anyone can now describe to a coding agent is worth dramatically less than it was. Those businesses sold convenience, and convenience got cheap.
It was sloppy in treating a flight school's training records system and a Slack integration as the same asset class. They are not. One is a subscription. The other is the system of record a regulator inspects, a plaintiff's attorney subpoenas, and an insurer prices off. You do not replace that with a weekend of vibe coding, not because the code is hard, but because the code was never the product.
The second reason is that we underwrite cash flow, not narrative. A vertical software business with 90 percent gross retention, 80 percent gross margins, and no meaningful capital requirement has always been one of the better business models available to a small buyer. What changed in 2026 is the price of admission and the number of people bidding against you, and both of those moved in the buyer's favor.
The third reason is the one nobody in the "SaaS is dead" camp wants to sit with: AI is a larger cost lever for the owner of a small software business than it is a threat to that business's revenue. More on that below, with numbers.
What counts as "regulated vertical SaaS"?
We use a specific test rather than a vibe. A business qualifies if the answer to all three is yes:
- Does the software produce a record that somebody outside the company inspects? A regulator, an auditor, an insurer, an accreditor, a court, or a payer.
- Does the customer face a defined consequence for getting it wrong? A fine, a license suspension, a failed audit, a denied claim, a lost certification, or personal liability for an officer.
- Is the software cheap relative to that consequence? If the subscription costs $400 a month and a failed inspection costs $50,000 and a week of shutdown, nobody is price shopping.
That third condition is the quiet one, and it is why this category resists both churn and AI substitution. Software that is a rounding error in the customer's budget but sits directly between them and a regulator is not a line item anyone volunteers to go rebuild in-house. The internal build has to be right, has to stay right through every rule change, and if it is wrong the person who approved building it owns that personally.
Here is the same idea applied to categories.
| Vertical | What the software is the record of | Who inspects it |
|---|---|---|
| HR compliance and workforce documentation | Employment eligibility, mandated training completion, leave and classification records, incident logs | Federal and state labor agencies, plaintiff's counsel, auditors |
| Pilot and aviation training management | Flight and ground instruction records, currency, checkride readiness, maintenance and duty tracking | FAA, insurers, certificate-holder audits |
| Commercial driver qualification and safety | Driver qualification files, hours of service, drug and alcohol testing records | DOT and FMCSA, insurers |
| Healthcare credentialing and practice compliance | Provider credentials, privileging, payer enrollment | Accreditors, payers, state boards |
| Childcare and early education | Staff-to-child ratios, staff background checks, licensing documentation | State licensing agencies |
| Environmental, health and safety | Incident reporting, emissions and waste documentation, safety training | EPA and state environmental agencies, OSHA |
| Fire, life safety and building inspection | Inspection records, deficiency tracking, certificates | Fire marshals, municipalities |
| Public sector and utility administration | Permits, licenses, billing, records retention | State agencies, auditors, the public record itself |
The specific vertical matters less than the structure. We would rather own unglamorous software for 400 flight schools than exciting software for 400 startups.
Why does regulation protect software from AI substitution?
This is the part of the thesis that has to be right, so it is worth stating precisely rather than assuming.
The threat model is: the customer uses an AI coding agent to build an internal replacement, or an AI assistant absorbs the workflow, and the subscription goes away. For a lot of software, that is a live threat. Public markets priced it as a live threat for everything, which is where the mispricing lives.
Regulated verticals have four defenses that are not about code quality.
Liability transfer. A compliance manager buying software is buying someone to point at. When an auditor asks how training records are maintained, "we use the industry standard system" is an answer. "Our ops lead built it with an AI agent last spring" is an incident. The vendor's existence is part of the product.
Regulatory maintenance is permanent, not one-time. Rules change constantly, and they change differently in each of the fifty states. An internal build is not a project, it is a subscription to keeping up with rulemaking, staffed by people whose job is something else. Most organizations discover this in year two.
Switching costs are revalidation costs. Migrating a system of record means moving historical records that are subject to retention requirements, re-proving chain of custody, retraining staff, and in some settings re-certifying the system itself. This is why vertical software churn is so low: industry analyses of the vertical market software model regularly describe annual churn below 5 percent, and note that the software is often less than 1 percent of the customer's revenue, which means no one has a budget incentive to go looking.
The data is the moat, and it accumulates. Ten years of a customer's inspection history inside your product is not something a new entrant, AI-assisted or otherwise, can recreate. Buyers of software have started pricing this explicitly. Analysts covering the category attribute vertical software's relative resilience to proprietary data, regulatory complexity, and embedded workflow rather than to feature sets.
The honest version of this argument has a boundary. Regulation protects the record, not the workflow around it. If your product's value is mostly a nicer interface for a task the regulator does not care about, you are horizontal software wearing a vertical badge, and you should be priced accordingly.
How much did SaaS multiples actually fall?
Ben's instinct here was half right, and the half that is wrong is the half most buyers get burned by. Public software fell hard. Private software did not.
Public markets, the de-rating:
| Series | Measure | Level |
|---|---|---|
| SEG SaaS Index | EV/TTM revenue, 2Q25 | 5.7x |
| SEG SaaS Index | EV/TTM revenue, 2Q26 | 3.2x |
| Aventis SaaS Index | EV/revenue, June 2026 low | 3.2x |
| Aventis SaaS Index | EV/revenue, August 2026 | 4.6x |
| Public SaaS peak, for reference | EV/revenue, late 2021 | 18.6x |
The February event has a name in the industry now, and roughly $285 billion of market value came off software companies in a matter of weeks. Aventis attributes the mid-2026 trough directly to the AI disruption sell-off, following the release of capable coding agents earlier in the year.
Private markets, the non-event:
| Series | Measure | Level |
|---|---|---|
| SEG private M&A median | EV/revenue, 3Q25 | 4.1x |
| SEG private M&A median | EV/revenue, 2Q26 | 4.0x |
| Aventis private median | EV/revenue, Q1 2026 | 3.1x |
| Acquire.com confirmed sales | Median profit multiple, 2024 | 3.9x |
| Acquire.com confirmed sales | Median profit multiple, 2025 | 3.9x |
Read those two tables together and the conclusion is uncomfortable for anyone selling a "buy software at a discount" story: private sellers did not cut price. What actually changed for a buyer is subtler and, we think, more useful.
- The competing buyer pool thinned. Capital that used to chase any recurring revenue is now asking whether the category survives AI. Fewer bidders per deal at an unchanged asking price is a better deal even when the headline multiple does not move.
- Seller motivation rose. Advisors covering the market report increased founder willingness to sell driven by concentration risk and uncertainty about AI disruption, particularly where most of a founder's net worth sits in one software asset. That is the sentence that creates negotiating room.
- Dispersion widened enormously. The market stopped paying for the category and started paying for the company. Businesses with strong retention and a defensible position hold their price. Thin businesses have no bid at all. A buyer who can tell the difference is being paid more than ever for that skill.
What small vertical SaaS actually trades at. For owner-operated software in the range our readers buy, pricing is on earnings rather than revenue:
| Profile | Typical range |
|---|---|
| Micro SaaS under $1M ARR, owner-operated | ~3 to 4x SDE (Acquire.com median 3.9x profit) |
| $1M to $3M ARR vertical SaaS, owner dependent | ~3.5 to 5x SDE |
| $1M to $5M ARR, management in place, strong retention | ~4 to 5x ARR, above the ~4.0x private median cited above |
For comparison, a waste hauling route book trades at roughly 3.0 to 5.5x SDE and requires trucks. Read our waste hauling thesis for the other side of that trade.
What do the economics of a small vertical SaaS look like?
The figures below are illustrative. They show the shape of the math for a hypothetical compliance software business, not any specific company. Every real deal needs a model built from the seller's actual customer list, pricing, churn, and hosting bill.
| Line | Illustrative figure |
|---|---|
| Customers | 450 |
| Average monthly price | $390 |
| ARR | ~$2,106,000 |
| Hosting and infrastructure | ~$180,000 |
| Support, 2 FTE | ~$160,000 |
| Gross profit | ~$1,766,000 (84%) |
| Engineering, 4 FTE fully loaded | ~$620,000 |
| Sales and marketing, 2 FTE plus spend | ~$330,000 |
| Regulatory content maintenance, 1 FTE | ~$110,000 |
| G&A, SOC 2, insurance, legal | ~$190,000 |
| Owner compensation | ~$180,000 |
| EBITDA | ~$336,000 |
| SDE (EBITDA plus owner comp) | ~$516,000 |
Three things about that table.
There is almost no capital in it. No trucks, no containers, no fleet replacement schedule. The closest thing to capex is engineering payroll, some of which is capitalized as internal-use software development under US accounting rules, which means it shows up as investing rather than operating cash flow and can make a seller's EBITDA look better than the cash actually generated. Check this in diligence. It is the single most common way a software P&L flatters itself.
The cost structure is almost entirely people. Engineering, support, regulatory content, and sales are roughly $1.2M of a $1.77M gross profit. That is the whole business. Which is precisely why the next section matters.
Revenue per customer is small and the customer count is not. 450 customers at $390 a month is a durable structure: no concentration, no enterprise sales cycle, and a price point low enough that renewal is automatic and price increases go unchallenged. Annual escalation in this category is as available as it is in waste hauling, and far more small software owners have never used it.
Where does AI actually lower the cost of running one of these?
This is the part of the thesis we think is genuinely underpriced, and it is an operating lever rather than a valuation opinion.
Every expense line above is a place where an AI-literate owner spends less than the retiring founder did. Applied to the illustrative business:
| Cost line | Before | After | Annual change |
|---|---|---|---|
| Engineering, 4 FTE | ~$620,000 | 3 FTE with agent-assisted development | ~-$155,000 |
| Support, 2 FTE | ~$160,000 | Deflection of routine tickets, 1.5 FTE equivalent | ~-$64,000 |
| Regulatory content maintenance | ~$110,000 | Monitored and drafted with AI, reviewed by a human | ~-$70,000 |
| Onboarding and implementation | Inside sales and support cost | Guided and largely self-serve | ~-$25,000 |
| AI inference and tooling | $0 | New line item | ~+$45,000 |
| Net | ~-$269,000 |
On $516,000 of SDE that is roughly a 50 percent increase in earnings without selling a single additional subscription. At 4x SDE, about $1.0 million of created enterprise value, against a business that might be acquired for $2.0 to $2.6 million.
Now the honest caveats, because this is where these posts usually lie to you.
Inference is a real cost and it lands in COGS. If you ship AI features to customers, your gross margin goes down before your opex goes down. Model it. A vertical SaaS running 84 percent gross margins can find itself at 76 percent after an ambitious AI feature launch, and that shows up directly in what a future buyer pays.
Cutting engineering headcount in a compliance product is riskier than in a consumer app. The thing you are shipping is correctness. Agent-assisted development compresses the time to write code, not the obligation to be right about a regulation. Keep the domain expert.
Support deflection has a floor. The customer calling you is often calling because an auditor is standing in their office. That call needs a human, and answering it well is part of why they renew.
Legacy stacks resist all of this. A fifteen-year-old codebase running on a framework nobody hires for is not where agent-assisted development shines. Assess the stack before you underwrite the savings.
The version of this we believe: AI takes 20 to 40 percent out of the operating cost of a small software business that is run the old way, and the businesses run the old way are exactly the ones for sale from founders who have been at it for a decade. That is the arbitrage. It is not "AI makes software free to run."
How does vertical SaaS compare to the other categories we have written up?
| Dimension | Regulated Vertical SaaS | Waste Hauling | Fence Rental | Property Management |
|---|---|---|---|---|
| Revenue type | Recurring subscription, often annual contracts | Recurring, auto-renewing | Monthly, duration-extending | Recurring, contracted |
| Gross margin | 75 to 85 percent | 12 to 32 percent depending on density | Moderate | Moderate |
| Capital intensity | Very low | High (trucks, carts, containers) | Moderate | Very low |
| Demand driver | Regulatory mandate | Mandated, population-linked | Permit and safety mandated | Rental housing demand |
| Pricing power | High, low price point relative to consequence | Very high, annual escalation norm | Moderate | Moderate |
| Asset value floor at exit | None | Yes | Yes | No |
| Primary risk | Technology and AI substitution | Disposal costs, route density | Cyclicality | Owner dependence |
| Typical small-deal multiple | ~3 to 5x SDE | ~3 to 5.5x SDE | ~2 to 3.5x SDE | ~2.5 to 3x SDE |
| Named buyers at exit | Serial vertical software acquirers, PE platforms | WM, Republic, Waste Connections, GFL | United Site Services, WillScot | Regional platforms |
The comparison worth sitting with is the first and last rows together. Vertical SaaS has far better margins and needs almost no capital, but it has no asset floor. If the product becomes irrelevant, the equity is zero. A hauler that fails still owns trucks and a customer book. That difference should show up in how much leverage you are willing to put on the deal, and it is the main reason we are more conservative on debt in software than in route-based services.
Who buys these at exit?
The exit is not speculative in this category either, and the buyer's playbook is public.
The serial acquirers of vertical market software have industrialized exactly this purchase. Constellation Software has made more than 850 acquisitions across over 100 verticals since 1995 under a buy-and-hold model, explicitly targeting niche software that is mission critical inside one industry and useless outside it, and it operates through decentralized operating groups that each run their own acquisition programs. Those groups buy businesses that are too small for most private equity, which is precisely the size our readers buy.
Then there are the PE platforms doing add-on acquisitions inside a vertical, and the strategics in the vertical itself, which are often the customers' other vendors.
One nuance worth knowing. The same AI narrative that de-rated public software has hit the serial acquirers too, and some of them now trade at valuations that would have looked impossible two years ago. That has two effects on you as a seller down the road: their cost of capital for acquisitions went up, and their appetite for businesses with a clear AI answer went up with it. Plan to sell a business that has already absorbed AI into its cost structure and its product, not one that is waiting to.
Where do you find vertical SaaS businesses for sale?
Here is one we found, a $3.3M software business with 90 percent retention, and the part of the fine print that decides whether it is worth the ask 👇
Some of it is listed, and more of it than in any of our route-based categories.
The marketplaces that carry real software deal flow are Flippa at the small end, Empire Flippers and Quiet Light in the curated middle, and Website Closers and Acquire.com at the larger end. Our full marketplace comparison covers the rest.
Rather than checking each of those in turn, our on-market database pulls the listed software deals from every source we index into one searchable place, so you can filter by cash flow, asking price and location in a single pass and set an alert for new software listings instead of going back to check.
The catch is the same as always: the listed inventory skews toward businesses whose owners decided to run a process, and in software that increasingly means businesses with an AI problem they would rather not solve. The best assets in this category, like the best haulers, mostly transact quietly. A founder running $2M of ARR in compliance software for a niche industry knows the three companies who would buy it.
So the sourcing approach is a hybrid. Search the marketplaces for the listed deals, then build a direct list for the ones that are not listed. Vertical software companies are unusually easy to enumerate: pick a regulated industry, look at what its trade association endorses, what its conference exhibitor list contains, and what software the regulator's own guidance mentions by name. That is your target list, and it is rarely more than twenty companies per vertical.
Building that list by hand is the tedious part, so we built it into the product. Off-market sourcing has a digital search mode made for exactly this category: give it a niche like flight school management software or driver qualification compliance, and it works across directories, industry leaderboards and the open web to assemble the companies that fit, rather than the local business listings a map search would return. That distinction matters here. Software companies do not have a storefront on a street, so the usual approach of searching a radius around a city finds nothing. From there you can run a deep dive on the ones worth pursuing, which finds the owner, the decision makers and a direct way to reach them.
Then write to them. Our off-market outreach playbook covers what gets a reply, and outreach sends the letters from the same place you built the list.
What are the risks of buying a vertical SaaS business?
We would not publish a thesis without the other side, and in this category the other side is louder than usual.
AI substitution is a real risk, not just a narrative. The market repriced software for a reason. If the product's value is a data entry interface with a report button, an AI-assisted internal build is a genuine threat within a few years. Our whole thesis depends on picking businesses where the regulator, not the interface, is the moat. Get that wrong and you have bought a melting ice cube at a 4x multiple.
Deregulation is the tail risk nobody models. The demand driver here is a rule. Rules get repealed, agencies get defunded, mandates get delayed. A single compliance product serving a single mandate is a policy bet. Prefer businesses that serve several overlapping requirements, and check whether any of them are currently contested.
Technical debt is the diligence trap. The equivalent of a twelve-year-old fleet is a codebase on an unsupported framework with one developer who understands it. Commission an independent code and infrastructure review. Ask specifically about the hosting bill trend, the database, and whether anything critical runs on a machine somebody set up by hand.
Key person risk is worse in software than in services. In a hauler the owner drives a truck. In small software the owner often is the engineering team, the domain expert, and the reason the top twenty accounts renew. Structure for a real transition period and check whether the code, the architecture, and the regulatory knowledge are documented anywhere other than one person's head.
Security and data liability sit on your balance sheet now. You are holding records that other companies rely on for compliance. A breach is not just a cost, it is a breach of the thing you sell. Review the security posture, incident history, insurance, and customer contract liability caps before signing.
Growth is usually the weak spot. A lot of these businesses have not added a net new customer in years and are living on retention and price. That can be fine at the right multiple, but underwrite it honestly rather than assuming you will fix distribution. Bootstrapped software companies in the low millions of ARR have seen growth normalize into the mid-teens and below; assume flat until proven otherwise.
Seat-based pricing shrinks when your customers shrink. If you sell per employee and your customers are cutting headcount, your revenue falls without losing a single account. Check the mix of per-seat, per-location, and flat pricing.
Concentration. Same rule as everywhere: know what the top five customers are as a share of revenue, and know when their contracts renew.
How should an accredited investor evaluate a vertical SaaS acquisition?
Start with retention, by cohort, for three years. Gross dollar retention and logo retention, not net. Net revenue retention can hide churn behind upsell. In a regulated vertical you want to see gross logo retention above 90 percent, and if it is not there, find out which cohort is leaving and why.
Then do the mandate audit. Write down the specific rule, statute, or accreditation standard the software serves. Name it. If you cannot name it, this is not a regulated vertical SaaS business, it is a productivity tool, and you should be paying a productivity tool multiple.
Reconcile revenue to cash. Annual prepayments, deferred revenue, and capitalized development costs all distort a small software P&L. Get the bank statements, the Stripe or billing export, and the deferred revenue schedule. Confirm what the seller calls EBITDA against cash actually generated after capitalized engineering.
Get an independent technical review. Code quality, dependencies, infrastructure, security posture, and a realistic estimate of what it costs to modernize. Price the modernization into your model, not into your hopes.
Test the AI cost case with the actual team. Before you underwrite $250,000 of savings, ask the engineers what fraction of their work is new feature development, maintenance, support escalation, and compliance updates. Agent-assisted development helps most on the first category and least on the last.
Interview five customers. In this category customers will tell you the truth quickly, because they have no reason not to. Ask what happens if the product disappears tomorrow, what they would do instead, and whether they have ever considered building it internally. The answers are your moat assessment.
Then write down the value creation case in numbers. Price increase, cost reduction, support deflection, adjacent module, upsell into the existing base. If you cannot express it as a change in EBITDA with a timeline, you do not have a plan. The multiple you pay assumes the business as run today. Everything above that is your return.
Frequently Asked Questions
Are SaaS multiples down in 2026? Public SaaS multiples fell sharply. The SEG index median went from 5.7x EV/TTM revenue in 2Q25 to 3.2x in 2Q26, and Aventis recorded a trough of 3.2x in June 2026 before a recovery to 4.6x in August, against a 2021 peak of 18.6x. Private deal multiples did not follow: SEG's private median held near 4.0x, and Acquire.com reported a median 3.9x profit multiple on confirmed SaaS sales in both 2024 and 2025.
Why did SaaS valuations fall in 2026? The proximate cause was investor conviction that AI coding tools and assistants would let customers replace or absorb much of what software sells. Roughly $285 billion in public software market value came off in a matter of weeks in February 2026. Growth deceleration across the sector compounded it.
Is vertical SaaS safer from AI than horizontal SaaS? In our view, yes, for structural reasons rather than technical ones. Vertical software's defensibility comes from proprietary data, embedded workflow, regulatory complexity, and switching costs that include revalidation, none of which a coding tool removes. Horizontal tools whose value is a generic capability with an interface are the ones directly exposed.
What multiple do small SaaS businesses sell for? Owner-operated software under roughly $1M ARR commonly transacts around 3 to 4 times profit, with Acquire.com's confirmed-sale median at 3.9x. Larger, better-retained businesses with management in place price on revenue rather than earnings, generally in the 3x to 6x revenue band depending on growth, retention, and margin.
What makes a vertical SaaS business worth a higher multiple? Gross retention above 90 percent, gross margin above 75 percent, a named regulatory mandate driving demand, no customer above 10 percent of revenue, contracts rather than month-to-month, a modern and documented codebase, and an owner who is not the engineering team.
Can AI make a small SaaS business more profitable? It is the largest operating lever available in the category, because nearly all the cost is people. Agent-assisted development, ticket deflection, and automated regulatory monitoring can remove a meaningful share of operating cost. It is not free: inference costs land in gross margin, and correctness work in a compliance product still needs domain experts.
Who buys small vertical software companies? Serial acquirers of vertical market software, private equity platforms doing add-ons inside a vertical, and strategics already selling to the same customers. Constellation Software alone has completed more than 850 acquisitions across over 100 verticals and specifically targets businesses too small for most private equity.
Where can I find SaaS businesses for sale? Flippa, Empire Flippers, Quiet Light, Website Closers and Acquire.com all carry software listings. The best assets in regulated verticals frequently transact off-market, which is why we pair marketplace search with direct owner outreach.
Is vertical SaaS a better acquisition than a service business? They solve different problems. Software has far better margins and needs almost no capital, but it has no asset floor and carries technology risk. Route-based service businesses have worse margins and real capital requirements, but the assets and the customer book retain value even when the business is run badly. We would underwrite less leverage on software for that reason.
Disclaimer
This article is for informational and educational purposes only. It is not investment, legal, tax, or financial advice, and it is not a recommendation to buy or sell any business or security. Accredited is a publisher, not a broker-dealer, investment adviser, or licensed M&A intermediary. Valuation multiples, market indices, and cost figures cited here are general industry references that vary by deal, geography, and time, and should not be relied on for any specific transaction. The unit economics and AI cost illustrations in this article are hypothetical and are provided to demonstrate a method of calculation, not to project the results of any business. Software serving regulated industries is subject to federal, state, and local requirements that change frequently and vary by jurisdiction. Conduct your own due diligence and consult qualified legal, accounting, technical, and security professionals before pursuing any acquisition.
Sources referenced include the Software Equity Group SaaS index and private M&A data, Aventis Advisors SaaS valuation research, Acquire.com's biannual acquisition multiples report, SaaS Capital private company benchmarks, Bain's 2026 Private Equity Midyear Report, and public company disclosures.