The B2B SaaS Sales Model: Which of the Four You Are Actually Running

A company sells a product at around $7,000 a year. It is doing fine. Then a large logo appears, asks for a security review, a custom contract and a pilot across two business units, and after five months of work it signs for $11,000.

Everybody celebrates. The lesson drawn in the room is that the company can sell to enterprises. Two quarters later a small field sales team has been hired, the cycle on every deal has stretched, the reps are unhappy, and the diagnosis is that the hiring bar was too low.

The hiring bar was fine. The arithmetic never worked, and it never had a chance to. Nobody did it, because sales models are usually discussed as strategy — as a thing a leadership team chooses in an offsite — when they are almost entirely a consequence of a number that was set by the pricing page long before anyone was hired.

The four models, briefly

There are four recognisable go-to-market models in B2B SaaS. Everyone can list them, so this section is short and the rest of the article is about what selects between them.

• Self-serve, or product-led. No human is involved in the purchase. The buyer discovers, evaluates, and pays without speaking to anyone. Sales, if it exists at all, arrives after the fact.
• Inside sales, transactional. One rep, one or two calls, a cycle measured in weeks, conducted entirely over video and email. Usually a single buyer with the authority to sign.
• Mid-market, consultative. One rep plus occasional specialists. Real discovery, a handful of stakeholders, a cycle measured in a couple of months, and a procurement step that is light but real.
• Enterprise, field. A pursuit team rather than a rep: account executive, solutions engineer, sometimes an executive sponsor. Security review, legal redlines, a formal procurement process, and a cycle measured in quarters.

These map onto the axes in our piece on the types of sales, which sorts the general taxonomy into four independent questions — who the buyer is, who starts the conversation, where the meeting happens, and how much decision the buyer carries. This article takes the SaaS case one floor down: not which coordinates exist, but which combination your business can afford.

The four models are four solutions to one equation

Here is the whole thing.

A salesperson costs some amount per year, fully loaded — salary, commission at plan, benefits, tooling, a share of management, and the recruiting cost spread over their expected tenure. Call that cost C.

That salesperson can close some number of deals in a year. Call it N. N is bounded by two things: how long a cycle takes, and how many deals a person can genuinely run in parallel before the quality of each one degrades.

The cost of one closed deal, in sales alone, is C ÷ N. For a model to exist, the gross profit a customer produces over the years they stay must clear C ÷ N by a margin wide enough to also pay for marketing, onboarding, support, and the deals that did not close.

Why the bands sit an order of magnitude apart

The interesting part of the equation is not C. Salespeople do not cost ten times more at the top of the market than at the bottom. The interesting part is N, and N collapses far faster than ACV rises.

A rep working six-week cycles, running a dozen deals in parallel, closes something in the tens per year. A rep working nine-month enterprise cycles, who can genuinely hold maybe six or eight live pursuits at once, closes a handful. That is not a small difference in productivity. It is roughly an order of magnitude in N.

Which means ACV has to rise by roughly an order of magnitude just to hold the equation still. Not to improve the business — to keep it in the same place it was.

This is why the conventional bands sit where they do, roughly a factor of ten apart, and why the boundaries feel real even though nobody can cite a source for them. They are not convention. They are what falls out when you divide a roughly-constant cost by a collapsing deal count.

It also explains the model at the bottom. Below a certain contract value, C ÷ N exceeds the lifetime gross profit of a customer for any N a human being can achieve. There is no rep cheap enough and no cycle short enough. The conversation still has to happen — the buyer still needs to be persuaded — so it gets conducted by software instead.

Self-serve is not a strategy anyone selects. It is what remains when the equation has no human solution.

The touch budget

The equation produces one number that turns out to generate almost everything else about how the company is organised. Call it the touch budget: the amount of human attention one customer can absorb before they stop being worth having.

Model	Touch budget, per customer	What it funds
Self-serve	Effectively zero — measured in support tickets	Docs, onboarding flows, lifecycle email
Inside / transactional	A few hours, across a couple of weeks	One rep, one demo, light follow-up
Mid-market	Tens of hours, across a couple of months	Rep plus specialist time, real discovery, a pilot
Enterprise / field	Hundreds of hours, across quarters	Pursuit team, security, legal, executive sponsorship

The touch budget is the useful number because it is the one you can check against reality. ACV is a fact about your pricing. Touch is a fact about your operations, and the two are supposed to match.

A sales model is broken when the touch a customer demands exceeds the touch the price funds. That is the entire failure mode, and it presents as a long list of symptoms that get diagnosed as almost anything else:

• Reps complaining that they spend their time doing support
• Cycles stretching quarter over quarter with no corresponding change in win rate
• Deals that require a solutions engineer on a product priced for self-checkout
• The founder being pulled into calls for deals of unremarkable size
• Discounting at the end of deals that already cost four times the effort budgeted

Every one of those reads like an execution problem, and every one of them gets managed as one. More coaching, tighter qualification, a new stage definition, a sales methodology rollout. None of it works, because the customer is demanding enterprise attention at an SMB price and no amount of rep performance closes that gap.

There are exactly two fixes. Raise the price so the touch is funded, or reduce the touch the customer needs — which is a product and onboarding change, not a sales change. Choosing neither is the most common option and it is the one that quietly consumes eighteen months.

Headcount and compensation follow the deal count

Once you have N, the org chart is largely determined, including the parts that people usually treat as matters of taste.

Self-serve. The headcount is engineering, lifecycle marketing and support. If you have people with "account executive" on their business cards in a genuinely self-serve model, you have probably misnamed a support team, and the mismatch will show up as reps with no pipeline and no clear definition of what they are meant to close.

Inside sales. Volume makes specialisation pay early. Separating prospecting from closing works here not because it is fashionable but because the cost of a conversation has to stay low and two different jobs are being done badly by one person. The binding constraint is not the on-target earnings figure; it is the ratio of quota to OTE, because that ratio is what determines whether the model funds itself at all.

Mid-market. The awkward band, and the one where most companies are worst. The deals are large enough to require genuine discovery and multiple stakeholders, and too small to fund a pursuit team. The characteristic failure is to staff it like inside sales and sell it like enterprise, which produces reps running eight complex deals with none of the support a complex deal needs.

Enterprise. Pursuit teams, ramp measured in quarters, and a compensation mix weighted further toward base than anywhere else in the business. That last one is not generosity. A rep who closes three or four deals a year cannot have their mortgage depend on the timing of one of them — and if they do, they will behave rationally by chasing whatever closes fastest, which is precisely the behaviour that destroys the enterprise motion you hired them for.

Compensation mix follows deal count. Deal count follows cycle length. Cycle length follows how much governance the price trips on the buyer's side. Everything is downstream of the one number.

Running two at once

Almost every SaaS company past a certain size runs at least two models. Self-serve plus sales-assisted, or inside plus enterprise, is the normal state of the world rather than a sign of confusion.

Running two is not the problem. Running two through one set of shared resources is the problem, and it is close to universal.

The specific breakage is worth spelling out, because it is invisible in the aggregate. Give a rep a pipeline containing a $9,000 opportunity and a $90,000 opportunity, then ask them to hit a revenue number. They will work the large one. This is not a character flaw; it is the only sensible response to the incentive you built. The small band does not collapse — it simply stops growing — and because total revenue keeps rising, nobody investigates. The stalled segment gets written up as market saturation, or as a product-fit problem, or as evidence that the small customers were never very good anyway.

The other thing two models need is an explicit handoff, and the trigger for it is almost always chosen wrongly. Company size is the intuitive trigger and the bad one: plenty of large companies buy small, and a five-person team can have a genuine enterprise-shaped requirement. The right trigger is a behaviour that shows the buyer has hit a wall the product cannot answer on its own — a fifth teammate invited, a permissions or single sign-on requirement, a second department appearing on the same email domain, an export of data that looks like someone preparing a business case.

Those are observable in the product, they happen before the buyer gets frustrated, and they identify the moment when human attention starts being worth its cost rather than the moment an account crosses an employee-count threshold in a data provider's file.

How to tell which model you are actually in

Not which one you describe on a board slide. Four tests, in increasing order of how uncomfortable they are.

1. Do the division. Gross profit per customer per year, multiplied by the number of years a customer realistically stays, against your fully-loaded cost per closed deal. Both numbers are in systems you already own. Most teams have never put them next to each other, and the pairing usually settles the argument in about ten minutes.

2. Count the hours that actually happened. Take your last ten closed-won deals and count every hour of human time in them — the rep, the solutions engineer, the founder, support, legal, the person who filled in the security questionnaire. Not the hours you planned. The hours that occurred. This number is the real touch budget, and for most companies it is materially higher than the one their pricing funds.

3. The founder test. If a founder or a senior executive is on the closing call for a median deal, you are running an enterprise motion regardless of what your price list says. Founder time is the most expensive touch in the company and it appears in no model, because nobody expenses it. It is also the most common hidden subsidy in early SaaS: the motion appears to work right up until the founder stops being available, at which point it is discovered that the founder was the model.

4. Ask who started the last twenty conversations. Did the customers arrive with intent already formed, or did someone manufacture it? That answer tells you whether you have a demand-capture business or a demand-creation one, and the two require different people, different content and different pipeline arithmetic.

If the four tests disagree with each other, believe the hour count. It is the only one of the four that is measured rather than asserted.

What this does not settle

The model tells you the shape of the motion. It does not tell you whether the motion is any good.

Two companies in the same band, at the same ACV, with the same touch budget and the same org chart, will produce very different results, and the difference lives entirely in what happens inside the conversations the model pays for. A mid-market model funds a rep to run real discovery across several stakeholders. Whether that rep can actually run it — whether they can ask a question that changes what a buyer believes, hold a price under pressure, or equip a champion to present without them — is a separate problem that no go-to-market design solves.

This is the half SalesArmor works on. The model decides how many conversations you can afford and with whom. Practice decides what those conversations are worth. Both matter, and it is worth noticing that companies spend enormously more planning attention on the first, which is arithmetic, than on the second, which is craft. For the SaaS-specific version of these conversations, our SaaS sales practice page covers the scenarios that show up in this market specifically.

Common questions about the B2B SaaS sales model

What are the four B2B SaaS sales models? Self-serve or product-led, where no human is involved in the purchase; inside or transactional sales, where one rep closes in a short cycle over video; mid-market consultative sales, with genuine discovery and several stakeholders; and enterprise field sales, run by a pursuit team through formal procurement. They differ less in technique than in how much human attention each customer can be given before they stop being profitable.

What ACV do I need for each model? The honest answer is that there is no lookup table, because the boundary depends on your gross margin, your retention, and how many deals one of your reps can genuinely close in a year. What is reliable is the shape: the bands sit roughly an order of magnitude apart, because deals per rep per year falls by about that much as cycles lengthen, and ACV has to rise by the same factor to compensate. Run the division on your own numbers and the boundary appears.

Can a company run product-led growth and sales at the same time? Yes, and most successful ones do. The constraint is that they need separate funnels, separate compensation plans and separate definitions of a qualified opportunity. Where it fails is when both motions feed one pipeline and one rep, at which point the larger deals absorb all the attention and the self-serve band silently stops growing.

When should a self-serve product add a sales team? When observable behaviour inside the product shows buyers hitting walls the product cannot answer — team expansion, permissions and single sign-on requirements, a second department on the same domain, a security questionnaire arriving. Those are the moments human attention becomes worth its cost. Employee-count thresholds are the wrong trigger, because plenty of large companies buy small and some small ones buy in a thoroughly enterprise-shaped way.

Why do companies fail when they move upmarket? Usually because moving upmarket is treated as a hiring decision when it is a pricing decision. Enterprise motions consume hundreds of hours of human attention per customer, and that has to be funded by contract value that has already risen — not by contract value that the new reps are expected to produce. The sequence matters: the price change enables the model, not the other way round. It is also worth checking what actually changes above the threshold, since it is the buyer's internal approval rules, not the price itself, that restructure the deal.

How many deals should one SaaS rep close per year? This is the wrong question in isolation, because the number is determined by cycle length and parallel capacity rather than chosen. The useful version is to measure your own: take deals closed per rep over the last four quarters, divide the fully-loaded cost of a rep by it, and compare the result to lifetime gross profit per customer. If the second number does not comfortably exceed the first, the model is not viable no matter how the reps perform.

Does the sales model change how we should forecast? Substantially. A transactional model with dozens of deals per rep per quarter has enough volume for stage-based conversion rates to mean something. An enterprise model with three or four deals per rep per year does not — the sample is too small for averages to be informative, and forecasting has to rest on evidence within each deal instead. The stage definitions in your funnel need to reflect that difference rather than being copied across from one motion to the other.

What is the most common mistake in choosing a SaaS sales model? Choosing at all, as though it were a free decision. The model is selected by contract value, gross margin and retention, and a leadership team that picks a model those numbers do not support has not made a strategy decision — it has committed to a subsidy whose size nobody has calculated. The productive version of the conversation starts with what the price funds and works outward from there.

A note on sources

No ACV thresholds, CAC payback benchmarks, average-deals-per-rep figures or cycle-length statistics appear here as facts. Numbers of that shape circulate constantly in SaaS writing, and the traceable ones come from surveys of self-selected respondents where "a deal", "fully loaded" and even "annual contract value" are defined differently by any two contributing companies. Borrowing one would describe somebody else's cost structure rather than yours.

What this article claims instead is arithmetic, and arithmetic you can run in an afternoon on data you already hold. The order-of-magnitude spacing between bands is not a researched finding — it is a consequence of dividing a roughly constant rep cost by a deal count that falls sharply as cycles lengthen, and if your rep cost or cycle length is unusual, your bands will sit somewhere else. That is the point of running it yourself rather than reading a table.

The four tests near the end are all measurements rather than benchmarks: your gross profit against your cost per closed deal, the real hour count on your last ten wins, whether a founder sits on median deals, and who started your last twenty conversations. If those four produce a coherent answer, you know your model. If they disagree, believe the hour count, because it is the only one that was observed rather than reported.