SaaS Pricing Tiers: How to Design the Ladder

Most SaaS pricing tiers are three numbers chosen in an afternoon and defended for five years. That is why the middle one stops selling.

A price ladder is not a set of prices. It is four jobs assigned to four rungs: an entry or free rung that owns the funnel, a volume rung priced at the impulse point, a middle rung with a defensible job of its own, and a high anchor rung whose main function is to make everything beneath it read as reasonable. Miss a job and the rung underneath absorbs it.

The mechanics are lawful rather than arbitrary. Verified in July 2026, adjacent paid B2B rungs step at roughly 2x to 5x, consumer rungs step far tighter at 1.16x to 1.69x, and annual discounts cluster in a 16.7% to 20% band. The numbers are downstream of the fences, because what makes a tier a tier is the thing that stops a buyer helping themselves to the rung above.

This piece reads five public price cards captured July 28, 2026, computes every ratio from them, and ends with a named audit for your own card.

Key takeaways

• Four rungs, four jobs. Slack runs Free, Pro, Business+, Enterprise+; Notion and Figma run the same four-rung stack (vendor pricing pages, July 28, 2026).
• B2B steps wide, consumer steps tight. Adjacent paid rungs, computed: Notion 2.00x, Slack 2.06x, Figma 1.64x to 3.44x on the Full seat, Claude 5.00x. Spotify’s consumer ladder runs 1.16x to 1.69x.
• The fence matters more than the number. Figma sells a $3 Collab seat and a $90 Enterprise Full seat off the same product, an 18.0x spread held up by seat role, not feature count (computed).
• The anchor rung has academic backing. Simonson and Tversky (JMR, 1992) documented extremeness aversion; Huber, Payne and Puto (JCR, 1982) documented asymmetric dominance. The tier nobody buys works for the tiers people do buy.
• Charm pricing does not cross the fence. All four Spotify consumer prices end in .99; none of the nineteen published B2B list prices here do (computed). Anderson and Simester (QME, 2003): the effect needs a buyer without a reference price.
• A working ladder is the cheapest revenue you will buy. Benchmarkit’s 2025 medians put the expansion CAC ratio at $1.00 against a new CAC ratio of $2.00, a computed 2x advantage for revenue that climbs the ladder rather than entering it.

---

How many pricing tiers should a SaaS product have?

Four is the working default, and the reason is jobs rather than aesthetics. You need an entry or free rung that owns the funnel, a volume rung priced near the impulse point, a middle rung with a defensible job of its own, and a high anchor rung that makes everything beneath it read as reasonable. If you cannot state a rung’s job in one sentence, you have a price point, not a tier.

Three is what most founders pick, and it is one short. Three rungs force the middle to be both the reasonable default and the upgrade destination. It cannot be both: a default is chosen by comparison, an upgrade is chosen by need.

• Rung 0, entry or free. Owns the funnel. Its output is qualified accounts, not revenue, and it needs a metered ceiling.
• Rung 1, volume. Priced where one decision-maker says yes without a meeting. It sets the reference price the whole card is read against.
• Rung 2, middle. Fails most often. Its job is to serve a buyer with a named requirement, not to be the average of its neighbours.
• Rung 3, anchor. Sets the reference point. It needs a real buyer, but its highest-value output is the frame it puts around rung 2.

One decision sits upstream of all of it. Whether you meter seats, usage, or both sets the retention shape of the whole ladder before you assign a single job, the tradeoff mapped in usage-based pricing versus seat-based pricing.

---

What do real SaaS pricing tiers look like on a live price card?

More disciplined than the folklore suggests. Across five public cards captured July 28, 2026, four of five run exactly four rungs, the paid rungs step in recognizable multiples, and annual discounts sit in a band narrow enough to look coordinated. What varies is not the shape but the fence.

Vendor	Tier	Billed monthly	Billed annually
Slack	Pro	$8.75/user/mo	$7.25/user/mo
Slack	Business+	$18/user/mo	$15/user/mo
Slack	Enterprise+	Contact sales	Contact sales
Notion	Plus	$10/member/mo	Not displayed
Notion	Business	$20/member/mo	Not displayed
Notion	Enterprise	Custom	Custom
Figma	Professional	Full $16, Dev $12, Collab $3	Not displayed
Figma	Organization	Not shown monthly	Full $55, Dev $25, Collab $5
Figma	Enterprise	Not shown monthly	Full $90, Dev $35, Collab $5
Claude	Pro	$20/mo	$17/mo ($200/year)
Claude	Max	From $100/mo (5x and 20x usage)	Not displayed
Claude	Team	Standard $25/seat, Premium $125/seat	Standard $20/seat, Premium $100/seat
Claude	Enterprise	$20/seat plus usage at API rates	Not displayed
Spotify	Student	$6.99/mo (1 verified account)	Not displayed
Spotify	Individual	$12.99/mo (1 account)	Not displayed
Spotify	Duo	$18.99/mo (2 accounts)	Not displayed
Spotify	Family	$21.99/mo (up to 6 accounts)	Not displayed

Sources: Slack, Notion, Figma, Anthropic (Claude) and Spotify public pricing pages, captured July 28, 2026. All four B2B vendors also publish a $0 rung (Slack Free, Notion Free, Figma Starter, Claude Free). “Not displayed” means the vendor did not render that figure in page content.

Four readings fall out before any arithmetic:

• Some vendors hide the annual price deliberately. Notion advertises “Save up to 20% with yearly” without rendering the per-seat result; Figma’s Professional rates sit behind a toggle. Printing both invites the buyer to treat the discount as the decision.
• Slack is the clean stack. One price per rung, both cadences printed: easy to read, easy to benchmark, a trade in favor of self-serve conversion.
• Figma is a matrix, not a ladder. Three seat types per plan at three prices, so the buyer chooses a mix. That is how Figma charges $90 and $5 inside one plan.
• Spotify’s ladder is a household. The rungs are counts of accounts plus a verified-status rung for students. Nobody gets more product by moving up. They get more people.
• Claude prices usage multiples, not features. Max sells as 5x and 20x the Pro allowance, and Enterprise prices a seat and a meter at once, the leading edge of what AI is doing to packaging, examined in why AI is breaking per-seat SaaS pricing.

---

How big should the price gap between SaaS pricing tiers be?

In B2B, adjacent paid rungs in this sample step between roughly 1.6x and 5x. Consumer rungs step much tighter, between 1.16x and 1.69x. A step under about 1.5x in B2B usually means the fence is missing, not that the tier is a bargain.

Ladder	Adjacent step	Ratio (computed)
Slack	Pro to Business+	2.06x monthly, 2.07x annual
Notion	Plus to Business	2.00x ($20 / $10)
Figma Full seat	Professional to Organization	3.44x ($55 / $16)
Figma Full seat	Organization to Enterprise	1.64x ($90 / $55)
Figma Full seat	Professional to Enterprise	5.63x ($90 / $16)
Figma Dev seat	Professional to Organization	2.08x ($25 / $12)
Figma Dev seat	Organization to Enterprise	1.40x ($35 / $25)
Figma Collab seat	Organization to Enterprise	1.00x ($5 / $5)
Claude	Pro to Max entry	5.00x ($100 / $20)
Claude	Team Standard to Team Premium	5.00x ($125 / $25 monthly, $100 / $20 annual)
Claude	Pro to Team Standard, monthly	1.25x ($25 / $20)
Spotify	Individual to Duo	1.46x ($18.99 / $12.99)
Spotify	Duo to Family	1.16x ($21.99 / $18.99)
Spotify	Individual to Family	1.69x ($21.99 / $12.99)

All ratios computed by siliconcent from the vendor pricing pages captured July 28, 2026.

B2B tolerates a wide step because the person approving the invoice is rarely the person using the product. A 3.44x jump from Figma Professional to Organization is not a 3.44x better day for a designer. It is a different buyer with a different budget line. Consumer does not work that way: Spotify’s entire paid ladder spans 1.69x, because the household wallet is the budget and substitution is one tap away.

Two anomalies are diagnostic. Claude’s Pro to Team Standard step is only 1.25x on monthly list, the narrowest paid step here: the price of a business wrapper rather than a new product. When a step is that small the fence is administrative, and the vendor is removing friction from the individual-to-team migration on purpose.

Figma’s Collab seat does not step at all between Organization and Enterprise (both $5, a 1.00x ratio). The plan changes; that seat’s price does not. The Collab seat is a distribution mechanism, and you do not charge more to add viewers because viewers are how the file spreads.

The rule that falls out: if a paid B2B step is under 1.5x, merge the rungs or find a real fence. A narrow step produces indifference rather than upgrades, and indifference defaults to the cheaper rung.

---

Fencing is the only thing that makes a tier a tier

A tier is not a price. It is a boundary a customer cannot cross without paying, plus a reason they accept the boundary as fair. Remove the fence and the tier is a suggestion.

Fence type	What it gates	Durability	Live example (July 28, 2026)
Feature fence	A capability switched off below the line	Weakest	Notion Plus at $10 against Business at $20 per member/month
Seat-type fence	Price by role rather than headcount	Strong	Figma Enterprise Full seat $90 against Collab seat $5, an 18.0x spread (computed)
Usage or meter fence	Credits, calls, storage, compute	Strong and self-adjusting	Notion AI at $10 per 1,000 monthly credits; Figma bundling 3,000 AI credits into a Professional Full seat
Admin and compliance fence	SSO, audit logs, roles, data residency	Most durable	Slack Business+ and Enterprise+; Claude Enterprise at $20/seat plus usage at API rates

Prices and product terms from the vendor pricing pages cited above, captured July 28, 2026. Durability rankings are analytical judgment, not vendor disclosure.

Feature fences are the default because they are easiest to build, and the first to erode: they bet that no competitor will include the same capability one rung lower, and that bet gets called constantly. Seat-type fences hold better because they map to who a person is, not what they can click. Figma’s spread is the clearest case, 18.0x between an Enterprise Full seat and a Collab seat and 5.33x at Professional ($16 against $3), both computed. Nobody argues that a stakeholder who comments weekly should pay what a designer pays.

Meter fences are the most self-adjusting. When the boundary is credits, the fence moves with the customer automatically and cost of goods moves with revenue, which is why Notion sells AI agents at $10 per 1,000 monthly credits and Figma allocates 3,000 monthly AI credits to a Professional Full seat against 500 for Dev and Collab seats.

Admin and compliance fences are the most durable, for a permanent reason: SSO, audit logs, role-based access and data residency are wanted by a different person than the one who wants features. The user wants capability; the buyer wants control. A competitor shipping one more feature cannot undo that.

The test: name the person blocked by each fence. “A power user who would like more” is soft. “An IT administrator who cannot sign without it” holds for years.

---

The anchor rung: what choice research says about the tier nobody buys

The high rung is not there to be bought. It is there to change how the rung below is read, and that claim has more academic support than most pricing folklore.

Simonson and Tversky, “Choice in Context: Tradeoff Contrast and Extremeness Aversion” (Journal of Marketing Research, Vol. 29, No. 3, pp. 281-295, 1992), found that options at the extremes of a choice set lose appeal while intermediate options gain it. Add a credible expensive rung and the previously expensive rung becomes the moderate one. Huber, Payne and Puto (Journal of Consumer Research, Vol. 9, No. 1, pp. 90-98, 1982) supply the companion mechanism: an option clearly inferior to one alternative but not another raises the probability of choosing the dominating option, a violation of regularity.

The vivid illustration comes from Dan Ariely’s Predictably Irrational (2008): an Economist page offering online-only at $59, print-only at $125, and print plus web at $125 split 16% / 0% / 84%, then flipped to 68% / 32% once the print-only decoy was removed. That is a reported student experiment, not a peer-reviewed field result, so cite it as illustration rather than as effect size for software.

Map the theory onto the cards. Claude prices Max entry at 5.00x Pro ($100 against $20) and Team Premium at 5.00x Team Standard; Figma prices Enterprise at 5.63x Professional on the Full seat ($90 against $16), both computed. Whether or not those vendors call the rung an anchor, the arithmetic is doing anchor work: a $20 subscription read against a $100 option is a different object than a $20 subscription read alone.

Two disciplines keep it honest. The anchor must have a real buyer, because a rung nobody purchases is a decoy and decoys get noticed. And it must be legible: “contact sales” is a weak anchor because it has no number. Slack’s Enterprise+ and Notion’s Enterprise both sit behind contact sales, trading framing power for negotiating room. Claude’s published “From $100” does the opposite.

---

Does charm pricing survive contact with procurement?

Mostly not. Anderson and Simester found $9 price endings increased demand in all three of their field experiments, but with two boundary conditions the marketing blogs skip: the lift was larger for newly introduced items, and it shrank when the price was paired with a “Sale” cue. The mechanism depends on the buyer lacking a reference price. Procurement buyers have reference prices.

That is the finding from “Effects of $9 Price Endings on Retail Sales: Evidence from Field Experiments” (Quantitative Marketing and Economics, Vol. 1, pp. 93-110, March 2003), which attributes the effect to consumers using the ending to infer value when they do not know what an item normally costs.

Card	Published list prices	Ending in .99
Spotify (consumer)	$6.99, $12.99, $18.99, $21.99	4 of 4
Slack	$8.75, $7.25, $18, $15	0 of 4
Notion	$10, $20	0 of 2
Figma	$16, $12, $3, $55, $25, $5, $90, $35	0 of 8
Claude	$20, $17, $100, $25, $125	0 of 5
B2B total	19 published price points	0 of 19

Counts computed by siliconcent from the vendor pricing pages captured July 28, 2026.

Nineteen published B2B price points, zero charm endings. Four consumer prices, four charm endings. The sample is small and not a controlled test, but it matches the mechanism Anderson and Simester identified rather than the folk version of the tactic. Slack is the interesting middle case: $8.75 and $7.25 are neither round nor charm-ended, but the output of a discount calculation, a third register that signals arithmetic instead of a deal.

The transferable rule is about buyer information, not digits. When your buyer compares three vendors in a spreadsheet and forwards the page to finance, the .99 only makes the total harder to add up.

---

Why did my middle pricing tier stop selling?

Three usual causes. The middle is the entry tier plus unfenced features, so it reads as a tax rather than a product. The step from the rung below is too narrow, typically under 1.5x, so nobody is pushed to move. Or the middle is aimed at a company size that does not exist in your funnel. The diagnostic is mix: under roughly 15% to 20% of paid accounts, the rung has no job.

• The middle is a tax. The entry rung does the work and the middle adds capabilities nobody asked for, so the buyer sees a price increase without a reason. The fix is a fence, not a discount: find the one capability a specific buyer role cannot operate without, move it up, move everything else down.
• The step is too narrow. Under about 1.5x, upgrading feels like paperwork rather than a decision. Widening the gap can raise middle-tier conversion, because it forces the buyer to evaluate the fence instead of shrugging at it. Extremeness aversion cuts both ways: a middle that no longer sits between two credible extremes stops harvesting compromise demand.
• No buyer exists. The middle was designed for “companies between 50 and 500 people” because that felt like a segment. If your funnel is 90% teams under 20 and 8% enterprises, the rung is aimed at empty space.

The cost of an idle middle is larger than it looks, because the middle is where expansion revenue lives. Benchmarkit’s 2025 medians put the expansion CAC ratio at $1.00 against a new CAC ratio of $2.00, a computed 2x advantage, with expansion ARR at 40% of total new ARR. A dead rung forces you to buy growth at the expensive end of that trade. It is also how healthy gross retention sits beside stagnant net retention, the divergence unpacked in gross retention versus net retention in SaaS IPOs.

---

The free rung has to pay rent

Free is not a marketing decision. It is a rung with a job, a cost, and a required output, and it belongs on the card only if it delivers conversion or distribution. Everything else is a cost of goods line with a friendly name.

• Free tier. Permanent, capped. Best when the product has network or artifact effects, because non-paying users pull paying ones in. Figma’s Starter tier and the $5 Collab seat are distribution instruments, not revenue.
• Free trial. Time-boxed full access. Best when value is obvious within days and the buyer is one person. Converts faster, distributes nothing.
• Reverse trial. Full access for a window, then automatic downgrade to a capped free tier. Captures both jobs at the cost of a more complex activation story.

Whichever you pick, the fence question is the same: what is capped, and does the cap bind where the customer’s success creates your cost? Capping storage when the pain is collaborators produces free users who never feel the wall. AI made that urgent, because inference is variable cost rather than near-zero marginal cost, and the cards show vendors metering the giveaway: Notion sells AI Custom Agents as “free to try, then $10 per 1,000 monthly Notion credits,” and Anthropic’s Enterprise plan is a $20 seat price plus usage at API rates.

Activation is what turns the free rung from a cost into a funnel, covered in free trial activation and SaaS growth. The rung earns its place when it produces qualified accounts below the cost of paid acquisition. When it does not, it is a subsidy paid to people who were never going to buy.

---

The annual discount is a financing decision, not a marketing one

Across every vendor here that publishes both cadences, the annual discount lands in a 16.7% to 20% band. Slack discounts 17.1% on Pro and 16.7% on Business+. Claude discounts 16.7% on Pro and 20.0% on both Team seat types. Notion advertises “up to 20%”. All computed from the July 2026 cards.

That clustering is not aesthetic convergence. A sixth to a fifth of a year’s revenue is what a software business will pay to hold twelve months of cash today, and to remove eleven monthly opportunities for the customer to leave. Read it as two instruments bundled together:

• A cost of capital. Prepaid annual revenue is customer-funded working capital with no covenants, no dilution and no interest schedule. The discount is the coupon. Frame it against your alternative cost of capital, not a competitor’s card.
• A churn-suppression instrument. Monthly billing gives a customer twelve exit points a year; annual gives one. The retention step-change is often worth more than the revenue given up, particularly on self-serve rungs where cancelling is a two-click decision.

The costs arrive later. An annual contract locks your price against your own future increases for the term, which is expensive when Simon-Kucher’s Global Pricing Study 2025 (2,200+ business leaders across 28 countries and 39 industries, fielded early 2025) found companies realize less than half of their intended price increases on average. Freezing the card for twelve months compounds a shortfall you already have.

One presentation detail is worth copying. Claude’s card shows “$17/mo” against a $200 annual plan; the exact monthly equivalent is $16.67 (computed), rounded up for display. Rounding up to a legible number rather than down to a flattering one is a small credibility signal.

---

The Tier Ladder Blueprint: a framework for auditing your price card

Call this the Tier Ladder Blueprint. One matrix assigns each rung its job, its pricing rule, its fence, and the specific way it fails. Run any price card through it and the broken rung identifies itself. The structure is original analysis; the pricing rules are calibrated to the step ratios computed above.

Rung	Its job	How to price it	How to fence it	Failure mode
Rung 0: entry or free	Own the funnel	$0, or the lowest price that filters out non-buyers	Cap the unit that scales your cost (credits, collaborators, history)	Uncapped cost of goods; users never hit a wall, so they never convert
Rung 1: volume	Convert at the impulse point, set the reference price	The highest number one decision-maker approves without a meeting	One capability gate the buyer can name	Priced above the impulse point, so it needs a sales motion it cannot afford
Rung 2: middle	Serve a named buyer with a specific requirement	1.5x to 3x rung 1 (Notion 2.00x, Slack 2.06x, Figma Dev 2.08x, computed)	Buyer-role fence: admin, roles, SSO, audit, support tier	Collapses into rung 1; mix falls under 15% to 20% of paid accounts
Rung 3: anchor	Set the reference point, capture the largest accounts	3x to 5x rung 2 (Claude Max 5.00x Pro, Figma Enterprise 5.63x Professional, computed)	Compliance, residency, scale limits	Priced without a real buyer, so it reads as a decoy

Rung jobs, fences and failure modes are siliconcent analysis. Step-ratio guidance is computed from Slack, Notion, Figma and Anthropic public pricing pages captured July 28, 2026.

The Ladder Test: eight questions, scored

One point per yes. Six of eight passes; five or fewer means the ladder is your growth constraint.

1. Can you state the job of every rung in one sentence, without the words “more” or “better”?
2. Is every adjacent paid step between 1.5x and 5x? Below 1.5x, merge or re-fence. Above 5x, you are missing a rung.
3. Does every fence map to a buyer role rather than a feature list?
4. Does the anchor rung have a real buyer, even a small one, and a published or quotable number?
5. Is the middle rung above 15% of paid accounts? If not, it has no job.
6. Does the free rung have a metered cost ceiling on the dimension that drives your cost of goods?
7. Is the annual discount consistent across rungs and defensible as a cost of capital rather than a reflex?
8. Can a customer self-serve up one rung without talking to sales?

Operator playbook: shipping a repricing without breaking the base

• Grandfather existing accounts publicly and generously. Grandfathering is a known, finite revenue delay. A migration revolt is unbounded and lands in public.
• Ship the new card to new logos first, for a full sales cycle. New-logo data tells you whether the fence holds; migration data only tells you how angry people are.
• Instrument tier mix weekly. Mix tells you whether the middle rung took; a quarter is three months of shipping the wrong card.
• Migrate on renewal, never on announcement. Renewal is a moment the customer already expects to think about price. An announcement is an ambush.
• Brief support and customer success first, then customers, then press. The team answering tickets needs the answer first.
• Publish the fence, not the price change. A higher price attached to a boundary customers understand lands far better than one attached to nothing.

---

What I learned designing a price ladder from scratch

I designed the price ladder for a consumer SaaS platform I operate, and I built it the way this piece recommends: by pulling real competitor price cards first and working backwards from their shape rather than from my own cost model. That step changed almost every number I had originally guessed.

The ladder uses roughly 1.7x steps between rungs. Consistent multiples make the jumps legible, so a buyer scanning the card feels the rhythm even without doing the division. Arbitrary steps read as pricing by mood, and buyers who sense that start negotiating.

It is charm-priced, because it is a consumer product with a consumer buyer, exactly the condition Anderson and Simester’s result depends on. I would not use those endings on a card sold to procurement.

The top tier exists deliberately as an anchor. It is a real product with real buyers, but its job is to make the middle rungs read as reasonable rather than to be the volume seller. I do not expect most buyers to pick it, and the ladder would be worse without it.

Annual plans carry a deep effective discount, in the 44-50% range, far outside the 16.7% to 20% band the B2B cards cluster in. That is on purpose, and the goal is not the headline saving. It is annual mix: cash up front, plus the retention step-change from converting twelve monthly cancellation decisions into one. On a consumer product with a low-friction cancel button, that is the largest retention lever available. The cost is honest and I price it in: a year of locked revenue at a low rate, with no ability to reprice that cohort until renewal.

Pricing is banded by purchasing power parity by region rather than one global number, because a single worldwide price is either unaffordable in most of the world or underpriced in the richest markets.

The hard-won lesson matches the data above. The middle tier is by far the hardest rung to sell. It has to have a job of its own, and “the one between cheap and expensive” is not a job. Every time I defined the middle by what it was not, it underperformed. Defining it by the buyer it serves is the only thing that has worked.

---

Worked example: rebuilding a collapsed middle, line by line

The numbers below are illustrative and hypothetical, chosen to show the arithmetic of a repricing rather than to describe any real company.

Illustrative, before	Price per account/mo	Accounts	MRR	Mix
Starter	$15	830	$12,450	83.0%
Growth	$22	90	$1,980	9.0%
Enterprise (custom)	$250	80	$20,000	8.0%
Total		1,000	$34,430

Run the Ladder Test. The Starter-to-Growth step is 1.47x, below the 1.5x floor. Growth mix is 9.0%, under the 15% threshold. Growth’s fence is “more of everything,” failing question three, and nothing sits between $22 and a custom contract, so question two fails at the top too. Four out of eight.

Illustrative, after	Price per account/mo	Job	Fence
Free	$0	Own the funnel	Metered ceiling on the cost-driving unit
Starter	$15 (unchanged)	Volume, reference price	One named capability gate
Growth	$35	Serve the admin buyer	SSO, roles, audit log, admin API
Scale	$90	Self-serve step before sales	Scale limits, priority support, residency
Enterprise	Custom	Anchor and procurement	Contract terms, security review, custom SLA

Computed steps after the rebuild: Starter to Growth 2.33x, Growth to Scale 2.57x, Starter to Scale 6.00x. All inside or adjacent to the Blueprint bands.

Now the part most repricing plans skip. Revenue neutrality first, accretion second. Growth previously produced $1,980 of MRR; at $35, matching it takes 57 accounts ($1,995, computed) against the 90 the old rung held. The change is revenue-neutral even if 36.7% of middle-tier accounts refuse to climb. At a target 20% mix the middle produces 200 x $35 = $7,000, up $5,020. At a pessimistic 10% mix it produces 100 x $35 = $3,500, still $1,520 above the old rung (all computed). The downside case is accretive, which is what makes a repricing safe to ship.

Methodology: how to read this model

• Inputs: current tier mix, average revenue per account, gross churn by tier, expansion rate by tier. Here: 1,000 paid accounts, $34,430 MRR, 9.0% middle-tier mix. Real operator data should replace every one of those before the model drives a decision.
• Assumptions: that the new fence is one existing customers would have paid for, that grandfathered accounts do not churn faster, and that migration happens on renewal across a twelve-month cycle. Each is testable and each can be wrong.
• Sensitivity: the model breaks even at 57 middle-tier accounts and is accretive at 100. The dominant variable is mix, not price. A 10-point mix swing moves middle-rung MRR by $3,500 here, more than any plausible price adjustment does.
• What this misses: competitive response, sales compensation distortion, support load from the new admin fence, the risk that a published Scale rung cannibalizes custom contracts that would have priced higher, and the assumption of uniform elasticity across the base, which never holds.

The prize is sized by the acquisition-cost gap. Benchmarkit’s 2025 medians put the expansion CAC ratio at $1.00 against a new CAC ratio of $2.00 (a computed 2x advantage), with expansion ARR at 40% of total new ARR and net revenue retention at a 101% median. A ladder that moves accounts upward buys revenue at half the cost of one that does not, and the recovery clock on that spend is the subject of CAC payback period, the SaaS metric that matters.

---

The bear case: what the skeptics get right

The strongest objection is not that a ratio here is wrong. It is that the ladder is the wrong structure for a meaningful share of software businesses, and that the psychology underneath it is thinner than the confident version admits.

For some businesses the ladder is theater. Pure usage products, infrastructure sold on a rate card, and enterprise software where every deal is bespoke do not have a ladder in any operative sense. They have a rate card and a negotiation, and the public tiers exist to qualify inbound traffic.

The choice-effect evidence is mostly lab evidence. Simonson and Tversky (1992) and Huber, Payne and Puto (1982) replicate well in experimental settings, but field evidence for pricing-page decoys at software scale is thin, for a structural reason: vendors who A/B test pricing pages almost never publish results. Treat the anchor rung as a well-supported prior, not a measured effect size for your card.

Wide steps can be a symptom rather than a strategy. A 5x jump might reflect a genuinely different buyer, or a vendor that never built the product between the two rungs. Reading Claude’s 5.00x Pro-to-Max step as sophisticated anchoring assumes an intent the card does not disclose; the same number is consistent with usage tiers priced as literal multiples of an allowance.

Adding a rung below can expand the market rather than cannibalize it, which the protect-the-anchor instinct gets backwards. The consumer proof case is the ad-supported streaming tier, where a cheaper rung brought in customers who were not in the market at the old price, examined in why the Netflix ad tier changes everything.

Weighing it honestly: the bear case is right that the ladder is a structure rather than a law, and that a published card can be decoration on a negotiated business. It is not right that fencing and step discipline are optional. Even a business that negotiates every deal needs a defensible list price, because that is where the negotiation starts.

---

Where this is vulnerable and what it misses

Four specific holes, named plainly.

The sample is five cards, on one day. These are large, mostly product-led vendors, not a random sample of software, and July 28, 2026 is a snapshot. Cards change without notice, and a promotion (Slack was running 50% off for three months on monthly billing for new teams at capture) distorts what a card appears to say.

Some annual figures are not published, so part of the discount band is inferred. Notion states “up to 20%” without rendering a per-seat annual price, and Figma does not display distinct annual Professional rates. Applying Notion’s stated saving to list prices would suggest roughly $8 and $16 per seat, but that is an illustrative derivation, not a quoted price.

AI consumption pricing dissolves the rung itself. When the meter is credits, a tier becomes a credit bundle and the step ratio is arithmetic rather than psychology. Claude’s Max sells as 5x and 20x usage, Figma allocates 3,000 monthly AI credits to a Professional Full seat, and Notion prices agents at $10 per 1,000 credits. If seats keep giving way to consumption, the fencing chapter survives and the step-ratio chapter partly does not.

Procurement defeats anchoring, and multi-year agreements freeze the card. Anchoring works on a buyer reading a page, far less well on one with a benchmark spreadsheet and a mandate to extract 20%. A three-year agreement then locks your card against your own repricing for the term, the exposure Simon-Kucher’s 2025 finding describes at portfolio level.

What would falsify the thesis: a credible field test showing that removing the anchor rung leaves middle-tier mix unchanged, or a broad shift in which fenced tiers are replaced by pure meters with no packaging layer above them. Neither has happened in the cards examined here.

---

What operators should take from this

The instinct is to treat pricing tiers as a packaging exercise finished once and revisited when growth stalls. That is backwards. The ladder is the machine that converts acquired customers into expanding ones, and expansion is the cheapest revenue on the card at a computed 2x efficiency advantage over new logos (Benchmarkit, 2025). Six moves for this quarter:

1. Score your card on the Ladder Test before touching a number. Below six of eight, the ladder is your growth constraint, and the failing rung is almost always the middle.
2. Assign a job to every rung, in writing, without comparatives. If the only description you have is “more than the one below,” it has no job. Rewrite it around the buyer it serves, then check that buyer exists in your funnel.
3. Audit fences by naming the blocked person. Feature fences erode as competitors ship; seat-role and admin fences hold. Figma’s 18.0x seat spread (computed) is what a durable fence looks like in numbers.
4. Fix step ratios before you fix prices. Merge or re-fence anything under 1.5x in B2B; add a rung wherever the gap exceeds 5x. Ratios are the diagnostic, absolute numbers are downstream.
5. Reprice the annual discount as a cost of capital. Write down what a year of prepaid cash and one cancellation decision instead of twelve is worth to you, then compare it to the discount you give. A large gap either way means the card is mispriced against your own balance sheet.
6. Ship to new logos first, migrate on renewal, instrument mix weekly. Grandfather generously and publish the fence rather than the increase. A repricing that is revenue-neutral in its downside case ships without a governance debate.

---

How the pieces fit together

A price ladder is a sequence that only works in order: choose the metering model first, assign a job to each of the four rungs, build the fence before you pick the number, set step ratios inside the observed bands, price the anchor to be credible and legible, and treat the annual discount as financing rather than a discount.

Three arbitrary numbers on a page will always look like a price card. Four jobs, four fences and four legible steps behave like one. The difference shows up two quarters later, in whether anyone climbs.

---

Analysis, not investment advice. Prices are drawn from the public vendor pricing pages cited above, captured July 28, 2026, and are subject to change; academic findings are cited by journal, volume and year. All ratios, discounts and counts labeled computed are siliconcent calculations from those published figures. Frameworks here, including the Tier Ladder Blueprint, are for understanding SaaS business models and tradeoffs, not for making buy or sell decisions.

Want the full toolkit for pricing work like this, the tier-mix worksheet, the fence audit, and the Tier Ladder Blueprint template used above? It’s in the Tech Business Analysis Playbook.