Build Expansion Into the Product Logic, Not the Sales Motion

Law 6

69% of benchmark companies

The old playbook: NRR as a metric for the CS team to manage at renewal. What the best companies did: build expansion into the product's own logic so that as customers use the product more, they pay more — automatically. Sierra and Decagon: more support interactions means more revenue. No upsell call needed. Deel: more employees at the customer company means more EOR/contractor revenue. Automatic. Ramp: more transactions means more interchange and seat revenue. Structural. Harvey: one workflow proven, then the adjacent department, then the next — seat expansion follows adoption, not the renewal calendar. Glean: departmental pilots at $60K expanded to enterprise-wide deployments at $300–500K+. The NRR data across the cohort: eleven of sixteen companies exceeded 120%. Writer achieved 209% NRR in early 2024. Hebbia showed inferred NRR above 200% from seat expansion patterns. Gong reached 140% historically. After the first 12–18 months, expansion revenue in these companies materially outpaced new-logo acquisition. The machine became self-financing. The fragility case: Gong's 2023 NRR deceleration — from strong historical levels to 16% ARR growth — when SaaS hiring froze shows the exposure of pure seat-based models to headcount contractions. Outcome-based and per-employee models showed more stable NRR through the same period. The distinction between structural NRR (Deel, Ramp, Sierra, Decagon) and sales-motion NRR (Harvey, Glean, Hebbia) matters for capital allocation. A company with structural NRR needs fewer expansion salespeople per dollar of NRR generated. A third variant — consumption-metered expansion — is visible in Cognition (Devin). ACU (Agent Compute Unit) billing means revenue scales with usage intensity: as engineers run more Devins on more tasks, ACU consumption grows multiplicatively (per-engineer usage × team size × task scope expansion). The >5x and >10x contract expansions reported at Cognition happen through increased consumption, not seat growth or interaction volume. This is structurally analogous to cloud-computing expansion (AWS, Snowflake) and may represent a distinct expansion mechanism for usage-based AI products.

Key examples

harvey sierra glean hebbia moveworks deel writer cognition

Anti-pattern

Treating NRR as a vanity metric rather than a business model input. Underinvesting in post-sale because "the product sells itself." Seat-based pricing in headcount-sensitive categories without expansion vectors that are independent of headcount.

Cross-Company Comparison

NRR, expansion mechanisms, and structural vs. sales-motion expansion

Company	NRR	Expansion mechanism	Type
Deel	120%+ (every year since inception — confirmed)	Headcount growth triggers automatic EOR/payroll revenue; no upsell call needed	Structural (headcount-indexed pricing)
Sierra	>120% (inferred; not disclosed)	More channels = more interactions = more resolutions = more revenue; voice launch 2024 multiplied cohort revenue 3-4x without new contracts	Structural (outcome-indexed pricing)
Writer	209% (early 2024); ~160% (normalized 2025)	Embedded account team discovers adjacent departments; L'Oreal 1 brand → 21 brands → 30+ use case families	Sales-motion (high-touch CS-driven expansion)
Harvey	>130% (inferred — not disclosed)	Seat doubling within 12 months + practice area expansion + Vault workflow agents + corporate in-house expansion	Sales-motion (3 independent vectors)
Glean	140–170% (inferred; not disclosed)	Department pilot → company-wide rollout ($60K → $500K+) + Agents upsell on same knowledge graph	Sales-motion (adoption-data-triggered expansion)
Hebbia	>200% (unverified estimate — gitnux.org)	3-5 Professional seats ($10K) → Lite seat proliferation ($3K) across firm; contracts roughly double within 12 months	Sales-motion (pricing-architecture-driven expansion)
Moveworks	115–130% (inferred — never disclosed)	IT land → HR/Finance/Facilities expand; headcount growth automatically increases per-employee ACV	Mixed (headcount auto-expansion + department sales-motion)
Gong	~140% (2017–2022); collapsed to ~16% YoY growth (2023)	Seat expansion tied to sales org headcount growth — fragile when headcount froze	Fragile structural (headcount-dependent; negative case)

How This Law Worked in Practice

Evidence from each benchmark company where this law was observed — how it manifested, what the mechanism was, and what sources confirm it.

Deel

Deel's NRR is the clearest example of structural expansion in the cohort. Every time a customer hires a new employee in a foreign country, Deel's revenue from that customer grows automatically — no account manager call required, no renewal negotiation, no upsell motion. At $599/employee/month for Employer of Record, headcount growth is a direct revenue event. This is pricing architecture, not a sales strategy. The 120%+ NDR held every year since inception (confirmed by multiple sources including a16z). The mechanism has four layers: (1) existing contractor-management customers convert contractors to EOR employees as their teams mature; (2) headcount growth at existing customers compounds per-seat revenue automatically; (3) customers expand from single-product to multi-product (contractor → EOR → global payroll → HRIS) as their HR complexity increases; and (4) account management QBRs unlock latent expansion from customers who don't know what products exist. The most striking data point in Deel's history illustrates the power of the expansion motion. When Deel finally built an account management function and ran QBRs in late 2021, Shuo Wang noted: "There's a huge potential and there's a big part of the revenue that we're leaving on the table." The result: ARR went from $50M to $100M in three months — not three years. The expansion flywheel had been running at partial capacity for years because customers simply didn't know about other products. The structural element distinguishes Deel from most companies in this cohort. Even without active account management, Deel's revenue from existing customers grows as those customers grow. The sales motion amplifies a baseline that compounds on its own. Approximately 60% of Deel's revenue is cross-sell/upsell (Sacra, unverified), but the foundation is a pricing model where customer growth is automatically Deel's growth.

Key evidence

120%+ NDR every year since inception★

$50M → $100M ARR in 3 months after account management + QBR activation★

~60% of revenue from cross-sell/upsell★

EOR pricing at $599/employee/month — headcount growth = automatic revenue growth★

Shuo Wang: 'There's a huge potential and there's a big part of the revenue that we're leaving on the table'★

Sierra

Sierra's expansion model is the most structurally elegant in the cohort: outcome-based pricing means expansion revenue requires no new contract, no renewal negotiation, and no sales motion. When a customer adds a second support channel — say, voice after deploying chat — their monthly resolution volume multiplies and Sierra's revenue multiplies with it. Bret Taylor described the mechanism directly: "If the AI agent resolves the case, no human intervention, there's a pre-negotiated rate for that. If we do have to escalate to a person, that's free." The structural expansion mechanism is confirmed by Sierra's trajectory. The 2024 year-end cohort (approximately $26M ARR) expanded to over $100M by Q4 2025 — a 4x expansion in 12 months — primarily driven by existing customers adding voice channels after the October 2024 launch. Voice overtook text as the primary interaction channel within 11 months of launch. Existing customers did not sign new contracts; their resolution volumes multiplied. No NRR figure has been publicly disclosed by Sierra. The inferred structural mechanics make >120% NRR nearly certain: outcome pricing is expansion-native; voice addition multiplied resolution volumes; use-case deepening (containment → subscription retention → proactive outreach) raises volumes further. Sierra's playbook analysis infers the 2024 cohort alone could represent $78M–$104M ARR by end of 2025 purely through expansion, before new logos. Sierra's model creates a structural moat against incumbents (Zendesk, Salesforce Service Cloud) that Taylor explicitly articulates: "Closing a technology gap in your product is hard, but not impossible. Changing your business model is really hard. There's a graveyard of CEOs who've been fired for failing to make this transition." Incumbents on seat-based pricing cannot copy outcome-based pricing without destroying their existing revenue base.

Key evidence

Outcome pricing: resolved interaction triggers revenue, escalations are free★

Voice overtook text as primary channel within 11 months of October 2024 launch★

$26M ARR end-2024 → $100M+ ARR Q4 2025 — primarily expansion, not new logos★

NRR not disclosed; inferred >120% from structural mechanics★

Taylor on incumbents: 'There's a graveyard of CEOs who've been fired for failing to make this transition'★

Writer

Writer's NRR is the best-documented in the cohort: 209% in early 2024, normalizing to approximately 160% by 2025 as the customer base scaled beyond 300 logos. Both figures are confirmed by multiple sources including First Round Review (May Habib) and Contrary Research. At 160% NRR, existing customers alone drive roughly 60% of incremental ARR — new logo acquisition amplifies rather than funds growth. The expansion mechanism is CS-motion rather than structural. An embedded account team (Solution Architect + Customer Engineer + CSM) works inside the customer organization and discovers adjacent use cases before the customer thinks to ask for them. The L'Oreal account is the canonical case: entered as one brand, expanded to 21 brands, now runs 30+ use case families. May Habib's internal ambition: "This should be a 400% NRR company, the team knows that" — implying L'Oreal is the norm, not the exception. The expansion is triggered by initial team type. Writer's data shows that when marketing teams are the first adopters, NRR = 160%. This finding directly shaped homepage and trial messaging to target marketers specifically. The product tour serves different content to end-users vs. directors, pre-selling the organizational expansion story before the first sales call. The 209% → 160% NRR normalization is a leading indicator worth monitoring: early-cohort effects (small customer count, high expansion per account) normalize as the base grows. Writer's growth story changes from "existing customers fund the company" to "existing customers materially supplement new logo acquisition" — still exceptional (35+ points above enterprise SaaS median) but structurally different.

Key evidence

NRR 209% early 2024; ~160% normalized 2025★

Marketing-team-first adoption → 160% NRR★

L'Oreal: 1 brand → 21 brands → 30+ use case families★

Internal NRR target: 400% — 'the team knows that'★

Andy Shorkey: 179% YoY NRR growth (Science of Scaling, May 2025)

Harvey

Harvey's expansion model operates through three independent vectors: seat count growth, practice area expansion, and workflow agent proliferation. Median seat count doubles within 12 months of initial deployment (cited in Sacra analysis). Gross revenue retention is 98% (Sacra), which is exceptionally high for enterprise SaaS and reflects the difficulty of displacing Harvey once it is embedded in a firm's workflows, precedent libraries, and billing systems. NRR is not publicly disclosed. The analysis infers "likely well above 130%" based on: 98% gross retention + doubling seat counts + practice area expansion. The three expansion vectors are independent: a firm that doubles seats in litigation can simultaneously expand to corporate transactional work (new practice area) and deploy Vault workflow agents for M&A due diligence (new product capability). All three compound simultaneously. Product utilization data from 2024 is particularly significant: utilization jumped from 40% to 70% within existing seats, meaning expansion was not just new logos or new seats — existing users were using Harvey more intensively. This deepens switching costs and renewal confidence simultaneously. The "selling the work" model emerging in 2024–2025 — where Harvey co-builds workflows with firms and shares revenue from AI-enhanced flat-fee services — represents a fourth expansion vector that is still nascent but structurally large. Unlike Deel or Sierra, Harvey's expansion is primarily sales-motion driven. The account expansion from research to drafting to due diligence to regulatory compliance requires active engagement from Legal Engineers (former Big Law attorneys in sales roles), not automatic triggering. The 98% gross retention contains the floor; the expansion multiple above 100% is generated by people, not by product mechanics.

Key evidence

Median seat count doubles within 12 months★

Gross revenue retention: 98%★

NRR 'likely well above 130%' — inferred from seat doubling + expansion data★

Product utilization: 40% → 70% in 2024★

Vault workflow agents: 25,000+ custom agents operating on platform; 400,000+ daily agentic queries★

Glean

Glean's expansion model is data-triggered rather than purely sales-driven. The machine works as follows: a departmental pilot generates clean usage data (queries per day, DAU/MAU, success rate); the CSM brings that data to the executive sponsor; the exec approves company-wide rollout. Initial departmental pilots at approximately $60K scale to $300–500K+ company-wide within nine months. The $1M+ contract segment grew 3x in one fiscal year; company-wide deployments doubled year-over-year. NRR is not directly disclosed by Glean. Proxies from public data suggest 140–170%: initial pilots convert to company-wide ($60K → $500K+ = 8x within 9 months), $1M+ contracts are the fastest-growing segment, and Glean Agents (launched February 2025) creates a second upsell layer on top of the same knowledge graph infrastructure — no new data integration required. The Forrester TEI puts the 3-year ROI at 141% with a payback period under six months, which makes expansion approval straightforward for executive sponsors. Glean's architecture creates a compounding data moat that makes expansion both natural and sticky. The Enterprise Graph (organization-wide knowledge) and Personal Graph (individual activity patterns) mean the platform becomes more useful as more employees onboard. Company- wide deployments are stickier than departmental ones because the network effect kicks in — every person using Glean improves retrieval quality for everyone else. "If I can't find it on Glean, it doesn't exist" is the behavioral signal that precedes the expansion conversation. Unlike Sierra or Deel, Glean's expansion is not automatic — it requires a CSM-led conversation using adoption data. But the data is self-generating (search produces its own metrics), and the expansion argument (here is your adoption at 80%, here is what company-wide would look like) is grounded in proof rather than aspiration.

Key evidence

Initial $60K pilot → $300–500K+ company-wide within 9 months★

$1M+ contract segment grew 3x in one fiscal year★

Company-wide deployments doubled YoY★

NRR inferred 140–170% from expansion mechanics★

Forrester TEI: 141% ROI, <6 month payback★

wDAU/wMAU 40% — 2x SaaS industry benchmark★

Hebbia

Hebbia's expansion model is built into the pricing architecture. Entry is 3–5 Professional seats at $10,000/year each — paid by senior analysts and associates who build workflow agents. Once those agents are embedded into daily processes, Lite seats proliferate across the enterprise at $3,000–$3,500/year. The NRR is reported as >200% (gitnux.org, unverified), implying contracts roughly double in size within 12 months. The $500K average ACV as of 2024 (confirmed by TechCrunch and Series B announcement) implies Lite seat expansion from initial Professional-only deals. The Bloomberg Terminal anchor ($10K/seat is already a budgeted line item in every finance firm) eliminates the pricing objection for the initial Professional seats. The Lite seat proliferation then follows from usage: when Professional users have built agents that the whole team needs to run, the economic case for Lite seats is self-evident. Danny Wheller (VP Business & Strategy) describes it directly: "Once Professional users have embedded agents into daily processes, Lite seats proliferate across the enterprise, creating a durable, high-margin annuity stream anchored by deep workflow integration rather than generic search." Forward-deployed AI Strategists (ex-bankers and lawyers) accelerate this motion. Their job is not traditional customer success — it is to design templates, configure agents, and drive Professional seat holders to build and share use cases. They own the expansion motion: Lite seat proliferation across adjacent teams, starting from the PE due diligence team to credit, M&A advisory, and complex litigation. Gross retention is reported as >90% among top asset managers (Sacra). Oak Hill Advisors documented 6x ROI and 75% reduction in review times — the kind of documented outcome that makes churn uneconomical and expansion approval straightforward. The NRR >200% figure, if confirmed, would make Hebbia the highest-NRR company in this cohort.

Key evidence

NRR >200%★

Gross retention >90% among top asset managers★

Average ACV $500K (confirmed TechCrunch / Series B announcement, 2024)

Pricing: Professional $10K/seat → Lite $3–3.5K/seat; contracts roughly double in 12 months★

Danny Wheller: 'Once Professional users have embedded agents into daily processes, Lite seats proliferate across the enterprise'★

Oak Hill Advisors: 6x ROI, 75% reduction in review times★

Moveworks

Moveworks' expansion model is mixed: one vector is structural (headcount-based pricing), the other requires active sales effort (department expansion). The per-employee pricing means that as customer headcount grows, Moveworks' revenue from that customer grows automatically. A customer with 5,000 employees that grows to 7,000 generates 40% more Moveworks revenue without any upsell call. The 90%+ whole-org deployment rate means this structural vector applies broadly — most customers deploy Moveworks to the entire company, not just IT. The department expansion vector — IT → HR → Finance → Facilities — requires the active engagement of account managers using QBRs and champion referrals. The typical trigger: CIO shares IT metrics with CHRO; CHRO initiates HR use case. Creator Studio (2023) opened arbitrary new use cases through customer-built no-code workflows, deepening lock-in and creating an expansion surface that account managers could activate. NRR was never disclosed publicly. The inference from public data is 115–130%: 90%+ whole-org deployment rate implies minimal mid-contract churn; land-and-expand by department adds ACV; headcount growth at customers adds revenue automatically. This is a lower bound than Sierra or Deel because Moveworks' price-per-employee creates a ceiling — you cannot expand revenue above 100% of total employees. The Forrester TEI documents $11.5M benefit over 3 years with a 256% ROI and sub-1-year payback for a 30,000-employee organization — the same type of documented ROI that makes expansion approval easy and churn uneconomical. ServiceNow's $2.85B acquisition at ~20x ARR is partly a validation of NRR quality: a business with high churn does not command premium multiples.

Key evidence

90%+ whole-org deployment rate★

Inferred NRR 115–130% — not publicly disclosed★

Flat per-employee pricing: headcount growth = automatic ACV growth★

Forrester TEI: 256% ROI, $11.5M benefit over 3 years, <1 year payback★

Land IT → expand HR/Finance/Facilities → Creator Studio (2023) for arbitrary use cases★

Gong

Gong is the cautionary case in this cohort. From 2017 to 2022, Gong operated at approximately 140% NDR (Sacra, unverified estimate): every year, the installed base generated more revenue than the prior year, primarily because sales teams grew and new reps required new seats. This is a structural expansion mechanism — as the customer's sales organization grows, Gong's revenue from that customer grows automatically without active upsell. The fragility was latent and structural: seat-based expansion tied to sales org headcount means Gong's NRR depends on a factor it cannot control — its customers' hiring decisions. The Gong playbook analysis describes what happened when the model hit its structural limit: "Sales hiring freezes across SaaS companies in 2023 directly eliminated Gong's seat-expansion revenue. The ~140% NDR degraded because customers contracted headcount, not because they churned. The seat-based model had a structural dependency on sales headcount growth." YoY growth collapsed to approximately 16%. The lesson is not that seat-based models are bad. It is that seat-based models tied to a specific customer function (sales teams) are exposed to that function's macro cycle. When the SaaS sector cut sales headcount 15–25% in 2023, Gong's expansion model was directly and immediately impaired. Gong's CEO Amit Bendov acknowledged the issue and announced a pivot toward value-based pricing: "If you're an AI company and you're selling based on seats, you're leaving a ton of value on the table." The distinction between Gong and Deel is instructive: both are seat/employee-based models. But Gong's seats track the customer's sales team size; Deel's seats track the customer's total headcount (specifically, international employees). Sales headcount is a discretionary cost that compresses in downturns; total employee count is stickier. The expansion flywheel's durability depends on what economic variable drives the expansion trigger.

Key evidence

~140% NDR (2017–2022) — Sacra, unverified estimate★

Sales headcount freeze 2023 → NDR collapsed → growth fell to ~16% YoY★

Seat-based model dependency confirmed: 'customers contracted headcount, not because they churned'★

Bendov: 'If you're an AI company and you're selling based on seats, you're leaving a ton of value on the table'★

← Law 5: Price Against Labor Cost, Not Software Alternatives All Growth Laws ↑