Durability
Durability
The year-10 gate asks one binary question: will revenue and adjusted free cash flow be higher in ten years than today, with very high conviction? Docusign clears the mechanical disqualifier — revenue has risen every fiscal year since inception, so the three-year-decline flag reads false. But the conviction sources Ruchir prizes — regulatory entry barriers, capital intensity, structural monopoly — do not apply to an asset-light software leader, and Docusign's own FY2026 10-K names AI substitution as a threat to its foundational capability. That leaves a genuine doubt on the revenue side.
The conviction sources, graded for this company
Ruchir's year-10 conviction comes from five specific sources. Graded honestly against Docusign, most do not apply.
Market structure — a leading share in a low-barrier market, not a protected duopoly. Docusign is the e-signature category leader: over 1.8 million customers and more than a billion users as of January 31, 2026, with roughly 280,000 direct enterprise and commercial customers [1]. The Business tab details the share picture; the durability-relevant fact is what kind of structure it is. Docusign's filing names Adobe as its "primary global competitor for eSignature" and then a long tail — "other global software companies that have or may elect to include basic electronic signature capability in their products," and vendors focused on specific industries or product areas [2]. This is a brand-and-integration lead in a fragmented, low-switching-cost market, not the regulated monopoly/duopoly (banks, insurers) where a garage startup cannot take share. Applies weakly.
Regulatory entry barriers — the regime enables the category, it does not protect Docusign. The legal scaffolding for e-signature is the ESIGN Act and UETA in the U.S. and eIDAS in the EU — the framework that defines the "electronic signature" Docusign relies on [3]. As the 2018 prospectus put it, "both ESIGN and UETA establish that electronic records and signatures carry the same weight and legal effect as traditional paper documents and handwritten signatures," and the passage of the ESIGN Act in 2000 "solidified the legal landscape" for the whole category [4]. Crucially, these laws are technology-neutral and vendor-neutral: any compliant provider can offer a legally binding signature. Regulation lowers the barrier to the category rather than raising a wall around Docusign. Does not apply.
Capital intensity as a moat — absent. Ruchir's thesis that capital-heavy essentials survive the AI world does not fit an asset-light SaaS business. Docusign leases all of its facilities and owns no real property; its headquarters is roughly 141,000 square feet under lease [5]. Capital expenditure was $106.4 million in FY2026 against $3.22 billion of revenue — about 3.3% — funding capitalized software and data-center operations, not an irreplaceable physical network [6]. There is no replacement-cost or asset-base moat here. Does not apply.
Essentialness — real but discretionary at the margin. E-signature is embedded in customer workflows across real estate, financial services, insurance, and healthcare, and integrations and templates create switching friction. Revenue diversity is genuine: no single customer accounted for more than 10% of revenue in FY2026 [7]. But the honest measure of essentialness is expansion: dollar net retention has fallen from roughly 115% at the IPO to a recent 102%, meaning existing customers now expand only marginally faster than they churn (see the retention chart below). Essential enough to be sticky; not utility-grade. Applies moderately.
Operating history — moderate. Docusign was incorporated in 2003 and reincorporated in Delaware in 2015; it has been public since 2018 [8]. That is roughly 23 years of existence and one full demand cycle survived — the COVID pull-forward and its unwind — well short of the 30–50-year survival record Ruchir treats as conviction-grade. Applies weakly.
The structural threats, hunted
The self-check here is not perfunctory: the most material year-10 threat is named by Docusign itself, and it is exactly Ruchir's "is anyone's margin here an Amazon opportunity?" test.
AI substitution — the company's own filing says its foundational capability is becoming cheaper to replicate. The FY2026 10-K's competition section escalated the language sharply: "As AI advances make some foundational capabilities for agreements (such as analyzing, summarizing, comparing, and acting upon text) cheaper and easier to replicate, our eSignature and Intelligent Agreement Management platform may also increasingly compete with providers of LLMs, data platform companies and other enterprise software companies, and with homegrown solutions developed internally by customers" [9]. The dedicated risk factor is blunter still: "advances in AI may significantly lower the cost of developing software, enabling companies to quickly and cheaply create agents or other homegrown alternatives," and if LLM providers, "hyperscalers," or others "develop solutions that provide comparable functionality at lower cost or in more convenient formats, demand for our products would suffer" [10]. The FY2026 filing goes as far as naming "vibe coding" — software created from natural-language prompts — as a competitive substitute [11]. This is a company telling its owners that the core function — putting a legally binding signature and workflow on an agreement — is on a cost curve that AI is collapsing. On Ruchir's test, Docusign's own margin is the Amazon opportunity.
Quantifying the year-10 exposure. The threat lands on the base, not the growth edge. Docusign's AI-native pivot, Intelligent Agreement Management (IAM), reached about 12.6% of total company ARR by Q1 FY2027, with roughly 40,000 companies on the platform [12]; management targets IAM at approximately 18% of ARR (over $600 million) by the end of FY2027 [13]. Even if that target is met, roughly four-fifths of ARR remains legacy eSignature and CLM — precisely the "foundational capability" the 10-K flags as replicable. The substitution risk therefore applies to the large majority of the revenue base, with no capital, regulatory, or structural-monopoly moat to blunt it — only brand, integration depth, and product execution. And execution is not a year-10 moat: a company that merely out-builds competitors has protection only for as long as it keeps out-building.
The threat is already in the price, and in sell-side models. The stock fell from a peak of $106.99 (December 2024) to a trough of $41.75 (February 2026), a 61% drawdown, driven substantially by this AI-disruption overhang; the anatomy is in Dislocation. External analysts reset targets on an explicit "AI risk framework" — Jefferies moved to Hold with a price target cut from $105 to $45 in March 2026, and other desks trimmed similarly. The bear mechanism they cite is specific: AI agents that read, negotiate, and act on agreements could erode the seat-based licensing on which Docusign's model rests.
Regulatory reversal, customer concentration, pricing pressure. Regulatory reversal is a low year-10 risk — the ESIGN/UETA/eIDAS framework that legitimizes electronic signatures is entrenched, and evolving EU trust-services rules cut toward compliance cost rather than existential threat [14]. Customer concentration is not a threat — the book is diversified with no customer over 10% [15]. Pricing pressure is real: Adobe bundles Acrobat Sign into Acrobat subscriptions rather than selling it standalone, and Box and Dropbox fold signing into their storage suites — the competitive question the Competition tab frames as whether signing gets absorbed into a document suite for free.
The disqualifier check — revenue history
Ruchir's one mechanical disqualifier: revenue declining high-single-digit for three consecutive fiscal years after a long existence. Docusign does not trip it.
Source: fit_features.revenue_trajectory, derived from filed Consolidated Statements of Operations; FY2026 figures cross-checked to the FY2026 10-K [16].
Revenue has risen in every one of the ten fiscal years on record. The fit_features.revenue_trajectory flag confirms it: consecutive_decline_years is 0 and three_year_hsd_decline is false. The structural-decline test (X3) is therefore checked and absent — there is no realized revenue decline to point to.
What the flag does not capture, and what belongs here plainly, is the shape: year-over-year growth decelerated from roughly 49% (FY2021) to about 8% (FY2026). That is deceleration, not decline — a maturing category rather than a shrinking one — but it removes any claim that revenue growth is structurally protected. The falsifier to watch, in Ruchir's own terms, is the first year revenue growth turns negative; nothing in the record to date shows it.
The year-10 case, both ways
The strongest case that year-10 revenue and adjusted FCF are higher. Docusign enters the decade as category leader with 1.8 million customers, deep workflow integrations, and the brand that made "e-signature" synonymous with legality [17]. The balance sheet is net cash — $866.5 million in cash and short-term investments, additional long-term investments, and no borrowings under its credit facility [18], so it can outlast a long transition. Reported free cash flow has risen every year to $1.06 billion in FY2026 [19], and consensus (per fit_features.consensus_forward_yield) models FCF continuing up from roughly $854 million (FY2025) toward $1.33 billion by FY2029 — the sell side does not forecast decline. The secular tailwind (paper-to-digital agreements) is intact, and management is repositioning the platform to sell AI rather than be displaced by it, with IAM growing double-digits and platform partnerships spanning Anthropic's Claude and Salesforce's Slack [20].
The strongest doubt. Docusign's own FY2026 10-K states that AI is making the foundational capability underpinning most of its revenue "cheaper and easier to replicate" by LLM providers, hyperscalers, and customers' homegrown agents [21] [22]. That threat sits on roughly four-fifths of ARR, is defended only by brand, integration, and execution — none of which is a structural moat — and arrives as growth has already decelerated to ~8% and expansion (dollar net retention) has flattened to ~102%. The consensus that models rising FCF looks out three years; the gate asks about ten.
The read. The mechanical disqualifier does not fire, and higher year-10 revenue and adjusted FCF are a defensible base case for the next few years. But P1 does not ask whether growth is likely — it asks whether it is near-certain. Here there is a genuine, company-acknowledged doubt on the revenue side: AI-driven commoditization of the core signing-and-agreement function, against a business with no capital-intensity, no regulatory, and no monopoly protection. That doubt is exactly the kind the gate is built to catch, so the year-10 gate does not hold with very high conviction — the genuine doubt is AI substitution of the foundational capability, flagged by Docusign's own filings. The counter-evidence adjacent to that read — a net-cash balance sheet, still-rising reported FCF, and a credible AI pivot — is real, and is why the doubt is a doubt rather than a conclusion of decline.
FCF consistency (P2)
The profile's basis is adjusted ("real") free cash flow — reported FCF less stock-based compensation less the trailing five-year average of acquisition spend. The deterministic feature that would grade its stability, fit_features.fcf_stability, is not computable: stock-based compensation was absent from the structured cash-flow feed, so the profile could not build the rolling five-year adjusted-FCF series (recorded in data gaps). The primary filings do carry SBC, so the series can be reconstructed from them — shown below as illustrative, not as the deterministic feature.
Illustrative adjusted FCF = reported FCF − SBC − 5-yr avg acquisitions, derived from filed Consolidated Statements of Cash Flows (FY2022–FY2026 10-Ks) [23] [24] [25]. The deterministic fit_features.adjusted_fcf series is not computable (SBC absent from the structured feed).
Two things stand out. First, reported FCF is genuinely stable and rising — positive and growing in each of the last five years, with one negative episode (FY2017, pre-IPO). That negative was a young-company year, not a business-model-inherent underwriting loss on a 5–8-year cadence; there is no insurance- or bank-style cyclicality here, because subscription software cash flow recurs rather than cycles. Second, on the profile's real basis the record is much shorter and thinner: adjusted FCF was roughly breakeven in FY2022, negative by about $147 million in FY2023, and only turned solidly positive from FY2024 — as reported FCF outgrew a roughly flat ~$610–622 million SBC load. Stock-based compensation, at about 19% of revenue, is the swing factor; the SBC add-back is visible in the FY2026 non-GAAP reconciliation [26].
So P2 reads two ways. Predictability is not the problem — the cash flows recur and are not volatile. The problem is that the positive adjusted-FCF history is only about three years long, and its stability depends on reported FCF continuing to outrun an elevated SBC charge. That FY2026 adjusted figure of roughly $406 million against the ~$9.84 billion market cap frames a ~4.1% real yield — the Yield tab computes the yield against the bar. For durability, the point is narrower: adjusted FCF is consistent and rising now, but has too short a positive track record to describe as a proven five-year-stable series, and it rests on SBC not re-inflating.
The dollar-net-retention record
Sources: ~115% at IPO, 2018 prospectus [27]; 98% trough, Q4 FY2024 call [28]; 102% recent, Q1 FY2027 presentation [29].
Dollar net retention is the cleanest single gauge of essentialness for a subscription business: it measures whether the installed base expands or contracts. Docusign's fell from ~115% at the IPO to a 98% trough in Q4 FY2024 — below 100%, meaning the base was shrinking net of churn — before recovering to ~102% through Q1 FY2027. The recovery is real and supports the essentialness case; the level, barely above breakeven, is why essentialness grades moderate rather than utility-grade. Whether the base keeps expanding as AI reshapes the category is the year-10 question in one number.