Overview
DisputeNinja was a San Francisco-based fintech startup founded in 2023 by Max Wu and Bowen Xue. [1] Accepted into Y Combinator's Summer 2023 batch, the company built AI agents that automated chargeback evidence collection and dispute response drafting, targeting software and AI startups as its primary customer segment. [2] The two-person team operated out of San Francisco with no disclosed external hires and no publicly announced revenue milestones.
DisputeNinja's core value proposition was rendered structurally non-viable by its primary infrastructure dependency: Stripe, the dominant payment processor for its target customers. In June 2025, Stripe introduced a $15 counter fee for merchants who manually contest chargebacks using third-party tools — waived only for users of Stripe's own native product. [3] Five months later, Stripe launched Smart Disputes, an AI-powered chargeback automation tool trained on $1.4 trillion in annual payments volume, directly replicating DisputeNinja's product. [4]
No formal shutdown announcement has been published. The DisputeNinja website remains live, and both founders' LinkedIn profiles still list the company as active. [5] The absence of any follow-on funding, press coverage, or community discussion — combined with Stripe's platform moves — points to a company that quietly became non-viable rather than one that closed with fanfare.
Image 1 / 1
Founding Story
Max Wu and Bowen Xue came to chargeback automation through direct, painful experience. Before founding DisputeNinja, both worked at startups operating in the NFT and cryptocurrency space — a sector notorious for payment fraud. [6] At those companies, the founders discovered that fraudulent NFT sellers accounted for roughly 90% of fraud cases, yet the existing fraud-prevention vendors they relied on had no meaningful solution for the problem. [7] The vendors were built for traditional e-commerce patterns, not the fast-moving, pseudonymous dynamics of crypto marketplaces.
Wu graduated from UCLA, while Xue studied at the University of Waterloo — Waterloo being a well-established pipeline for technical founders in North America. [8] [9] Xue had previously worked as a software engineer at Paper, an education technology company, before co-founding the company. [10] Neither founder had a background in payments or financial services specifically — their expertise was in building software products and fighting fraud operationally, not in the regulatory or banking infrastructure that underlies chargeback processes.
The founding insight was experiential rather than analytical: the founders had lived the problem of inadequate fraud tooling and believed they could build something better. This is a common and often reliable origin story for B2B software companies — the founders are their own first customer. The specific pivot from general fraud prevention to chargeback automation appears to have been a product focus decision made around the time of YC application, though the precise timing is not documented publicly.
What makes the founding story structurally unusual is that Wu and Xue simultaneously launched two companies under the same YC batch: DisputeNinja (chargeback automation for software startups) and Trench (open-source fraud prevention for marketplaces). [11] Both were accepted into S23, which consisted of 247 companies selected from over 24,000 applications. [12] Whether this dual structure reflected a deliberate portfolio strategy, a hedge against product-market fit uncertainty, or simply two complementary tools within a single fraud-tooling thesis is not documented. YC's public announcement in August 2023 focused on Trench, describing it as an open-source, self-hosted fraud prevention platform for marketplaces — with DisputeNinja's chargeback product operating in parallel. [13]
Timeline
- 2023 — Max Wu and Bowen Xue found DisputeNinja and Trench in San Francisco, drawing on prior experience fighting payment fraud at NFT and crypto startups. [14]
- 2023 — Both DisputeNinja and Trench are accepted into Y Combinator's Summer 2023 (S23) batch, receiving standard pre-seed investment. [15]
- 2023-08-03 — YC publicly announces Trench as an open-source fraud prevention platform for marketplaces; DisputeNinja's chargeback automation product is also live at this time. [16]
- 2023 — Trench raises $500K in a single funding round from one investor. No separate external funding round for DisputeNinja is identified in any database. [17]
- 2025-03-27 — Stripe announces a new $15 dispute counter fee for merchants who manually contest chargebacks, effective June 17, 2025, waived only for merchants using Stripe's native Smart Disputes tool. [18]
- 2025-06-17 — Stripe's new fee structure takes effect; merchants using third-party chargeback tools now pay an additional $15 per contested dispute. [19]
- 2025-11 — Stripe launches Smart Disputes, an AI-powered native chargeback automation tool trained on $1.4 trillion in annual payments volume, directly replicating DisputeNinja's core value proposition. [20]
What They Built
DisputeNinja's product addressed a genuine operational pain point: chargeback management is labor-intensive, time-sensitive, and disproportionately burdensome for small software companies. When a customer disputes a charge with their bank, the merchant typically has 7–21 days to submit a response with supporting evidence — transaction logs, email correspondence, IP addresses, usage data, and signed terms of service. For a two-person startup, assembling this evidence package for every dispute can consume hours per case.
DisputeNinja's stated solution was to automate the entire evidence collection and response drafting workflow using AI agents. [21] The product worked in two modes: API integrations that connected directly to a merchant's existing software stack (payment processor, CRM, email platform) to pull relevant transaction data automatically, and browser agents that could navigate web interfaces to gather evidence where APIs were unavailable. [22] Once evidence was collected, the system drafted a formatted chargeback response document ready for submission to the payment processor.
The company's demo page positioned the product explicitly for a specific customer type: "Chargeback automation that actually wins. Built for software and AI startups." [23] This framing was deliberate — software and AI startups sell digital goods and subscriptions, which have different chargeback patterns than physical goods merchants. Digital goods disputes often hinge on proving service delivery and usage, rather than shipping confirmations, which plays to the strengths of an API-connected evidence collection tool.
DisputeNinja also offered a free-tier AI Chargeback Response Generator at a separate subdomain, which allowed any merchant to generate a chargeback response document in minutes without a full product subscription. [24] This free tool likely served as a top-of-funnel acquisition mechanism — a common PLG (product-led growth) pattern for B2B tools targeting developers and technical founders.
The product's key differentiator, relative to manual dispute management, was the labor efficiency claim: a single analyst could handle what previously required an entire team. [25] Relative to other chargeback automation competitors (Chargeflow, Disputifier, Midigator), DisputeNinja's differentiation appears to have been its explicit focus on software and AI startups rather than e-commerce merchants — a narrower ICP (ideal customer profile) that may have been intended to reduce competition for early customers.
No product changelog, feature history, or customer case studies are publicly available. The product's evolution between its 2023 launch and the present is undocumented.
Market Position
Target Customers
DisputeNinja targeted software and AI startups — companies selling digital subscriptions, SaaS products, and AI-powered tools. [26] This segment has several characteristics that make chargeback automation appealing: high transaction volumes relative to headcount, limited finance operations staff, and a customer base that is more likely to dispute charges (given the intangible nature of digital goods). The segment is also disproportionately concentrated on Stripe as a payment processor, which — as discussed below — proved to be a critical vulnerability rather than an asset.
Market Size
The global chargeback management software market is a subset of the broader fraud prevention and payment security industry. Chargeback volumes have grown significantly with the expansion of card-not-present (CNP) transactions, and the problem is structurally persistent: card network rules require merchants to respond to disputes or forfeit the funds automatically. For software companies, chargeback rates above 1% can trigger processor penalties or account termination, creating a compliance dimension on top of the revenue recovery motivation. DisputeNinja did not publish any market size estimates, and no third-party market sizing specific to their segment was identified in available sources.
Competition
DisputeNinja entered a market with established players and a structural vulnerability that became decisive.
Established third-party tools — Companies like Chargeflow, Disputifier, and Midigator had already built chargeback automation products before DisputeNinja launched. These competitors had more mature products, larger customer bases, and more evidence of win rates. DisputeNinja's differentiation — focus on software and AI startups — was a positioning choice, not a technical moat. Any of these competitors could have targeted the same segment.
The platform incumbency axis — The most important competitive dimension was not product depth or pricing, but distribution and data. Stripe processes payments for the overwhelming majority of DisputeNinja's target customers. Stripe has access to every transaction, every dispute, every piece of evidence that a merchant would need to submit — not as an integration, but natively. A third-party tool like DisputeNinja had to reconstruct this data picture through API calls and browser agents. Stripe already had it. This asymmetry was always present; the question was whether Stripe would choose to compete.
Stripe's platform moves — Stripe answered that question in two stages. First, in March 2025, Stripe announced a $15 counter fee for merchants who manually contest chargebacks, effective June 17, 2025, with the fee waived only for users of Stripe's own Smart Disputes product. [27] This fee restructuring did not merely add a cost to using DisputeNinja — it created a direct financial incentive to switch to Stripe's native tool. For a startup paying $15 per contested dispute on top of DisputeNinja's subscription fee, the economics of using a third-party tool deteriorated sharply. Second, in November 2025, Stripe launched Smart Disputes itself — an AI-powered chargeback automation tool trained on $1.4 trillion in annual payments volume. [28] The training data advantage alone is insurmountable for a two-person startup: DisputeNinja's AI agents were trained on whatever evidence its small customer base generated; Stripe's system was trained on the dispute history of millions of merchants.
The competitive map, plotted along the axes that mattered most — distribution reach versus data advantage — placed DisputeNinja in the worst possible quadrant: low distribution (no existing merchant relationships at scale), low proprietary data (dependent on customer-provided integrations), competing against an incumbent with maximum distribution and maximum data.
Business Model
DisputeNinja's revenue model is not publicly documented. The company never disclosed pricing, revenue figures, or customer counts. The absence of any revenue data is itself a signal: companies that achieve meaningful traction typically publish case studies, win-rate statistics, or at minimum, pricing pages that imply a monetization structure.
Based on the product description and standard patterns in the chargeback automation category, the most likely revenue models were: a monthly subscription fee per merchant account, a per-dispute success fee (a percentage of recovered chargeback amounts), or a hybrid of both. The free-tier AI Chargeback Response Generator suggests the company was using a freemium model to acquire users, with conversion to a paid tier for full automation.
Estimated unit economics (inference, not fact): The YC S23 standard pre-seed investment was approximately $500K. [29] For a two-person team in San Francisco, with no disclosed office lease and a software-only product, annual burn likely ranged from $200K–$350K (salaries, infrastructure, and YC program costs), implying 18–30 months of runway from the initial investment alone — assuming no revenue. If Trench's separate $500K raise was accessible to the same founding team, total capital could have extended this runway further, though the legal and entity structure between the two companies is unknown. [30]
The absence of a Series A or any follow-on round for DisputeNinja specifically is consistent with a company that did not demonstrate the revenue growth metrics — typically $1M+ ARR with strong month-over-month growth — required to raise in the 2023–2024 funding environment.
Post-Mortem
Primary Cause: Platform Squeeze by Stripe
The most structurally decisive factor in DisputeNinja's non-viability was Stripe's two-stage competitive move in 2025. This was not a case of a startup being outcompeted by a better product in a fair market — it was a platform owner restructuring the rules of the market to favor its own product.
Stage one: The fee penalty. On March 27, 2025, Stripe announced that starting June 17, 2025, merchants who manually contest chargebacks would pay a new $15 counter fee per dispute. [31] This fee was waived for merchants using Stripe's native Smart Disputes tool. The practical effect: any merchant using DisputeNinja (or any third-party chargeback tool) now faced an additional $15 cost per dispute on top of whatever they were paying DisputeNinja. For a software startup with, say, 50 chargebacks per month, that is $750/month in additional fees — a direct tax on using a competitor's product. DisputeNinja's target customers (software and AI startups) are precisely the merchants most concentrated on Stripe, maximizing the impact of this fee change on DisputeNinja's addressable market. [32]
There is no public record of DisputeNinja responding to this fee change — no blog post, no pricing adjustment, no announcement of support for alternative processors. Whether the team attempted to pivot toward non-Stripe processors (Adyen, Braintree, Checkout.com) to escape the fee structure is unknown.
Stage two: The native product. In November 2025, Stripe launched Smart Disputes — an AI-powered chargeback automation tool that does what DisputeNinja does, natively, within the Stripe dashboard, trained on $1.4 trillion in annual payments volume. [33] [34] The data advantage is not marginal — it is categorical. DisputeNinja's AI agents were trained on whatever dispute evidence its small customer base generated. Stripe's system was trained on the dispute history of millions of merchants across every industry, geography, and card network. A merchant choosing between the two products faces: a free native tool with superior training data and no additional fees, versus a third-party subscription with inferior data and a $15/dispute penalty for using it.
This is a textbook platform squeeze: the platform owner first makes third-party tools more expensive, then launches a competing native product that is exempt from those costs. The strategy is not unique to Stripe — it mirrors moves made by Apple (App Store), Amazon (private label), and Google (vertical search) — but it is particularly effective in payments because the platform controls the dispute data that makes the product work.
Secondary Cause: Dual-Company Structure and Focus Dilution
The simultaneous operation of DisputeNinja and Trench by the same two-person founding team raises a structural question about focus. Both companies were accepted into YC S23, both addressed fraud-related problems, and both shared the same founders. [35]
Running two companies with two people is not impossible, but it creates a resource allocation problem that compounds over time. Every hour spent on Trench's open-source marketplace fraud platform was an hour not spent on DisputeNinja's chargeback automation product — and vice versa. YC's public communications in August 2023 focused on Trench, not DisputeNinja, suggesting that at least externally, Trench was the more prominent vehicle. [36]
Whether this dual structure reflected a rational hedge (two bets on the same fraud-tooling thesis) or a failure to commit to either product is impossible to determine from available data. What is observable is that neither company achieved the scale or visibility that would attract follow-on capital or press coverage — no HN launch threads, no TechCrunch coverage, no Reddit discussions were found for either product.
Tertiary Cause: Structural Disadvantage in a Data-Intensive Category
Even absent Stripe's competitive moves, chargeback automation is a category where data network effects strongly favor incumbents. The quality of an AI chargeback response depends on the breadth of training data: what evidence patterns correlate with winning disputes across different card networks, dispute reason codes, and merchant categories. A startup with a small customer base generates a small evidence dataset. A payment processor with millions of merchants generates a dataset that is orders of magnitude larger and more diverse.
This structural disadvantage was present from day one. DisputeNinja's founding team had domain expertise and a real problem to solve, but they were building a data-intensive AI product in a category where the data was owned by the infrastructure they depended on. The only viable path to escaping this disadvantage would have been to accumulate a large enough customer base — and the proprietary dispute outcome data that comes with it — before the incumbent decided to compete. DisputeNinja did not achieve that scale.
Absence of Public Traction Signals
No customer count, MRR, ARR, win rate, or revenue data of any kind was found for DisputeNinja. The company published no case studies, no testimonials with specific metrics, and no press releases announcing customer milestones. This absence is not proof of failure — early-stage B2B companies often operate quietly — but it is consistent with a company that did not achieve the traction that would have made it defensible against Stripe's moves or attractive to acquirers.
Key Lessons
-
Building a point solution on a payment processor's infrastructure concentrates platform risk to a fatal degree. DisputeNinja's entire product depended on Stripe's dispute data, Stripe's merchant relationships, and Stripe's fee structure — all of which Stripe restructured in 2025 to favor its own native product. A startup that cannot operate independently of its primary infrastructure provider is not building a business; it is building a feature that the infrastructure owner can absorb or penalize at will. The lesson is not "avoid platforms" but rather: if your product depends on a platform's data to function, your defensibility horizon is bounded by the platform's decision to compete.
-
Targeting a customer segment that is maximally concentrated on a single platform amplifies, rather than reduces, platform risk. DisputeNinja explicitly targeted software and AI startups — the cohort most likely to use Stripe as their payment processor. This was a rational ICP choice for early customer acquisition (shared language, shared tools, shared distribution channels), but it meant that Stripe's fee restructuring hit nearly every potential DisputeNinja customer simultaneously. A more diversified processor mix in the customer base would not have eliminated the risk, but it would have preserved a portion of the addressable market after Stripe's moves.
-
In data-intensive AI categories, the entity that owns the training data owns the category — and payment processors own the dispute data. Stripe's Smart Disputes was trained on $1.4 trillion in annual payments volume. DisputeNinja's AI agents were trained on whatever its small customer base generated. This gap was not closeable through better engineering or faster iteration; it was structural from the moment the company was founded. DisputeNinja's experience mirrors what happened to early social graph analytics startups when Facebook restricted API access in 2018 — the platform that owns the underlying data can always build a better product than a third party working with a subset of that data.
-
Operating two companies simultaneously with a two-person team creates a focus deficit that compounds at exactly the moment when focus matters most. DisputeNinja and Trench shared the same two founders, the same YC batch, and the same fraud-domain thesis. Neither company generated the press coverage, community engagement, or follow-on funding that would indicate breakout traction. The dual structure may have been a rational hedge, but it also meant that neither product received the undivided attention that early-stage product-market fit discovery requires. The founders' fraud expertise was real; the question is whether splitting it across two products diluted the signal they needed to find.
Sources
- YC Company Directory — DisputeNinja
- DisputeNinja Website
- DisputeNinja Demo Page
- DisputeNinja AI Response Generator
- Max Wu — LinkedIn
- Bowen Xue — LinkedIn
- Bowen Xue — Crunchbase
- YC LinkedIn Post — Trench Launch Announcement, August 3, 2023
- YC Company Directory — Trench
- Tracxn — Trench Funding Data
- Stripe Smart Disputes Documentation
- Stripe Radar Dispute Management
- Chargeflow — Stripe Dispute Fees 2025 Analysis
- Hiretop — Trench Fraud Prevention Profile
- Extruct — YC S23 Batch Analysis