Yardstick Research tear-sheet / AI sales cohort

Methodology · how we score · rubric weights in plain sight · vendors received this sheet seven days before publication and could flag factual errors, never rankings

Puzzle

Identity

• Founded: 2019; product publicly launched July 2023. [VENDOR-CLAIMED + THIRD-PARTY corroboration https://puzzle.io/about, https://pitchbook.com/profiles/company/438696-55]
• HQ: San Francisco, California. [THIRD-PARTY https://pitchbook.com/profiles/company/438696-55]
• Founders: Sasha Orloff (Cofounder and CEO, formerly LendUp/Mission Lane, Grameen Foundation, CGAP, Citigroup) and John Cwikla (Cofounder and CTO). [VENDOR-CLAIMED https://puzzle.io/about, https://www.linkedin.com/in/sashaorloff]
• Funding: ~$66.5M total raised across Seed to Series B per Tracxn. Last publicly-confirmed round was a $30M Series B announced November 14, 2023, led by S32 and XYZ Capital with General Catalyst, Felicis, Kapor Capital, FOG Ventures, Sterling Road, Born Ventures, Soma Capital, Alumni Ventures, Gaingels, and 50+ CFO/operator angels from Stripe, Gusto, Plaid, Robinhood, and Y Combinator-backed companies. Valuation not publicly disclosed. [VENDOR-CLAIMED + THIRD-PARTY corroboration https://puzzle.io/blog/puzzle-raises-an-additional-30m-to-fuel-a-new-era-of-ai-powered-accounting, https://tracxn.com/d/companies/puzzle/__S4yCZBnPr-mpGqg7hbrU72aQdTnrSDtmsEqXuDcw_v8, https://www.thesaasnews.com/news/puzzle-raises-30-million-in-funding]
• Headcount: 86 to 91 (PitchBook 86, LeadIQ ~86 across 5 continents Dec 2025, Tracxn 91 as of Jan 31, 2026). [THIRD-PARTY https://pitchbook.com/profiles/company/438696-55, https://leadiq.com/c/puzzle/5e1f68ef62be9f43cade582b, https://tracxn.com/d/companies/puzzle/__S4yCZBnPr-mpGqg7hbrU72aQdTnrSDtmsEqXuDcw_v8]
• Customer count (claimed): 7,000+ startups and accounting firms. No third-party verification. [VENDOR-CLAIMED https://puzzle.io/]
• Recent news (last 12 months):
• 2026 AI Close Agents for Accounting Firms surfaced as a flagged-new product with a book-a-demo CTA on puzzle.io. [VENDOR-CLAIMED https://puzzle.io/]
• November 14, 2023 $30M Series B led by S32 and XYZ Capital; customer-growth claim 15% MoM new-customer growth at announcement. [VENDOR-CLAIMED https://puzzle.io/blog/puzzle-raises-an-additional-30m-to-fuel-a-new-era-of-ai-powered-accounting]
• 2025 Puzzle Bookkeeping partner program launched, positioning the platform as a workbench for outsourced bookkeeping firms. [VENDOR-CLAIMED https://puzzle.io/partners/puzzle-bookkeeping]
• Archetype: AI-native general-ledger and bookkeeping platform built for venture-backed startups and the outsourced bookkeeping firms that serve them. Positioned as a QuickBooks replacement, not a tax engine. 1-click Tax Package is an export-to-CPA workflow rather than an in-product filing capability.

Total score: 66.25 / 100

• Stage fit:
• Foundation (<40 readiness): yes Free tier covers the first $20K of transactions and Starter at $25/user/month annual makes Puzzle a low-friction first-AI deploy for a seed-stage startup currently on QuickBooks. The AI categorization + real-time burn/runway dashboard maps directly to weekly founder work.
• Pilot (40-59): yes Core ($60/user/month annual) and Complete ($100/user/month annual) cover the typical Series A controller pilot scope: insights, multi-entity via integration, revenue recognition, AI reconciliations, white-glove QB migration.
• Scale (60-79): conditional Scale tier delivers subledgers, priority agent support, and removes transaction and integration caps. Hard constraint is the cohort-fit ceiling: Puzzle is a GL platform, not a tax engine, and is bounded to startup-scale customers. Buyers needing sales-tax automation must add Anrok / Stripe Tax / Numeral on top.
• Optimization (80+): no Compliance gap (no publicly-disclosed SOC attestation, no ISO 27001, no public LLM disclosure, thin documented audit trail) and the absence of any enterprise-ERP bidirectional sync make Puzzle a poor fit for an Optimization-stage tax-cohort buyer in 2026.
• One-line verdict: The cohort's strongest AI-native accounting/GL platform for seed-to-Series-B startups already on Stripe + Brex + Mercury + Ramp + Gusto, and a tax-adjacent vendor by design (Puzzle owns the books that feed the tax engine; it is not itself a tax engine).

Cohort-fit caveat

Puzzle is an AI accounting/GL platform with a 1-click Tax Package export, not a tax automation product. The closest cohort precedent is Trullion (35/100), included with the platform-note "Tax-adjacent. Included in the AI tax cohort because the research context flags it as audit/revrec adjacent. The seven-dimension tax rubric penalizes it on jurisdiction breadth and filing because those aren't its product." The same logic applies to Puzzle: the rubric correctly penalizes Tax-rule accuracy (2/4) and Filing automation (2/4) because those are not the product. The 66.25 / 100 reflects credit for the genuine tax-adjacent surface (Tax Package + R&D-credit founder guidance + CPA partner network) on top of strong integration, cost, and time-to-value scores. A buyer evaluating an AI tax engine will not find one in Puzzle; a buyer evaluating "the books we send to our CPA" will find a strong candidate.

Headline numbers

Dimension scores

Pricing detail

Source: https://puzzle.io/pricing. All [VENDOR-CLAIMED].

• Starter $0/month (first $20K of transactions), then $25/user/month annual or $30/month monthly. 1 user. Up to 98% auto-categorization, auto-drafted financials, cash and accrual books, unlimited connections, 1-click tax report, CPA-approved chart-of-accounts templates, AI support and AI setup review, 25 AI credits. No multi-entity, no advanced reporting, no custom chart of accounts.
• Core $60/user/month annual or $72/month monthly. 5 users. Described as "Your QuickBooks replacement." Adds insights (cash, burn, runway, margin), spend and revenue tracking, variance analysis, multi-entity via integration, advanced reporting, custom chart of accounts, chat support, self-guided QB migration. 25 AI credits.
• Complete $100/user/month annual or $120/month monthly. Unlimited users. Adds 50% faster month-end close guarantee, AI-powered accuracy review, AI-powered reconciliations, AI insights and analysis, AI-enhanced categorization, categorize-from-your-inbox, classes/departments/projects, revenue recognition, priority chat support, white-glove QB migration. 100 AI credits/month.
• Scale $300/user/month annual or $360/month monthly (custom-pricing path available). Unlimited users. Adds subledgers, priority agent support, priority feature requests, no transaction limits, no integration limits, backup and restore, dedicated support and onboarding. 300 AI credits/month.
• Add-ons: AI Close add-on across all tiers (pricing not published).
• Fair-use: Starter capped at $20K of transactions. Core and Complete have implicit transaction limits not disclosed. Scale has no transaction or integration limits.
• Overages: No explicit overage fees on the public pricing page.

Integrations

Source: https://puzzle.io/integrations, https://puzzle.io/, https://puzzle.io/security.

• Banking and corporate cards: Brex, Mercury, Ramp, Rippling. [VENDOR-CLAIMED] Bank data ingestion via Plaid and Finicity. [VENDOR-CLAIMED https://puzzle.io/security]
• Payment processors: Stripe. [VENDOR-CLAIMED]
• Accounts payable: Bill.com, Ramp. [VENDOR-CLAIMED]
• Payroll: Gusto, Deel, Rippling. [VENDOR-CLAIMED]
• Cap table: Carta (coming soon). [VENDOR-CLAIMED]
• Financial planning and forecasting: Causal, Runway. [VENDOR-CLAIMED]
• Other: Central, Joiin, Meow. [VENDOR-CLAIMED]
• Sync directionality: [UNKNOWN] Integrations broadly described as read-only per the security page; per-integration bidirectional-sync detail not published.
• Enterprise ERPs NOT integrated: NetSuite, SAP S/4HANA, Oracle Fusion, Microsoft Dynamics 365, Workday, Sage Intacct, Xero. None of the cohort's enterprise tax-cohort buyers can use Puzzle as a GL overlay on top of these systems. [VENDOR-CLAIMED, by absence]
• POS NOT integrated: Shopify, Toast, Square, Lightspeed. [VENDOR-CLAIMED, by absence]
• CRM NOT integrated: Salesforce, HubSpot, Pipedrive. [VENDOR-CLAIMED, by absence]
• Tax engines NOT integrated natively: Anrok, Stripe Tax, Numeral, Avalara, TaxJar, Vertex. Tax-cohort buyers pairing Puzzle with a tax engine must wire the two up manually. [VENDOR-CLAIMED, by absence]
• Developer surface: [UNKNOWN] No public REST/SDK/webhook portal discoverable.

Editorial assessment

Puzzle is the strongest AI-native accounting/GL platform in the market today for the seed-to-Series-B venture-backed software, fintech, or DTC startup whose finance stack is already Stripe + Brex + Mercury + Ramp + Gusto and whose tax filing is being done by an external CPA. The vendor-claimed 98% auto-categorization is partially corroborated by the one independent hands-on test we found (Fahim AI's 90-day measured 85 to 95% range), the integration list is exactly the modern-startup finance stack, the pricing is fully published with a real free tier, and CEO Sasha Orloff's prior-founder credibility (LendUp lessons, Mission Lane spinout, Stanford/Princeton/Harvard lecturing on tech finance) is genuine. The Series B in late 2023 came with credible operator angels (50+ CFOs from Stripe, Gusto, Plaid, Robinhood, and Y Combinator companies) and the customer roster includes the outsourced bookkeeping firms (Burkland, Accountalent, Attivo Partners, Decimal) that move the most startup books in the market.

The hard question for the AI tax cohort is whether Puzzle belongs in it at all. The cohort's leaders (Anrok, Avalara, Intuit ProConnect, Thomson Reuters ONESOURCE, Vertex, Stripe Tax, Numeral) are tax engines or tax-preparation suites: they compute rates, file returns, manage exemptions, handle audit-trail evidence for the IRS or for state DOR audits. Puzzle does none of those things. Puzzle produces the books that those tax engines and tax-preparation suites consume, plus a 1-click Tax Package export to an external CPA at year-end. The closest cohort precedent is Trullion (35/100), included with the "Tax-adjacent" platform note. Puzzle's 66.25 / 100 reflects credit for the integration depth, the cost transparency, the fast time-to-value, and the genuine tax-adjacent surface (Tax Package + R&D-credit founder guidance + CPA partner network) on top of a structurally low score on Tax-rule accuracy and Filing automation that no GL-only platform could earn.

The strategic ceiling on Puzzle inside the tax cohort is twofold. First, any buyer running NetSuite, Sage Intacct, SAP, Oracle Fusion, Microsoft Dynamics, Workday, or QuickBooks Online as their committed GL cannot replace that GL with Puzzle without a migration, and Puzzle does not bidirectionally sync with any of them. Second, any buyer whose tax need is sales-tax automation (the canonical tax-cohort need) or enterprise-tier compliance (Sovos, ONESOURCE territory) will not find what they need in Puzzle. The buyer who actually wins with Puzzle is the seed-to-Series-B startup CEO who is currently running QuickBooks Online with a bookkeeper and wants to swap both for an AI-native GL plus light tax-package handoff, and the outsourced bookkeeping firm running 20 to 200 startup clients that wants AI categorization and exception-based review across the whole book.

The compliance posture is the other ceiling. No publicly-disclosed SOC attestation, no ISO 27001, no documented audit-trail surface, no disclosed LLM. For Puzzle's current ICP this is tolerable. For any procurement-team-led buyer in 2026 it is not. The third-party review base is also thin (G2 5/5 across only 4 reviews, no TrustRadius depth comparable to Numeric or FloQast), which means the independent signal available is mostly review-blog teardowns and a handful of Reddit posts rather than aggregated peer-review depth.

Revisit triggers: (a) Puzzle ships a SOC 2 Type II attestation (or Type I as a first step) and adds it to the public security page; (b) underlying LLM provider is disclosed; (c) documented public REST/SDK developer portal lands beyond the read-only security framing; (d) native tax-engine integrations to Anrok / Stripe Tax / Numeral / Avalara ship, which would let Puzzle function as books-source-of-record for a tax-cohort buyer; (e) Series C or strategic transaction is announced (Puzzle is ~30 months past its last disclosed raise); (f) independent third-party benchmark on AI categorization correction-friction (not just headline accuracy) is published; (g) native bidirectional sync to NetSuite, Sage Intacct, or QuickBooks Online ships, which would materially change the cohort-fit conversation.

Best for

• Stage: Foundation and Pilot. Conditional fit for Scale (cohort-fit and compliance ceilings). Not a fit for Optimization.
• Company profile: Seed-to-Series-B venture-backed software, fintech, or DTC startup; 1 to ~100 employees; finance stack already Stripe + Brex + Mercury + Ramp + Gusto + Rippling; tax filing already handled by an external CPA; founder-CEO wants real-time burn and runway dashboard without retraining the team.
• Secondary fit: Outsourced bookkeeping firms (Burkland, Decimal, Attivo Partners, Accountalent) running 20 to 200 startup clients who want AI categorization + exception-based review across all books on one workbench.
• Industry fit: Strong for B2B SaaS and FinTech. Moderate for DTC, marketplaces, digital services. Not a fit for Manufacturing, Healthcare-provider, Public Sector, Construction, Retail (POS-heavy), Energy, Real Estate.
• Annual tooling budget: $0 (free tier) to ~$20K at Scale-tier 5-user shape; AI Close add-on extra.
• Skip if: (a) you run NetSuite, Sage Intacct, SAP, Oracle Fusion, Microsoft Dynamics, Workday, Xero, or QuickBooks Online as your committed GL (no bidirectional sync); (b) you need a sales-tax engine (Puzzle does not have one, this is what Anrok, Stripe Tax, Numeral, Avalara, TaxJar do); (c) you need ONESOURCE/CCH-class enterprise tax-preparation; (d) you need SOC 2 Type II or ISO 27001 for procurement; (e) you need a vendor that can name its LLM provider; (f) you need POS integration (Shopify, Toast, Square); (g) your scale is mid-market or enterprise (10,000+ employees, multi-entity global); (h) you need a published REST/SDK developer portal today.

Right-of-reply gaps

Puzzle should be invited to fill these in before publication:

• Underlying LLM provider for AI categorization, AI reconciliations, AI accuracy review, AI Close. None disclosed publicly.
• SOC 1, SOC 2 Type I, or SOC 2 Type II attestation status. Neither claimed on the public security page; many of the cohort's mid-tier vendors carry at least SOC 2.
• ISO 27001 status. Not claimed publicly.
• HIPAA, GDPR posture, and EU data residency options. US-server hosting stated; EU residency not stated.
• AI training-data policy. Is customer transaction data used to train Puzzle's models? Opt-out terms?
• Audit-trail surface. Customer-facing documentation of how audit-trail evidence is captured and exported for IRS or state DOR audit response.
• Public developer documentation portal. REST endpoints, SDK languages, webhook event types. Currently [UNKNOWN].
• Sync directionality per integration. The "read-only" security framing is in tension with any workflow that pushes journal entries back into a connected payroll or AP system; per-integration sync detail not published.
• 1099 / W-2 / K-1 / federal-form generation scope. Not documented on the public tax-documents help-center article; division of labor with Gusto/Rippling/Deel unclear.
• AI Close pricing and exact launch date. Surfaced as a "NEW!" flagged product on the homepage; no formal launch-date post found.
• Customer count verification. Vendor claims 7,000+; no third-party verification of the figure.
• Funding update. Series B was November 2023. Any subsequent raise or strategic transaction in the 30 months since?
• Independent benchmark data on AI correction-friction. Headline categorization accuracy is partially measured; the practical "AI got it wrong and I cannot fix it" friction is well-documented by independent reviewers but has no benchmark number.
• Native tax-engine integrations on the roadmap. Anrok, Stripe Tax, Numeral, Avalara, TaxJar — any of these on the public roadmap?

Sources

Puzzle first-party: - https://puzzle.io/ - https://puzzle.io/about - https://puzzle.io/pricing - https://puzzle.io/security - https://puzzle.io/integrations - https://puzzle.io/testimonials - https://puzzle.io/blog/puzzle-raises-an-additional-30m-to-fuel-a-new-era-of-ai-powered-accounting - https://puzzle.io/blog/a-guide-to-tax-season-for-founders - https://puzzle.io/partners/puzzle-bookkeeping - https://puzzle.io/101-for-founders - https://help.puzzle.io/en/articles/8594277-puzzle-tax-documents

Press / investor / company databases: - https://www.thesaasnews.com/news/puzzle-raises-30-million-in-funding - https://www.crunchbase.com/organization/valencia-data - https://pitchbook.com/profiles/company/438696-55 - https://tracxn.com/d/companies/puzzle/__S4yCZBnPr-mpGqg7hbrU72aQdTnrSDtmsEqXuDcw_v8 - https://leadiq.com/c/puzzle/5e1f68ef62be9f43cade582b - https://www.linkedin.com/company/puzzlefin - https://www.linkedin.com/in/sashaorloff

Third-party reviews / comparisons: - https://www.g2.com/products/puzzle-puzzle/reviews - https://www.capterra.com/p/10028909/Puzzle/ - https://www.gotofu.com/blog/puzzle-io-review - https://www.fahimai.com/puzzle-io - https://beancount.io/blog/2025/07/27/puzzle-io-user-feedback-analysis - https://www.producthunt.com/products/puzzle/reviews - https://pilot.com/blog/puzzle-alternatives

Podcast / founder interviews: - https://lex.substack.com/p/podcast-from-mobile-credit-to-autonomous - https://podcasts.apple.com/us/podcast/crafting-puzzle-challenging-quickbooks-and-helping/id1649355311?i=1000678268021 - https://www.sliceoffinance.com/p/ep-4-sasha-orloff-co-founder-ceo-puzzle - https://www.rippling.com/resources/how-i-screwed-this-up-sasha-orloff - https://ltse.com/insights/the-puzzle-handbook-for-founders-top-accounting-and-finance-mistakes-to-avoid