Innodata Inc. [INOD]

Innodata's growth is organic, cash-collected and clean — zero acquisitions, negative accruals, and receivables that fell in the quarter revenue rose 24.5% sequentially. It is also 56% dependent on one unnamed customer under master service agreements that, in the 10-K's own words, “generally do not obligate customers to purchase services in future periods” and are “generally terminable by our customer upon 30 to 90 days’ notice.” There is no backlog, no RPO and no committed contract value disclosed anywhere; deferred revenue is 2.5% of TTM revenue and falling. The implied-path test passes at every exit multiple tested — but the screen's +51.8pp margin is overstated by roughly 32pp once the exit multiple is anchored on services comparables rather than a universe-wide set, and the share count is put on a diluted basis.

Key findings

• CUSTOMER CONCENTRATION, PRECISELY: one customer was 56% of Q1 2026 revenue (58% FY2025, 48% FY2024, 61% Q1 2025) and 65% of receivables at 2026-03-31. A second customer was 17% of Q1 2026 revenue. Top two = 73% of the quarter.
• CONTRACT STRUCTURE, VERBATIM FROM THE 10-K: master service agreements with project-based statements of work that 'generally do not obligate customers to purchase services in future periods', 'generally terminable by our customer upon 30 to 90 days' notice'. There is no backlog, no RPO and no committed contract value disclosed anywhere. Deferred revenue is $7.2m = 2.5% of TTM revenue, and falling. 97.5% of the revenue base is uncommitted.
• BUT THE GROWTH IS ORGANIC AND CASH-BACKED. FY2024, FY2025 and Q1 2026 investing cash flow is capex ONLY; goodwill flat at $2.0m; zero acquisitions. FY2025 CFO $46.8m on net income $32.2m (1.45x). Sloan accruals -10.1% of average assets. DSO went 57.8 -> 45.9 days in the quarter revenue rose 24.5% sequentially: receivables FELL while revenue surged. This is not the AAOI or 10x Genomics pattern.
• THE BRIEF'S 'SERVICES vs SOFTWARE TERMINAL MARGIN' TRAP LANDS ON THE WRONG PARAMETER. The screen's 15.8% terminal margin is already a services margin (peer median 14.9%) and INOD already earns 17.1% TTM / 18.8% in Q1. The software-like assumption is in the EXIT MULTIPLE: 27.2x EBIT from a universe-wide 72-name set, against a services comp set (EPAM/Genpact/Cognizant) trading at 6.8-9.0x.
• THE IMPLIED PATH SURVIVES THE CORRECTION ANYWAY. At the services-anchored 9.0x exit on diluted shares and verified net cash, today's price requires a 50.5% five-year revenue CAGR vs 70.3% demonstrated — margin +19.8pp. It passes at every multiple tested down to 6.8x (+11.1pp). But the screen's headline +51.8pp is overstated by ~32pp.
• SCREEN VALIDATION: TTM revenue $283,415,000 ties EXACTLY to four filed quarters. Share count 32,655,358 verified against the 10-Q cover and reconciled (35,839 issued - 3,184 treasury). Scale cross-check passes: 14,898 / 32,625 = $0.457 vs filed basic $0.46; 14,898 / 35,572 = $0.419 vs filed diluted $0.42.
• SCREEN DEFECT 1 — STALE NET CASH. The screen's $112,287k implies debt of exactly $5,079k, which matches no 2026 line item and is precisely the 2022-12-31 LongTermDebtNoncurrent XBRL tag. Actual: cash $117,366k, revolver undrawn, only obligations pension $9,314k + Microsoft licence financing $2,536k. Debt-only net cash $114,830k. Impact 0.4% of EV — immaterial, but exactly the stale-tag pattern the brief warned about.
• SCREEN DEFECT 2 — BASIC SHARES UNDERSTATE EV BY 9%. Diluted WASO is 35.572m vs 32.655m basic (SBC ran 6.6% of revenue in Q1). Market cap $2,047m not $1,879m; EV $1,933m not $1,767m; EV/Sales 6.82x not 6.23x.
• SCREEN DEFECT 3 — MIXED PERIODS. op_margin_pct 15.8 is FY2025's margin applied to TTM revenue. TTM operating income is actually $48.5m (17.1%), and Q1 2026 alone ran 18.8%.
• PRICE SERIES VERIFIED REAL. 894 SIP daily bars pulled independently. My realized vol: 252d 105.8%, 126d 125.8%, 63d 156.5%, 21d 57.6%. The screen's 121.5% is in family. The volatility is genuine and event-driven, not a data artifact: +86.0% in a single session on 2026-05-08 (Q1 earnings), peak close $121.50 on 2026-06-04, -52.6% since. Every large move maps to a dated filed event.
• MECHANISM, NAMED AND DISCLOSED: a new set of engagements with a Big Tech company expected to generate ~$51m of 2026 revenue, from zero twelve months ago. Verifiable and verified: Q1 2026 disclosed a second customer at 17% = $15.3m, where no second customer reached 10% a year earlier. ~$36m remains to be recognised across Q2-Q4 2026 — roughly $12m/quarter of already-identified incremental revenue.
• THE GUIDE LOOKS CONSERVATIVE ON ITS OWN ARITHMETIC. FY2026 guidance of ~40%+ growth on $251.7m is $352.3m. Q1 2026 annualises to $360.4m — already above the full-year guide. The raised guide therefore implies ZERO sequential growth for three quarters while $36m of a named programme is still to be recognised. Nothing in the filings indicates a Q1 one-off.
• INSIDERS SOLD $158.3 MILLION IN FIVE WEEKS AT ~$97.5 AVERAGE; THE STOCK IS $57.56. Parsed from Form 4 XML: CEO Jack Abuhoff 1,233,651 shares / $121.9m at $98.83 avg; Ashok Mishra $26.3m; director Louise Forlenza $6.2m; Interim CFO Marissa Espineli $2.1m at $105.88; director Stewart Massey $1.8m. Total 1,622,885 shares = 5.0% of shares outstanding and 8.4% of market cap, between 2026-05-12 and 2026-06-16. Selling began five days after the +86% earnings gap and ended twelve days after the all-time high. NONE of the 20 Form 4s in the window carries the Rule 10b5-1(c) checkbox and none of the 21 Form 144s reports a plan adoption date.
• FY2024 EPS IS TAX-BENEFIT-INFLATED — DO NOT USE IT AS A BASE. FY2024 net income ($28.7m) EXCEEDED operating income ($24.3m) on a NEGATIVE 17.1% effective tax rate; the 10-K discloses a one-time $6.0m valuation-allowance release in Q3 2024. Reported EPS went $0.89 -> $0.92 (+3%) while pre-tax income went ~$24.5m -> ~$42m (+71%).
• Q1 2026 OPERATING CASH FLOW IS WORKING-CAPITAL-FLATTERED. Of the $37.3m CFO, ~$16.5m is an accrual build (payables/accrued +$11.6m, accrued salaries +$4.9m). Ex-working-capital cash generation is ~$23m, reconciling to Adjusted EBITDA of $25.0m less cash taxes. Do not annualise $37.3m.
• UNRESOLVED: 'Accrued expenses and other liabilities' went $9.6m -> $27.2m (+183%) in one quarter against +24.5% revenue — 19% of quarterly revenue in one line, with no component breakdown filed. This is the single open accounting question and should be the first item on the Q2 call.
• SBC IS GROWING FASTER THAN REVENUE AND IS 24% OF HEADLINE ADJUSTED EBITDA. $4.0m FY2023 -> $11.1m FY2025 -> $5.9m in Q1 2026 alone (6.6% of revenue, ~$24m annualised). Adjusted EBITDA of $25.0m adds back $5.9m of it; GAAP operating income was $16.9m. Disclosed and reconciled (no Twist-style concealment), but the $25.0m is not a cash-economics number.
• DISCLOSURE DEGRADATION, FLAGGED: effective Q1 2026 Innodata retired three-segment reporting (DDS / Synodex / Agility) for a single segment, so the declining businesses (Synodex -8%, Agility +9% in FY2025) can no longer be separated from the growing one. Offsetting: the Q1 2026 concentration disclosure IMPROVED, voluntarily breaking out a second 10%+ customer.
• MENTION-FREQUENCY (14 quarterly earnings releases, normalised per 1,000 words; transcripts were unreachable so the substrate substitution is declared): 'observability' appears for the FIRST TIME in 14 quarters in 2026Q1, naming the new platform; 'agentic' 0 -> 1.9; 'diversif-' 0 -> 1.9; 'hyperscaler' re-enters after ten silent quarters. Decaying: 'Generative AI' 10.4 -> 0.9, 'LLM' 4.7 -> 0, 'annotation' 1.0 -> 0 (gone entirely since 2024Q2). The one GENERATIVE finding is negative: 'sovereign' and 'federal' spiked together in 2025Q3 and vanished, and the 10-K's Federal Practice discloses no contract, award or dollar of revenue.
• IN-HOUSE PRECEDENT FOR THE BEAR CASE: Synodex FY2025 revenue fell 8%, which the 10-K attributes 'primarily to termination of a customer contract'. The bear mechanism is not hypothetical.
• THE H1 2025 STALL IS THE BEST EXISTING EVIDENCE OF WHAT PROJECT REVENUE DOES WHEN A PROGRAMME PAUSES: revenue went $58,344k -> $58,393k (+0.1% sequential) across two quarters inside the '70.3% demonstrated CAGR', then re-accelerated to +24.5% sequential in Q1 2026.
• GOVERNANCE: the company ran with an Interim CFO through its highest-growth period; permanent CFO Jayant Chauhan (ex-SVP of M&A at Mphasis) became effective 2026-07-06, at a company with $117m of undeployed cash that lists 'possible business acquisitions' as a use. Auditor is BDO India Services Private Limited, an uncommon choice for a US registrant of this size.

Sections

Disclosed limitations

• THE TOP CUSTOMER IS NOT NAMED IN ANY FILING and is not named here. Nor is the second (17%, ~$51m) customer, described only as 'one of the world's leading Big Tech companies'. Any attribution would be fabrication.
• MANAGEMENT'S '453% growth in other Big Tech customers' IS NOT ADOPTED. It is arithmetic on a cut not disclosed in the financials. The verifiable analogue — the residual non-top-customer base — grew 74% ($22.8m -> $39.6m).
• THE $17.2m SINGLE-QUARTER JUMP IN ACCRUED EXPENSES AND OTHER LIABILITIES IS UNRESOLVED. No component breakdown is filed. The benign reading (offshore delivery accrual plus bonus accrual on a headcount ramp) is an inference, not evidence.
• NO STREET CONSENSUS, ESTIMATE REVISIONS, ANALYST COUNT OR EXTERNAL PRICE TARGET. Alpha Vantage returned the 25/day quota message. Consensus Criteria is INDETERMINATE and blocks nothing — but it also means valuation.md's 'sanity band' check of the 12-month target against an external professional target COULD NOT BE PERFORMED. Recorded as a gap, not silently skipped.
• NO FY2027 GUIDANCE EXISTS. The Q1 2027 and Q2 2027 quarters inside the NTM band are extrapolated from FY2026 guidance plus the disclosed ~$51m programme schedule and are labelled as such. They are not consensus.
• TRANSCRIPT MENTION-FREQUENCY USED EARNINGS PRESS RELEASES (8-K EX-99.1), NOT CALL TRANSCRIPTS. Transcripts are not on EDGAR and no transcript corpus was reachable. Consequence: prepared remarks only, no analyst Q&A. Release length collapsed from ~1,950 to ~900 words in mid-2024, so ALL counts are normalised per 1,000 words per the reference file's explicit ISRG length-artifact warning.
• WHETHER THE $158.3m OF INSIDER SALES WAS MADE UNDER PRE-EXISTING RULE 10b5-1 PLANS CANNOT BE ESTABLISHED. The Form 4s carry no such checkbox and the Form 144s report no plan adoption date. That is all the filings say; no inference is drawn.
• EXIT-YEAR GROWTH IS UNKNOWN, so whether the 9.0x services anchor is genuinely growth-matched is unknown. Above ~20% terminal growth the exit multiple is declared UNIDENTIFIED rather than defaulted.
• PEER SET IS FOUR NAMES REDUCED TO THREE. Accenture excluded (share-count XBRL tag stale to 2010-03-19), Globant excluded (revenue tag returned $64m, obviously broken), Endava and TaskUs excluded (no usable current share count). Excluded rather than estimated.
• NO Q2 2026 EARNINGS DATE HAS BEEN ANNOUNCED in any filing as of 2026-07-29. The catalyst calendar carries the historical analogues (2024-08-08, 2025-07-31) explicitly labelled INFERRED. No date is fabricated.
• FEDERAL / SOVEREIGN AI REVENUE IS ZERO IN THE MODEL. A Federal Practice is described in the 10-K; no contract, award or dollar of revenue is disclosed anywhere.
• THE EVALUATION AND OBSERVABILITY PLATFORM IS MODELLED AT ZERO. $1m closed to date, no GA date, no guidance.
• NO FINANCIAL MODEL .xlsx WAS BUILT. The model is documented as INOD_Financial_Model_Notes.md with every figure traced to a filed XBRL fact or statement, and every derived figure showing its derivation. Time-boxed run; the accounting-quality and valuation work was prioritised per the brief.
• NO CHART PACK. Same reason.
• NO POSITION IS RECOMMENDED and nothing is appended to trade_recommendations.jsonl — the memo issues no verdict, and the ledger records specific trade recommendations, not analyses.