iRobot

Phase A — Understand the business

Company Overview

iRobot Corporation (Bedford, MA; incorporated Delaware 1990) is the company that created the consumer robot-vacuum category with the Roomba (launched 2002). By 2024 it was a single-category consumer-hardware company: robotic vacuum cleaners (RVC) and robotic mops sold under the Roomba and Braava brands, plus a nascent robotic lawn mower, sold through retailers and distributors across the US, EMEA, Japan and other markets. It exited its once-diversifying bets years ago (Mirra pool cleaner, Ava telepresence, and it spun out its defense/PackBot business as Endeavor Robotics in 2016) — leaving it wholly exposed to one fiercely competitive consumer niche.

How it made money. Hardware sales, one-time, at retail. In fiscal 2024, total revenue was $681.8M, down 23.4% from $890.6M in 2023. It is fundamentally a product company with weak recurring revenue — an app and connectivity layer, but no meaningful subscription annuity to cushion hardware cyclicality. Contract terms are ordinary purchase orders, not take-or-pay; retailers and distributors can and did cut orders hard when consumer sentiment weakened.

Customer concentration. Dangerously high. One retailer accounted for 22.2% of total revenue in FY2024 (24.0% in 2023, 22.6% in 2022). That retailer is Amazon.com — which is also the counterparty to the failed 2022–2024 acquisition, meaning iRobot's largest customer and its would-be acquirer were the same entity. Selling channel: a network of retailers in the US, Canada, Japan, France and Spain (direct sales/marketing teams), with Germany, Italy, Poland, Latin America and parts of Asia run through distributors.

The "iRobot Elevate" pivot (2024). Management moved to an asset-light model — offloading non-core engineering to joint-design/contract-manufacturing partners, centralizing marketing, cutting R&D via offshoring, and concentrating IP in a new "iRobot Labs". The strategy's goal was survival through margin recovery, not growth. It failed: the asset-light dependence on partners (chiefly Picea, its primary contract manufacturer) is precisely how the company ended up owned by that partner.

Supply Chain

Map: contract manufacturers (China · Vietnam · Malaysia) → iRobot (design/brand/IP + distribution) → retailers/distributors (Amazon, big-box, specialty) → consumer.

Named / identifiable stakeholders along the chain:

• Upstream — primary contract manufacturer: Picea (Shenzhen PICEA Robotics Co. + Santrum Hong Kong Co.). iRobot's manufacturing agreement with Picea was scheduled to expire Aug 17, 2025. This is the single most important node in the entire dossier: the contract manufacturer that built the robots also held secured/manufacturing debt against iRobot — and, in Dec 2025, took 100% of the equity out of Chapter 11 by cancelling that debt. The supply chain literally swallowed the company.
• Upstream — capital / secured lender: TCG Senior Funding L.L.C. (an affiliate of The Carlyle Group), administrative + collateral agent on the ~$200–224M senior secured Term Loan. Carlyle's ~$190M claim was cancelled in the restructuring.
• Component inputs: vacuum motors, brushless motors, LiDAR/vSLAM sensors, batteries, semiconductors, plastics — sourced through the CMs; iRobot took non-cash reserves for obsolete/excess component inventory and losses on non-cancelable purchase commitments in FY2024, a tell that its supply commitments outran collapsing demand.
• Downstream buyers: Amazon (22.2% of revenue), other big-box/specialty retail, and regional distributors.

Chokepoints & single-source dependencies: (1) Manufacturing concentrated in Picea — a single-supplier dependency that became an ownership takeover. (2) Amazon as ~22% customer — demand concentration. (3) Vietnam tariff exposure — US import duties of 46% on Vietnam-made goods added ~$23M of cost in 2025, directly undercutting the point of diversifying out of China. This lens is not generic: the named nodes (Picea, Carlyle/TCG, Amazon) are the whole story.

Competitive Advantages (moats)

The moat evaporated — that is the thesis of the whole collapse. iRobot's historical advantages were (a) brand ("Roomba" as the genericized category name), (b) first-mover installed base and app data, and (c) a patent estate on navigation/mapping. By 2024 none was durable enough to defend pricing or share:

• Brand: still #1 in North America but eroding. Genericized brand recognition did not translate into pricing power once Chinese rivals offered equal-or-better features cheaper.
• IP: iRobot litigated aggressively — a patent-infringement suit against SharkNinja (five patents, later three, at the ITC + Massachusetts federal court, opened Oct 2019). But litigation is a weak, expensive, slow moat against a field of a dozen well-capitalized entrants; it did not stop share loss.
• Bargaining power: essentially gone. Against Amazon (22% of revenue), iRobot was the price-taker. Against its contract manufacturer Picea, iRobot was the debtor. Against Chinese competitors (Roborock, Ecovacs, Dreame, Xiaomi, Narwal), it was the slower innovator. Who needed whom more? iRobot needed all of them; none needed iRobot.

Named rivals (from the 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes.): Cecotec, Dreame, Ecovacs, Electrolux, LG, Midea, Roborock, Samsung, SharkNinja (Shark), Xiaomi. The competitive verdict is stark: the world's top five robot-vacuum makers are now all Chinese — Roborock, Ecovacs, Dreame, Xiaomi, Narwal — together ~70% of Q2-2025 global shipments. iRobot's moat was a decade's head start that its rivals closed while it was distracted by the Amazon deal.

Segments

our figures on disk is header-only (empty), so segment detail is drawn directly from the filings and labeled accordingly. iRobot reports one operating segment (consumer robots); the meaningful cuts are geography and product/ASP:

By geography (FY2024):

• Domestic (US) revenue −$95.8M (−22.4%)
• International revenue −$112.9M (−24.4%), driven by Japan −27.7% and EMEA −24.5%
• US sales specifically fell ~47% cumulatively into 2024 vs peak

By product / operating drivers (FY2024):

• Total robots shipped −18.8%
• Gross average selling price −5.6% (mix shift toward entry-level as new launches skewed cheaper — the opposite of premiumization; in FY2023 ASP had risen 6.8% on premium mix, so 2024 marks a deliberate down-market move to chase volume)
• Product line: Roomba vacuums (core), Roomba Combo (vacuum+mop), Braava mops, new robotic lawn mower — but the multi-function "wash/dry/heat the mop pad" dock category where EMEA was growing was exactly where iRobot was late.

Trend and cause: every geography decelerating simultaneously, volume and price both down — a textbook signal of share loss into a growing category, not a shrinking market. The RVC market grew in EMEA in 2024 (multi-function docks); iRobot shrank into it. iRobot's global unit share fell to 13.7% (−2.6pp) in 2024 as Roborock overtook it (16% units, 22.3% of revenue, 3.3M units shipped, +20.7% YoY).

---

Phase B — Measure performance

Earnings Result

The descent, quantified. Three consecutive years of large losses, accelerating cash burn, and a balance sheet that went negative.

Full-year P&L:

Cumulative net loss FY2022–FY2024 ≈ ($736.5M). FY2024's smaller loss was flattered by a one-off: G&A came in at negative ($17.1M) because the $94.0M Amazon "Parent Termination Fee" (received Jan 29, 2024) was booked against operating expense. Strip that out and 2024 operations were far worse than the headline; the fee, net of $18.8M professional costs, was the cash that bought 2024 another year of life.

Latest print — Q3 FY2025 (quarter ended Sept 27, 2025):

• Revenue $145.8M, −24.6% YoY (from $193.4M); 9-month revenue $375.0M, −26.4% (from $509.8M)
• Gross margin 31.0% (up from 32.2% YoY) — the one genuinely improving line; 9-mo GM 27.7% vs 24.7% (asset-light + cost cuts working on the margin, too late to matter)
• Total operating expenses $62.9M (up from $55.1M — note the prior year benefited from the Amazon fee offset, so opex "rose" optically)
• 9-mo operating loss ($100.1M) vs ($41.9M); operating margin (26.7)%
• Q3 net loss ($21.5M); 9-mo net loss ($131.6M) vs ($68.4M); 9-mo basic/diluted EPS ($3.97) on ~33.1M weighted shares
• Balance-sheet flags — this is the crux:
• Cash and equivalents $24.8M as of Sept 27, 2025, down from $134.3M at Dec 28, 2024 — and down from $40.6M just one quarter earlier (June 28, 2025), despite drawing $31.0M from a restricted-cash account. The final $5.0M of restricted cash was fully drawn Sept 30, 2025 — "the Company has no sources upon which it can draw for additional capital."
• Term Loan fair value $205.3M (up from $200.6M — PIK interest compounding the principal) vastly exceeds cash
• Total current liabilities $480.9M (up from $218.2M) — the jump is the Term Loan reclassified to current on default risk
• Total liabilities $508.5M > total assets $481.6M → stockholders' DEFICIT ($26.9M) (from +$61.2M at Dec 28, 2024) — balance-sheet insolvent
• 9-mo cash used in operating activities ($104.2M) vs ($30.5M) — burn 3.4× worse YoY

Market reaction. On Mar 12, 2025 (going-concern disclosure), the stock fell ~30%. On Oct 27, 2025, when the sale counterparty withdrew after exclusive negotiations, the stock fell ~30% again. On the Dec 14–15, 2025 Chapter 11 filing, IRBT plunged ~73%. Each leg priced in a step toward zero. Anything unusual vs history? Everything: negative G&A, negative equity, restricted-cash raids, a fair-value debt mark disclosed because the company was effectively pricing its own insolvency.

Earnings Calls (sentiment trend)

transcripts/ on disk is empty (0 transcripts ingested), so this lens is filing- and web-derived and labeled accordingly. The arc of management messaging across 2024–2025 is a case study in narrowing hope:

• Early 2024 (post-Amazon-collapse): new CEO Gary Cohen frames a turnaround — "iRobot Elevate," asset-light model, cost discipline, "rebuild the Company both financially and strategically".
• Q1 2025: Cohen launches "the largest new product launch in iRobot's history" (five+ new Roombas from Mar 18, 2025), the bet-the-company move.
• Mar 12, 2025: tone breaks — the board announces a review of strategic alternatives (sale/refinancing) simultaneously with the going-concern disclosure.
• Through 2025: language shifts from "turnaround" to "substantial doubt," "dependent on continued waivers... in the Lenders' sole discretion," and finally, in the Q3 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited., the explicit admission that absent new capital the company "would likely seek bankruptcy protection" and equity holders would "unlikely" receive proceeds.

Recurring phrases that intensified: "going concernGoing concernAn auditor’s formal doubt that the company can fund itself for the next year. It is a warning about survival, not about performance.," "waiver," "strategic alternatives," "cost optimization." What they stopped saying: growth, market-share gains, premiumization, addressable-market expansion. The founder's epitaph captures the sentiment better than any call: ex-CEO Colin Angle called the bankruptcy "a tragedy for consumers", and co-founders blamed "wrong-minded" FTC/EU opposition to the Amazon deal.

Comps

Read: the comp set tells the whole story without a single multiple. Five of the top-five global players are Chinese and ~70% of shipments (Q2-2025); the one American incumbent went bankrupt and was bought by a Chinese supplier. The relevant "comp" for iRobot is not a valuation multiple — it is relative competitive position, and iRobot's was terminal.

Stock-Price Catalysts

The >5% moves over the relevant window were almost entirely deal- and solvency-driven, not fundamentals-driven — which itself reveals what the market keyed on: survival, not earnings.

• Aug 2022: Amazon announces $1.7B acquisition (~$61/share) → stock spikes toward deal price.
• Jul 2023: Amazon cuts the deal price (to ~$51.75/share after a new debt raise) → stock falls.
• Jan 18–29, 2024: EU Commission signals it will block; deal terminated Jan 29, 2024 → stock collapses; the merger premium fully unwinds.
• Mar 12, 2025: going-concern + strategic-review disclosure → −30%.
• Oct 27, 2025: sale counterparty withdraws after exclusivity; disclosed that the offer was below recent trading price → −30%.
• Dec 14–15, 2025: prepackaged Chapter 11 filing + Picea RSA → −73%; delisted from Nasdaq, trades as IRBTQ from Dec 22, 2025.
• Terminal: common cancelled in the plan; residual quote collapses to ~$0.05 and effectively zero.

Pattern: for the last three years, IRBT was a binary event stock — every material move was a deal headline or a covenant/solvency headline. Fundamentals (revenue, margins) mattered only as inputs to the survival question. That is the signature of a company the market had already decided was a going-concern coin-flip.

---

Phase C — Judge people & books

Management

CEO: Gary Cohen (appointed May 2024). Prior: CEO of Timex and Qualitor Automotive; senior roles at Energizer, Playtex, Gillette — a consumer-products turnaround specialist, explicitly hired to restructure, not to grow. His archetype is the professional restructuring manager, the right profile for triage but not a robotics visionary. What he actually delivered: executed the ~51% headcount cut (1,113→541), the asset-light pivot, inventory reduction ($152.5M→$76.0M), gross-margin recovery (to 31% in Q3-2025), and the largest-ever product launch — and it still wasn't enough. The turnaround improved unit economics at the margin while the top line fell faster than costs could be cut. Verdict: competent execution of an un-winnable hand; the damage was done before he arrived.

Founder: Colin Angle — co-founded iRobot in 1990, CEO for ~33 years, architected both the Roomba era and the fateful Amazon deal. He departed after the deal collapsed (2024). His capital-allocation and strategic legacy is mixed-to-negative in hindsight: he bet the company's independence on an acquisition that regulators killed, leaving iRobot exposed and cash-poor when the window to self-fund a competitive response had closed. Post-bankruptcy he publicly blamed the FTC/EU — a founder externalizing a strategic miss.

Skin in the game / insider ownership: our figures not present on disk; not independently sourced here → n/a. In any case, insider equity was cancelled in the plan alongside all common — management's stock was wiped with everyone else's.

Capital-allocation history — the core indictment: Over FY2022–2024 the company burned ~$736M of net losses while share and margin eroded. The one large discretionary inflow — the $94M Amazon breakup fee — was consumed by operating losses within roughly a year rather than funding a decisive strategic reset. R&D was cut (from $165M in 2022 to $93M in 2024) at the exact moment Chinese rivals were out-innovating on multi-function docks and navigation — arguably the most damaging call: starving the product engine to preserve cash accelerated the competitive loss. This is value-destroying allocation forced by a value-destroyed position.

Red flags: the deep reliance on a single contract manufacturer that was also a creditor (Picea) is a governance/dependency red flag that became the endgame — the counterparty on the other side of your manufacturing and your debt is the counterparty that can take your company. Not fraud, but a structural conflict that materialized against shareholders.

Forensic Red Flags

Not an accounting-fraud story — a solvency story. The books were clean enough that the auditor's own going-concern paragraph is the headline forensic finding. Where earnings/cash-flow diverged, it diverged in the direction the disclosures already flagged:

• Going-concern (the defining item): the FY2024 audit opinion carries an explicit explanatory paragraph expressing substantial doubt about the ability to continue as a going concern. This is the most severe non-fraud red flag an auditor can raise, and it was reaffirmed every quarter through Q3-2025.
• Covenant breach / debt reclassification: the ~$205M Term Loan was moved to current liabilities, and the company survived only on a chain of waivers — Amendments No. 1 through 5+ to the Credit Agreement (Mar–Sept 2025), each extending the waiver period by weeks, each carrying fees: a $3.6M PIK fee (2.0% of principal), warrants for 1,840,503 shares (6% of outstanding, $0.01 strike), a deferred $4.0M "use fee," and required more-frequent reporting. Dilutive, coercive, lender-in-control terms — the fingerprints of a borrower with no leverage.
• Cash vs. earnings divergence: operating cash burnCash burnHow much more cash goes out than comes in, per period. The clock on a company with no profits. ($104.2M 9-mo) exceeded the 9-mo net loss framing once working-capital and the restricted-cash draws are accounted for; the company drained restricted cash to fund operations — a liquidity-of-last-resort tell.
• Inventory / purchase commitments: non-cash reserves for obsolete/excess component inventory and losses on non-cancelable purchase commitments in FY2024 — demand fell below what the supply chain was contractually committed to build.
• Goodwill / negative carrying value: as of Sept 27, 2025, the single reporting unit had a negative carrying value; the Oct-27 counterparty withdrawal was flagged as a triggering event for impairment testing.
• Stock-based comp / dilution: antidilutive shares excluded from diluted EPS (options underwater), and lender warrants layered on — SBC was not flattering non-GAAP so much as the equity was being handed to lenders.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. regulatory/regulatory-findings.md (fetched 2026-07-06 via SEC EDGAR EFTS, LR + AAER, 2021–2026) returned 0 SEC findings for iRobot.
• 10-K Item 3 (Legal Proceedings): the material litigation is offensive IP — iRobot's own patent-infringement suits against SharkNinja (five→three patents, at the ITC and Massachusetts federal district court, filed Oct 15, 2019); SharkNinja counter-sought declarations of non-infringement. This is iRobot as plaintiff defending its moat, not a defendant in an enforcement action.
• Non-SEC enforcement (FTC/DOJ/EU) web search: the material government action was antitrust review of the Amazon acquisition — the EU Commission's signaled block (competition concerns: marketplace foreclosure of rival vacuum sellers + Alexa access) and US FTC opposition, which together killed the deal on Jan 29, 2024. Note the irony: the regulatory action that "protected competition" removed iRobot's lifeline and the company then failed anyway, ceding the market to Chinese incumbents. No fines/penalties/consent decrees against iRobot itself were found.
• Summary: No material regulatory or accounting-fraud findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-07-06. The forensic story is going-concern and covenant distress, fully and repeatedly disclosed, not concealment.

---

Phase D — Project & stress-test

Forward Projection

No EPS projection is meaningful — the public equity is extinguished. A forward EPS model for IRBT is undefined: the common stock was cancelled in the confirmed Chapter 11 plan (Feb 2026), holders received no recovery, and the entity is now private, wholly owned by Picea. There is no share count, no public consensus, no tradeable claim. Per skill discipline, no our model forecast is logged (there is nothing scoreable — the outcome, zero, is already realized).

For completeness, the three scenarios that were on the table pre-resolution, and how they settled:

• Base (played out): strategic review fails to find a solvent buyer at a viable price → Chapter 11 → equity zero. Realized.
• Bull (did not happen): a strategic/financial buyer takes out the whole company at a premium (the 2022 Amazon path) → this was the only route to shareholder value, and it died when the counterparty withdrew Oct 2025 offering below market.
• Bear (also realized): disorderly liquidation with even worse recoveries — narrowly avoided; the prepackaged deal with Picea preserved the operating business (a going-concern sale) but still zeroed the equity.

The one number that mattered was runway, not EPS: $24.8M cash vs a $205M+ current Term Loan vs $100M+/yr operating burn ⇒ months, not years. The company said so itself: expected default Dec 1, 2025, "no sources upon which it can draw".

Post-resolution capital structure (private): Picea cancelled ~$264M of debt ($74M manufacturing debt + $190M Carlyle/TCG term loan) in exchange for 100% of the reorganized equity. The Roomba business continues under Picea, delevered, private, and outside public markets.

Bull vs Bear

Bull case (as it stood, now moot): iRobot owned the #1 brand in North American robot vacuums, a genericized name, a real installed base, and a patent estate; a turnaround CEO was cutting costs, margins were recovering (GM 24.7%→27.7% 9-mo), the largest product launch in company history was shipping, and a strategic/financial buyer could take the whole thing out — as Amazon nearly did at $1.7B. If any buyer had closed, equity had upside from distressed levels.

Bear case (realized in full): three structural risks, all permanently impairing:

1. Category commoditized by faster, cheaper Chinese rivals — Roborock, Ecovacs, Dreame, Xiaomi, Narwal now ~70% of global shipments; iRobot's share fell to 13.7% and Roborock passed it. A single-category consumer-hardware company that loses its category has no second act.
2. Balance-sheet insolvency + lender control — negative equity, a Term Loan exceeding cash by ~8×, survival on discretionary weekly waivers. Once the strategic buyer withdrew, default and bankruptcy were arithmetic.
3. Structural dependence on a creditor-supplier (Picea) — the manufacturer/lender that could, and did, take the company.

Pre-mortem (it broke — what happened?): exactly as written above. The thesis that "a turnaround + a buyer saves the equity" broke because (a) the turnaround improved margins but couldn't stop a 25%+ annual revenue decline, (b) cash ran out faster than the strategic review could close, and (c) the only rational buyer (Picea/the lenders) had no reason to pay for equity when they could take the assets through Chapter 11 by forgiving debt.

Were the multiples too high? Any positive equity value was too high — the market repeatedly overestimated recovery odds, which is why each disclosure re-rated the stock down 30–73%.

Contrarian view (what the market refused to see, until it did): the "Roomba brand is worth something / someone will buy it" hope ignored that the brand's value accrues to whoever controls manufacturing and cost, and that party was already a creditor. The market kept pricing a shareholder rescue; the actual value transfer went to the supply chain, over the shareholders' heads. The residual contrarian point today: the Roomba business is not dead — it survived, delevered, under Picea — but that is entirely orthogonal to IRBT equity, which is zero.

Devil's Advocate (short-seller)

The short thesis needs no construction — it already paid out to zero. Reconstructed as it would have read at the going-concern disclosure (Mar 2025), every plank proved correct:

• What structurally breaks how it makes money? It already broke: a single-category hardware company lost its category to lower-cost, faster-innovating Chinese OEMs. There is no recurring revenue to fall back on.
• Where is revenue concentrated, and what if it shifts? ~22% in one customer (Amazon) and heavy EMEA/Japan exposure — both fell 24–28% in 2024. Concentration amplified the decline.
• Why is the moat weaker than bulls think? Brand recognition ≠ pricing power; patents ≠ share protection against a dozen entrants; iRobot had to litigate (SharkNinja) precisely because the product moat was gone.
• Most dangerous competitor bulls underestimated? Roborock — passed iRobot in global units in 2024 and commands premium revenue share, proving the Chinese entrants win at both ends (price and premium features).
• Worst capital-allocation / incentive problems? Consuming the $94M Amazon fee in a year of operating losses; cutting R&D from $165M→$93M into a technology race; and, structurally, financing survival with coercive lender warrants and PIK fees that handed equity to creditors.
• What must hold for the price? A solvent buyer at a premium — which required more time and cash than existed. The moment the buyer withdrew (Oct 2025), the equity was arithmetically zero.
• Growth disappoints by 20–30%? It disappointed by ~25% per year, and that is what killed it.
• The single scenario that permanently impairs — and its plausibility? Chapter 11 with equity cancelled. Plausibility in Mar 2025: high. Realized: Dec 2025. The devil's advocate was simply right.

Management Questions

Ordered by information value. Given the entity is now private under Picea, these are addressed to Picea/reorganized iRobot management — the questions whose answers most determine whether the Roomba brand has any future (and whether a future public re-listing could ever be investable):

1. Cost structure post-deal: now that Picea owns both manufacturing and the brand, what is the true delivered cost of a Roomba versus a Roborock, and does vertical integration finally give the brand price-competitive economics?
2. Product roadmap & cadence: can the reorganized company match the ~6-month innovation cycle of Chinese rivals, or does it remain a fast-follower — and what is the specific plan for multi-function docks and next-gen navigation?
3. R&D rebuild: R&D was cut to $93M; what is the go-forward R&D budget under Picea, and where does IP now legally reside (iRobot Labs vs. Picea)?
4. Channel & Amazon: does the ~22% Amazon concentration persist, and how does Chinese ownership affect the Amazon relationship (which was itself the failed acquirer)?
5. Tariffs: with US duties of 46% on Vietnam-made goods and Chinese ownership, what is the tariff-optimized manufacturing footprint, and what is the landed-cost exposure to further US–China trade action?
6. Brand equity durability: how much of the Roomba brand premium survived the bankruptcy headlines, and what is the retention/repurchase data on the existing installed base?
7. Recurring revenue: is there any credible path to a services/consumables/subscription annuity, or does this remain a one-time-hardware business forever?
8. Geographic strategy: after Japan (−27.7%) and EMEA (−24.5%) collapses, which markets does Picea defend vs. abandon?
9. Warranty & support liabilities: what are the post-emergence obligations to existing customers (the "tragedy for consumers" concern), and how are they funded?
10. Capital plan: how much fresh capital is Picea injecting beyond debt forgiveness, and toward what?
11. Management continuity: does Gary Cohen (or a Picea-appointed team) run it, and what are the incentives now that there is no public equity?
12. Patent portfolio: what happens to the SharkNinja litigation and the broader patent estate — asset to monetize, or lapsed?
13. Governance / conflicts: with the supplier now the owner, how are transfer-pricing and related-party terms governed?
14. Re-listing intent: is a future IPO/relisting contemplated, and on what milestones — or is Roomba now permanently a private Chinese-owned brand?
15. The counterfactual: what, specifically, would have had to be different (the Amazon deal, the R&D cuts, the timing) for iRobot to have survived independent — the lesson Picea must not repeat?

---