A hardware-plus-subscription model that works precisely where a service-only model doesn't — and the two structural attacks now aimed at it, four months after a confidential IPO filing. Public sources only. August 2026.
Twelve months in which the company quadrupled revenue, won a patent case, lost the exclusivity it bought with it, and filed to go public.
Not found: audited financials of any kind, gross margin, EBITDA, attach rate, CAC, marketing spend, multi-year retention, current contract manufacturers, enterprise revenue share, the RingConn royalty rate, and the outcome of Ultrahuman's Federal Circuit appeal or the USPTO challenge to the underlying patent.
The category reorganised itself around a single competitive claim: no subscription.
The demand signal is real but the category is small, and the two research houses that track it disagree about whether it is still accelerating.
Ōura's measurement accuracy holds up better than sceptics assume. The harder problems are what people do with the data, whether anyone's health actually improves, and who else can see it.
A 2025 systematic review and meta-analysis in OTO Open pooled six validation studies against polysomnography, n=388, and found no statistically significant difference on any of seven sleep parameters — total sleep time, efficiency, wake after onset, onset latency, light, deep and REM. That is a stronger result than the usual dismissal of consumer sleep tracking allows for, and it should be conceded plainly.
The caveats matter, though. The sample is small, it pooled earlier ring generations rather than Ring 4 or 5, and null results with wide confidence intervals are not demonstrated equivalence — the deep-sleep interval spans −10 to +13 minutes, a large fraction of a typical night. Agreement on nightly totals is also a much weaker test than epoch-by-epoch staging agreement, which the meta-analysis does not report.
A 2024 JMIR meta-analysis of 23 pulse-oximetry studies (n=59,684) found SpO2 accuracy breached the FDA's 3% threshold across every skin-tone group — light 3.96%, medium 4.71%, dark 4.15% — and wearable pulse-rate error ranging from ±13.5 to ±33.7 bpm against a ±5 bpm industry standard. The finding is broader and worse than the familiar "PPG fails on dark skin" framing. Ring 5's marketing claim of "more powerful LEDs to improve accuracy across varied finger sizes and skin tones" reads as an implicit acknowledgement.
A December 2025 systematic review in Biomimetics covering 107 studies and roughly 100,000 participants found adherence falling from 80% at three months to 43% at twelve months. Ōura reports above 80% membership renewal after year one. Both can be true at once only if paying consumers behave very differently from study participants — or if renewal is measuring billing continuation rather than active use. A subscription that keeps billing while the ring sits in a drawer is churn that has not yet expressed itself, and it is the single largest risk to the lifetime value in the model below.
The same review found 65% of studies carried moderate-to-high risk of bias, 89% relied on undisclosed proprietary algorithms, only 35% documented participant race or ethnicity, and Ōura alone accounted for 72% of all studies — a degree of vendor concentration that makes the evidence base hard to read as independent.
A 2026 national survey published in SLEEP (n=1,280) found 32.4% of US adults use digital sleep tracking, and 30.9% of those screened positive for orthosomnia risk — anxiety about their own tracker data — roughly twice prior estimates. A validated instrument for it now exists. For a company whose core product is a nightly score, one in three users showing anxiety about that score cuts both ways: it drives engagement in the short run and abandonment in the long run.
The best available evidence — a Lancet Digital Health review of 39 systematic reviews covering 163,992 participants — found activity trackers add roughly 1,800 steps and 40 minutes of walking a day, and about a kilogram of weight loss. Effects on blood pressure, cholesterol and HbA1c were "typically small and often non-significant." And 34 of the 39 reviews were rated critically low confidence. That literature is about wrist-worn step trackers; there is essentially no equivalent outcomes base for sleep-and-recovery-score rings, whose mechanism is quite different.
The Department of Defense has been Ōura's largest enterprise customer since 2019, with tens of thousands of rings deployed, a Fort Worth facility built to serve it, and an enterprise platform hosted on Palantir's federal compliance infrastructure. Ōura describes the Palantir arrangement as a small inherited security layer and says it does not share or sell user data. The backlash in late 2025 was nonetheless real and named — cancellations covered in national press, a petition, and academic write-ups.
The acute version is reproductive data. Ōura markets aggressively to women — cycle, fertility, menopause and pregnancy tracking, with 26M cycles and 350,000 pregnancies logged — and women in their early twenties are its fastest-growing segment. Wearables are not covered by HIPAA. Building the most trusted women's-health data product and being the Pentagon's largest wearable vendor are strategies in tension with the same customer, and 2025 showed the tension is not theoretical. It belongs in an S-1 risk-factors section, which means it will shortly be public.
Eleven priced competitors, and Ōura and Whoop are the only two that require a subscription for core functionality. That is a structural alignment, not a coincidence.
| Product | Hardware | Subscription | 3-year cost | Scale | Position |
|---|---|---|---|---|---|
| Ōura Ring 5 | $399–499 | $5.99/mo required | $615–715 | 5M+ members ~8M rings cumulative | Highest hardware price in the category and one of two products gating features behind a fee |
| Whoop One / Peak / Life the model comparison | included | $199 / $239 / $359 per yr | $597–1,077 | ~2.5M members ~$1.1B run-rate | Hardware bundled into subscription. Converges on the same three-year value as Ōura from the opposite direction |
| Ultrahuman Ring Pro | $349 | none (optional add-ons) | $349 | ~$150M run-rate #2 by share | Designed around Ōura's patent and back in the US since June. A 43% discount on three-year cost of ownership |
| RingConn Gen 2 / Air | $314 / $199 | none | $199–314 | not disclosed | Licensed by Ōura and paying royalties — a permanently sanctioned price anchor at half the cost |
| Samsung Galaxy Ring | $399 list | none | $399 | not disclosed | Persistent deep discounting suggests weak sell-through; successor not expected before 2027 |
| Circular Ring 2 | $304 | freemium | $304 | not found | ECG and AFib, HSA/FSA eligible, "140+ biometrics free forever" |
| Garmin CIRQA | $199.99 | none | $200 | sold out at launch | Incumbent distribution, explicitly subscription-free, July 2026 |
| Amazfit Helio | $149.99 | none | $150 | not disclosed | Helio Ring 2 confirmed for H2 2026 with ECG and body composition in patent filings |
| Fitbit Air (Google) | $99 | $10/mo optional | $99–459 | not disclosed | Same architecture as Ōura at a quarter of the hardware price, with the AI layer paywalled |
| Movano EvieMED | not found | none | — | company in distress | The only FDA-cleared smart-ring pulse oximeter, held by a company facing Nasdaq deficiency notices |
Three-year cost assumes one device and continuous membership at list, before HSA/FSA treatment or discounting. Ōura's realised hardware price is materially below list — see the assumption register.
Ōura won at the ITC in August 2025 and the orders became enforceable in October. Eight months later Ultrahuman was back, selling a designed-around product at $349 with no subscription, and RingConn had converted from defendant to licensee — legitimised, paying royalties, and selling at $199. Ōura now collects an undisclosed royalty from a sanctioned competitor that undercuts it by half. The patent bought time and a revenue line; it did not buy the market. And the patent itself is under USPTO challenge.
Ultrahuman, RingConn, Circular, Amazfit, Renpho, Leep, Garmin and Samsung all market subscription-free as their headline differentiator, and they name Ōura when they do it. Survey data says 92% of prospective buyers would pay extra for health features — but that is stated preference, and the revealed preference visible across 2026 is that the fastest-moving products in the category are the ones that removed the fee.
These are the same business with the revenue recognised in opposite orders. Ōura charges for hardware and rents the software; Whoop gives away hardware and charges more for the software. Holding the customer's holding period constant shows how close the two land — and how differently they get there. These update live with the model in Section 8.
Whoop hardware cost is a modelled assumption — the company does not disclose it, and the device is given away, so it sits in cost of revenue rather than as a price. Ōura figures use the derived realised selling price, not list.
This comes before the model deliberately. Ōura has published a revenue trajectory, a member count and one retention figure. Everything else below is derived or mine.
| Input | Basis | Where it comes from |
|---|---|---|
| Ring price $399–499 (Ring 5), $349 (Ring 4) | Reported | Company store, verified August 2026. Ring 5 raised base finishes $50 and several premium finishes $100 |
| Membership $5.99/mo or $69.99/yr | Reported | Company membership page. Unchanged through 2026. Required on Gen3 and newer for anything beyond three daily scores |
| Realised hardware ASP $273 | Estimated | Derived by an analyst from Finnish statutory filings. Against a $349–499 list this implies 22–45% of realised give-up to discounting and retail channel margin. The pivotal and least-corroborated number in the model |
| Hardware cost $131 per ring | Estimated | Two unrelated methods converge: a TechInsights teardown put Ring 4 bill of materials below 39% of the $349 price (≈$136) at 50,000 units, and the Finnish filings imply $130.55. At ~3M units a year realised cost should be lower |
| Hardware gross margin ~52% | Estimated | Follows from the two lines above. No company-disclosed gross margin of any kind exists |
| Subscription gross margin 88% | Assumption | Not disclosed. Described only as "high" by third parties. Covers app infrastructure, clinical content and support |
| First-year renewal 80% | Reported | The only retention figure Ōura has ever published, stated twice in May 2026. Note it measures renewal, not use |
| Steady-state renewal 65% | Assumption | Mine. First-year renewal typically overstates steady state for hardware-tethered subscriptions because device attrition compounds. Set it to 80% to model the company's own figure persisting |
| Subscription attach 60% | Assumption | Never disclosed; the CEO declined to give it in February 2026. Cumulative rings sold is not an installed base, so the 5M members against ~8M rings is not an attach rate. Calibrate using the implied-members tile in the model |
| CAC $90 | Assumption | Never disclosed. No marketing spend figure exists either. The only adjacent datapoint is that roughly 22% of employees sit in sales and marketing |
| Units 3M per year | Reported | Company cites ~3M rings in 2025; an independent read of Finnish filings gives 2.5M. Treat 2.5–3.0M as the range |
| Revenue trajectory | Reported | ~$250M (2023, implied), >$500M (2024), ~$1B (2025), "close to $2B" 2026 guidance. All CEO statements; none audited |
| Hardware / subscription split 80 / 20 | Reported | Attributed to the CEO; implies roughly $400M of subscription revenue on ~$2B in 2026 |
| Fixed operating cost $700M/yr | Assumption | Mine, scaled to 1,200+ employees and a company guiding to ~$2B. Used only for the company-level bridge |
| Competitor prices | Reported | Each vendor's own store or press release, verified August 2026 |
Not found, and therefore not estimated: audited financials, gross margin, EBITDA, net income, attach rate, CAC, marketing spend, multi-year retention or cohort curves, enterprise and defence revenue, the RingConn royalty rate, Ring 5 bill of materials, current contract manufacturers, and tariff exposure. Note also that the $96M Department of Defense contract widely cited in 2024 was cancelled in March 2025 and should not be modelled as revenue.
Two revenue lines with very different margins, and a customer who pays the acquisition cost on the way in. Toggle any line off to remove it, or drag to change it. The Whoop comparison in Section 6 updates with every change.
Ōura has never published one, and the CEO declined to give it in February 2026. Dividing 5M members by ~8M cumulative rings gives about 63%, but cumulative shipments are not an installed base — that figure ignores returns, dead and retired devices, replacements, gifts and enterprise units. Use the implied paid members tile instead: adjust attach and renewal until it lands near the 5M the company reported for Q2 2026, and you have a pair of assumptions that is at least internally consistent with the one number they disclosed.
Read down the ladder. Each line is live.
The line that makes this business different. Look at contribution at the point of sale. In a service-only subscription business that line is deeply negative — you pay to acquire a customer and recover it over years. Here the hardware gross profit lands the moment the ring ships, so the customer effectively funds their own acquisition and everything the subscription earns afterwards is incremental. That is the whole structural argument for selling hardware at a margin rather than giving it away.
What it costs. The same structure makes revenue dependent on selling new units every year rather than on a base that compounds. Subscription revenue is roughly 20% of the total, so growth still has to come from moving boxes — into a category where every competitor is cheaper and most charge nothing recurring. Drag the realised price down toward $200 and watch how quickly the advantage disappears.
Two modelling caveats. CAC is charged entirely at the point of sale, which is the conservative treatment. And the model assumes one ring per customer per lifetime — no repurchase on the roughly 18-month hardware cadence — so lifetime value is understated for loyal upgraders and overstated for anyone whose ring stops fitting.
Judge each against the structural fact established above: hardware margin funds acquisition, and subscription is only a fifth of revenue. Click "Model this" to load any case.
The current strategy, and it has been vindicated: Ring 5 raised prices into a market full of cheaper rings and revenue is still guided to roughly double. The defence is feature cadence — 14 new features in a year, Health Radar, physician messaging, lab uploads, an in-house clinical team of 40+. What makes it fragile is that the premium is being defended against products that are 50–75% cheaper and free of a recurring fee, and the realised selling price is already 22–45% below list. Drag the realised price toward $220 and the hardware margin that funds acquisition compresses fast.
Every competitor markets subscription-free against Ōura by name, and the CEO has publicly refused twice, arguing the fee funds ongoing development. The model shows why he is right on arithmetic: the subscription is high-margin and removing it costs roughly a fifth of revenue at close to 90% margin, which is most of the profit. It also shows the trap — the fee is the single most attacked feature of the product, and defending it forces the hardware cadence that raises R&D and inventory intensity. The preset models a partial capitulation: attach falls and realised price drops to meet Ultrahuman.
This is the part a $199 Shenzhen ring genuinely cannot copy. Ōura already has 1,200+ organisational partnerships, the Department of Defense as its largest enterprise customer, and an Essence Healthcare Medicare Advantage deal that as of June 2026 feeds nighttime breathing data into physician workflow for sleep apnoea screening. Procurement relationships, clinical integration, SOC 2 and HITRUST certification and a 40-person clinical team are the assets that compound. It is not a preset because the unit is a negotiated contract rather than a retail sale — and note that enterprise revenue share has never been disclosed, so its current contribution is unknown.
Health Panels at $99 for 50 biomarkers, the Dexcom Stelo integration at $99, GLP-1 journey tools, lab uploads. The strategic problem is that Whoop launched Advanced Labs through the same lab partner — both companies resell Quest — so this adjacency is convergent rather than differentiating. It is also membership-gated and US-only, which makes it an ARPU lever on the existing base rather than an acquisition channel. Model it by raising the membership price rather than the attach rate.
The confidential S-1 went in on 21 May 2026, and Ōura's last private mark of $11B on roughly $2B of guided revenue is about 5.5× forward revenue — cheaper than Whoop at roughly 9× on $1.1B, despite Ōura being the larger business on both revenue and members. The case for going now is that revenue has quadrupled in two years, the company says it is profitable, Apple has retreated and the FDA has just widened what can ship without clearance. The case against is that the S-1 will have to disclose the attach rate, the real retention curve, the gross margin and the enterprise concentration — every number in the register above marked "not found" — into a market where the competitive set reorganised around undercutting the subscription eight weeks before filing.
The patent behind the exclusion order is under post-grant challenge at the USPTO by Samsung and others. If it survives, Ōura keeps a royalty from RingConn and a credible deterrent against anyone else. If it falls, the royalty stops, the deterrent goes, and the eight months of exclusivity it bought against Ultrahuman turn out to be the entire return on the litigation. The outcome is not disclosed as of this writing and is not something Ōura controls — which is exactly why it belongs in a risk section rather than a model.