Levels prescribes a continuous glucose sensor through an affiliated telehealth network, ships it to you, and pairs it with an app that logs meals, draws your curve, and scores each meal. The app is the product. The sensor underneath is commodity hardware from Dexcom or Abbott — the same hardware every consumer platform in this category ships, with the same accuracy.
That framing sets up the only decision that matters here. You are not choosing a measurement; you are choosing a software layer and a delivery mechanism, and paying roughly ten times what the sensors alone would cost. Whether that is a good trade depends on something about you rather than about the product: whether a well-designed interface is the difference between running a two-week experiment and leaving the sensor in a drawer. This review covers the cost math, the accuracy limits, what the output supports, and where a different purchase fits better.
The Verdict
Scorecard
| Category | Score | Why |
|---|---|---|
| Sensor hardware | Not scored | Dexcom or Abbott, depending on supply — the same hardware every platform in the category ships |
| App design and food logging | 9/10 | The most polished interface in the category, and the main thing the subscription buys |
| Metabolic scoring | 7/10 | Per-meal scores are legible and motivating; the scoring model itself is proprietary |
| Onboarding and prescription | 8/10 | Questionnaire through an affiliated telehealth network; no visit with your own physician |
| Human interpretation | 4/10 | Self-directed by design — no dietitian consult included in the standard membership |
| Data export and portability | 6/10 | You can review your own history in-app; check current export options before committing |
| Long-term value | 5/10 | Value is front-loaded, and the subscription outlives the insight for most users |
| Cost transparency | 7/10 | Plan structure is stated clearly; the annualized figure is the number to check |
What it costs, and what the same data costs without a platform
Plan structures change, so confirm current terms directly. The comparison below is the one worth running before you subscribe.
| Option | Typical price | Annualized | What you get |
|---|---|---|---|
| Sensors-included membership | Roughly $199 per month | Roughly $2,400 | Continuous sensor supply plus full app access |
| App-only membership with your own sensor | Lower monthly fee, plus sensor cost | Varies with how many sensors you buy | App and scoring layered onto an over-the-counter sensor you purchase yourself |
| Over-the-counter sensor bought directly, no platform | Roughly $50–$100 per 14–15 day sensor | Roughly $200 for two two-week blocks | Raw glucose data in the manufacturer app, no scoring or logging layer |
The gap is roughly a factor of ten, and it buys software rather than measurement. Two other cost notes matter. Insurance almost never covers a consumer CGM platform for someone without diabetes, so this is out-of-pocket spending. Health savings and flexible spending accounts sometimes apply, which is effectively a 20–35% discount depending on your tax situation and worth checking before you pay.
Onboarding, the prescription pathway, and setup
- Intake questionnaire. You complete a health questionnaire online. An affiliated telehealth network reviews it and issues the prescription. Your own physician is not involved, and the prescription is tied to your enrollment rather than to you.
- Sensor ships. Delivery typically takes a few days. Which manufacturer you receive depends on supply at the time, not on your plan choice.
- Application. The sensor goes on the back of the upper arm with an applicator. It takes a couple of minutes and stays on for 14 to 15 days, through showers and exercise.
- Warm-up. Readings begin after an initial warm-up period, and the first 12 to 24 hours are the least reliable. Treat that data as calibration rather than signal.
- Logging. The value comes from logging meals as you eat them. Retrospective logging is where most users quietly stop, and where the experiment fails.
Realistically you are five to seven days from signup to usable data. Plan the first sensor around an ordinary two weeks — not a holiday, a travel stretch, or an illness — because those produce data that describes the circumstances rather than you.
Accuracy limits worth knowing before you interpret anything
The sensor is good. It is not a laboratory instrument, and reading it as one produces bad conclusions.
- It measures interstitial fluid, not blood. Interstitial glucose lags blood glucose by roughly 5 to 15 minutes, which is why a sensor and a fingerstick disagree during a fast rise or fall with neither being wrong.
- Mean absolute relative difference runs roughly 8 to 10 percent against laboratory reference values. Two sensors worn at once on the same person can differ by 10 to 20 mg/dL.
- The first 12 to 24 hours are unreliable. Insertion trauma affects local readings.
- Compression lows are real. Sleeping on the sensor restricts local blood flow and produces a false plunge. If your curve shows a dramatic 3 a.m. drop only on nights you slept on that arm, the sensor was compressed, not your blood sugar.
- Exercise raises glucose, healthily. Hard training prompts the liver to release stored glucose. That spike is the system working, and avoiding it would be a mistake.
The practical rule that follows: treat the shape of the curve as the signal and any individual number as approximate. Judge a food by a pattern across two weeks, never by one reading.
What the output actually tells you
A CGM makes several things visible that nothing else does — and leaves the most important metabolic question unanswered.
| What you are looking at | Reasonable range | What it suggests | Common misreading |
|---|---|---|---|
| Fasting glucose on waking | 70–90 mg/dL | Repeatedly above 100 suggests early insulin resistance | A dawn rise of 10–20 mg/dL is normal cortisol physiology |
| Peak after a meal | Under 140 mg/dL | Repeated peaks above 160–180 in a non-diabetic are worth investigating | One high reading after a celebration means nothing |
| Return to baseline | Within 2–3 hours | Slow return matters more than peak height | Fiber, fat and protein legitimately extend the curve without harm |
| 14-day average glucose | 85–105 mg/dL | The most stable number a consumer CGM produces | Large disagreement with your HbA1c suggests sensor drift |
| Standard deviation | Under roughly 15 mg/dL | Wild swings track poorly with metabolic health | Flatness is not the goal, and exercise spikes are healthy |
Beyond the numbers, the genuinely useful findings tend to be behavioral rather than nutritional. A ten-minute walk after dinner flattening a peak by 20 to 30 mg/dL. A poor night's sleep raising the next morning's fasting glucose. Alcohol suppressing overnight glucose and then rebounding. Eating protein and vegetables before the carbohydrate portion of the same meal changing the curve. Those effects are large, personal, and immediately actionable, and they are the strongest argument for wearing a sensor at all.
Who it fits, and who should buy something else
| Your situation | Verdict | Reasoning |
|---|---|---|
| You learn well from a well-designed interface on your own | Good fit | The app is the product, and it is the strongest one in the category |
| You want to run structured self-experiments on meals | Good fit | Logging and per-meal scoring make comparisons easy to set up |
| You plan one to two months of wear, then act | Good fit | Matches where the insight actually is, without the multi-year subscription |
| You want a professional interpreting your data | Poor fit | No dietitian consult is included in the standard membership |
| You only want the raw glucose curve | Poor fit | An over-the-counter sensor delivers that for a fraction of the annual cost |
| You have a history of disordered eating | Caution | Real-time meal scoring is a strong reinforcement loop and can narrow a diet |
| You have diabetes or take insulin or a sulfonylurea | Poor fit | Glucose management belongs with your own clinician and a prescribed CGM |
| You have never run a blood panel | Poor fit | Fasting insulin and HOMA-IR detect insulin resistance a CGM cannot see |
Trade-offs to weigh
- Interface quality versus price. The app is the best in the category and costs roughly ten times what the sensors alone would. That trade is about you, not the product.
- Self-directed versus supported. No dietitian consult is included, which keeps the price lower than coaching-inclusive platforms and leaves interpretation to you.
- Front-loaded value versus recurring cost. Insight peaks in the first 30 to 60 days while billing continues monthly.
- Prescription tied to enrollment. Access to sensors depends on staying subscribed, unless you switch to the over-the-counter route.
- Proprietary scoring. Per-meal scores are motivating and their underlying model is not published, so an odd score cannot be traced to a reason.
How it compares to the direct alternatives
Listed alphabetically, with the direct-purchase route included because it is the relevant price anchor.
| Option | What it adds | Typical price | Main value | Fits someone who |
|---|---|---|---|---|
| Levels | Polished consumer app, food logging, per-meal metabolic scoring | Roughly $199/mo with sensors included | Self-directed learning | Wants a strong interface and will interpret their own data |
| Nutrisense | App plus registered dietitian consults | Roughly $250–$400/mo depending on tier and term | Human interpretation | Wants a professional reading the curve with them |
| Over-the-counter sensor bought directly | Manufacturer app only | Roughly $50–$100 per sensor | Raw data | Only wants the glucose curve, with no subscription |
| Signos | Weight-management framing with activity prompts | Roughly $140–$160/mo | Weight-focused coaching prompts | Frames the goal as weight management |
Against Nutrisense, the difference is human interpretation. Nutrisense includes registered dietitian consults, and Levels does not; Levels prices lower and invests in the interface instead. Pick the coaching model if you want someone reading the curve with you and would not otherwise act on it. Pick the self-directed model if you are comfortable interpreting your own data and would rather not pay for consults you will not book. Both statements carry the same weight — the choice is about how you work, not about which platform is stronger. Our Nutrisense review runs the same rubric.
Against buying an over-the-counter sensor directly, the question is narrower: would you interpret raw data on your own? The sensor is identical and the data is identical. What you lose is the food database, the logging structure, the scoring, and the convenience of not thinking about supply. For roughly $2,200 a year in difference, that is a real question rather than a rhetorical one.
Related
- CGM accuracy — MARD, interstitial lag, and compression lows
- Over-the-counter CGMs — Libre Rio, Lingo and Stelo
- Nutrisense review
- CGM platforms compared — including the over-the-counter route
- Biomarker guides — fasting insulin, HOMA-IR, and what glucose alone misses
- Reversal vs management — where continuous glucose data earns its cost
- Nutrition guides
Frequently Asked Questions
What is Levels and what do you get for the subscription?
Levels is a consumer glucose monitoring platform. It prescribes a continuous glucose sensor through an affiliated telehealth network, ships it to you, and pairs it with an app that logs meals, displays your curve, and assigns each meal a metabolic score. The sensor itself is commodity hardware manufactured by Dexcom or Abbott, and it is the same hardware other platforms in this category ship. What the subscription buys is the interface, the food database, the scoring layer, and the logistics — not a different or better measurement.
How much does Levels cost per year?
The sensors-included membership commonly runs around $199 per month, which annualizes to roughly $2,400. Some plans separate the membership from sensor supply, letting you attach the app to an over-the-counter sensor you buy yourself at a lower monthly rate. Confirm current plan structure directly, since it changes. The useful comparison is against buying sensors alone: two separate two-week wearing blocks a year costs roughly $200 in over-the-counter sensors, so the subscription premium is buying software and convenience, not data quality.
How accurate is the sensor?
Modern continuous sensors run a mean absolute relative difference of roughly 8 to 10 percent against laboratory reference values, which is good but not precise. Two sensors worn simultaneously on the same person can differ by 10 to 20 mg/dL. The first 12 to 24 hours after insertion are the least reliable and are best treated as calibration. A CGM also measures interstitial fluid rather than blood, and interstitial glucose lags blood glucose by roughly 5 to 15 minutes — which is why a sensor and a fingerstick disagree during a fast rise without either being wrong.
Do I need a prescription, and does my own doctor get involved?
A prescription is required for the sensors Levels ships, and it is issued through the platform’s affiliated telehealth network after you complete an intake questionnaire. Your own physician is not involved, and the prescription is tied to your enrollment. Since 2024 there is also an over-the-counter route: continuous sensors intended for adults not using insulin can be bought from a pharmacy or manufacturer without any prescription, paired with the manufacturer’s own app. That route removes both the telehealth step and the subscription.
Is the app worth ten times the cost of a sensor?
It depends entirely on whether you would interpret raw data on your own. The app is genuinely the best-designed in the category, and for some users that is the difference between running a two-week experiment and abandoning the sensor in a drawer. For a user who is comfortable reading a glucose curve, logging meals in a notes app, and drawing their own conclusions, the premium buys convenience rather than information. Neither answer is wrong — the question is a self-assessment, not a product comparison.
How long should I stay subscribed?
Most of the learning arrives in the first 30 to 60 days. Week one is mostly novelty and calibration. Weeks two through four surface the real findings: the three or four regular meals that spike you hardest, the effect of meal order and a post-meal walk, and what alcohol and short sleep do to your fasting number. Weeks five through eight test fixes and confirm they hold. After that the curve largely repeats. A defensible pattern is one to two months of wear, act on what you learned, then a repeat block six or twelve months later.
Can a CGM tell me if I have insulin resistance?
Not reliably, and this is the most important limitation to understand. A CGM shows how your glucose behaves, but says nothing about how much insulin your pancreas is producing to keep it there. Insulin resistance develops in a sequence: fasting insulin rises first while the pancreas compensates, then post-meal glucose drifts, then HbA1c climbs, and fasting glucose crosses a threshold last. Someone with a comfortable-looking glucose curve can already have significant resistance. Fasting insulin and HOMA-IR catch it years earlier, which is why bloodwork should come first.
Who should buy something else?
Anyone who wants a professional interpreting their data, since no dietitian consult is included in the standard membership — a platform that bundles consults answers that directly. Anyone who only wants the raw glucose curve, which an over-the-counter sensor supplies for roughly $50 to $100 per two-week wear. Anyone who has not run a blood panel, because fasting insulin and HOMA-IR answer the more important question first. And anyone with diabetes, on insulin, or on a sulfonylurea, who should be managed by their own clinician with a prescribed CGM rather than a consumer wellness subscription.