Dexcom G7 Goosenecking, Accuracy & What's Coming with G8
Inside Dexcom: goosenecking, accuracy, and a peek at G8
I sat down with Dexcom's CEO and CTO to ask the questions you've been asking me — about sensors that don't make it, day-one fuzziness, replacements, and what the next sensor is going to fix.
This is one of those conversations I think a lot of people have been waiting for. I had Jake Leach, Dexcom's president and CEO, and Peter Simpson, their chief technology officer, on the show together. Peter is one of the founders of the technology — he's been at Dexcom for about twenty-four years and has had a hand in basically every generation of the sensor. So when I asked questions about why a sensor does what it does, I was talking to the person who would actually know.
We covered the stuff you keep bringing to me: the G7 "goosenecking" failures, day-one accuracy, why some sensors don't make it to the end, how replacements and support are changing, the new adhesive, and a little bit about the next sensor, G8. I'm not going to give you the whole thing here — the details are worth hearing in their own words — but here's what stuck with me.
The setupWhere these answers came from
Some of this came out of the Dexcom Customer Advisory Council, which the company formed in January 2026 to hear directly from the people who use the product every day. I'm part of that council, and I'll be honest — I didn't just bring my own opinions to the table. I brought a pile of feedback from this audience. The fact that I got invited back for a second meeting told me they actually wanted to hear it.
Dexcom is publishing a report based on those meetings, laying out what they heard and the commitments they're making around product performance, communication, transparency, and replacements. This episode is, in a lot of ways, me asking them to explain those things out loud.
The big one
What "goosenecking" actually is
If you've seen this come up online, here's the plain-language version Peter gave me. When you deploy a G7, the applicator brings the sensor and the needle down under the skin, then retracts the needle and leaves the sensor behind. In a goosenecking failure, the sensor gets pulled back out with the needle — so it ends up not actually in your body. The result is a sensor that won't connect or won't read glucose. Essentially an out-of-the-box failure.
Two things I wanted people to hear clearly. First, Dexcom says their algorithm detects this, so it isn't a safety concern — it's a usability concern. But they take it seriously, because when you put a sensor on, you expect it to work. Second, because the devices are connected and constantly reporting performance, they saw the failure rate start climbing early in 2025 — before they understood the cause.
What they did next is the part I found genuinely interesting. They pulled people off their normal projects into what they call "tiger teams," meeting once or twice a day to chase it down. The root cause turned out to be the kind of thing that shows up when you scale: as manufacturing volume grows, you get more process variability. Once they identified where that variability was coming from, they were able to bring the rate back down.
"It does happen very rarely still, but we want to try and drive as close to zero as possible."
— Jake Leach, CEO, DexcomWhere it stands now: they say the failure rate is back to historically low levels, and there's technology coming later this year meant to push it even further — to the best they've ever had in G7. Could it still happen to someone? Yes, rarely. But the bigger takeaway, in my read, is that this problem taught them to monitor a lot more in real time. Peter said they're now watching dozens of attributes and reacting to small signals before they ever reach the field.
The part people misread
When a sensor "fails," it isn't always broken
This is the section I'd most want a newly diagnosed person to hear, because the word "failed" is misleading. A lot of what we call a sensor failure isn't a defective sensor at all — it's physiology.
Peter explained it like this. On a benchtop, these sensors are extremely accurate — we're talking low single-digit MARD (that's the standard accuracy measure). The variability shows up when the sensor meets a living body. Put the same sensor on ten different people and you can get ten different experiences.
When you insert a sensor, you create a tiny wound, and your body kicks off a wound-healing response — sending in cells that are very metabolically active. The sensor is working fine. Your body is just temporarily interfering with glucose reaching it. Once that settles, the environment stays stable for a long time. At the very end of wear, the opposite happens: the body starts to encapsulate the sensor, cutting it off from the glucose and oxygen it needs.
In both cases, the algorithm is watching signal quality. If it doesn't trust what it's seeing, it will temporarily blank the data or shut the sensor down rather than show you a number that might be wrong. So "sensor failed" can simply mean the algorithm decided it could no longer stand behind the reading.
On the question everyone asks — how often does a sensor not make it? — Jake said the goal is for at least 80% of sensors to reach full wear time, that most do make it to day ten, and that the fifteen-day sensors are actually surviving a little better than they did in the clinical study. When one ends early, it's often at day 13 or 14, and usually it's the physiologic environment, not a defect.
Support & replacements
Why a replacement call can feel like an interrogation
I asked directly about support and replacements, because that's where a lot of the frustration lives. On support, Jake pointed to the new MyDexcom account and better tools for their agents, with the goal of fixing the thing where you have to repeat your information over and over. His line was that they want to set the standard for customer service in the whole category — and that the company should be relieving the burden of diabetes, never adding to it.
On replacements, here's a reality I didn't fully appreciate: there are people calling Dexcom who aren't even users, trying to get replacement sensors to resell. That's part of why a genuine customer can feel like they're being grilled — the company is trying to confirm you're a real person who actually needs a sensor. Dexcom says they're building better, lower-friction ways to verify users so real customers can get what they need faster. The stated north star is simple: nobody should go a day without a sensor.
"We're relieving the burden of diabetes — we should never be adding to your burden as a user."
— Jake Leach, CEO, DexcomWhat's next
A better adhesive now, and a peek at G8
On adhesive — Dexcom is in the middle of rolling out the third version of the G7 adhesive, and it's already reaching people in the field, with plans to bring it across the portfolio.
And then there's G8. I tried to ask about it in a thoughtful way: if G8 had been your first sensor ever, which of today's problems simply wouldn't exist? Jake described it as the next big advancement — it adds multi-analyte sensing, and it carries brand-new, first-of-its-kind technology that measures an additional signal so the sensor can essentially self-adjust while you wear it. Think of it as calibrating itself, on its own, instead of you doing it. That should take aim at a lot of the variability people have run into. He was candid that a little first-day variability will probably still exist — there's always more to improve — but the direction is clear.
I'll note: no firm release date was given in this conversation, so I'm not going to invent one. But it didn't sound as far off as you might think.
If you only remember five things
The short version
- Goosenecking is when the sensor gets pulled back out with the needle during insertion. Dexcom says it's detectable, not a safety issue, traced to manufacturing variability from scaling, and now back to historically low rates — with more improvement coming this year.
- "Sensor failed" often isn't a broken sensor. Day-one fuzziness is usually a wound-healing response; end-of-wear shutoffs are usually the body encapsulating the sensor. The algorithm hides data it can't trust.
- The goal is 80%+ of sensors reaching full wear, most make it to day ten, and fifteen-day sensors are surviving a bit better than in the trial.
- Replacement friction exists partly because non-users try to game the system. Dexcom says it's working on lower-friction verification for real customers.
- G8 adds multi-analyte sensing and a self-adjusting "auto-calibration" technology aimed at the variability people experience today.
There's a lot more nuance in the full conversation — including how Dexcom now monitors performance in real time, and Peter's deeper explanation of how the sensor and the interstitial fluid actually interact. Go listen to the whole thing.
Listen to the full episode
The details matter here, and they're best in Jake and Peter's own words. Tap play, then come ask questions in the community.
More from Dexcom: the Customer Advisory Council report · sensor replacement policyRead the full disclaimer
