Network Truth
Not All Androids Are Equal: Configuration Stability Measured by Brand
Across two independent deployments on the same host network in Mexico, the device brand predicts configuration stability better than any other attribute, the rates replicate to within a percentage point, and the divergence traces to Android 13.
Across two independent deployments on the same host network in Mexico, the device brand predicts configuration stability better than any other attribute, the rates replicate to within a percentage point, and the divergence traces to Android 13.
Operators choose which devices to homologate, subsidize, and ship. Those choices are usually informed by price, availability, and spec sheets. Almost never by the one property this report measures: how well a device holds its carrier configuration once deployed. Over twelve months we measured steady-state internet APN configuration loss, the share of devices that lose working data configuration after their first week, across roughly 40,000 Android devices from two independent deployments on the same host network in Mexico.
Finding 1: a five-fold spread between brands
The share of devices experiencing at least one APN configuration loss over the period ranges from about 11 percent (OPPO) to about 65 percent (Infinix). In between: Samsung 19 percent, Redmi 33 percent, Motorola 44 percent, POCO 46 percent. Device brand is the strongest single predictor of configuration stability in our data, stronger than device age, Android version alone, or Carrier Privileges status.
Finding 2: the numbers replicate
Because the devices come from two independent deployments, every brand rate was measured twice, on separate fleets, enrolled separately. The two measurements agree to within roughly one percentage point for every major brand: Motorola 44.7 vs 44.1 percent, POCO 46.5 vs 45.0, Redmi 33.3 vs 33.1, Samsung 18.8 vs 18.8, OPPO 10.5 vs 11.4. This is not one fleet's anecdote. It is a reproducible property of the devices themselves.
Finding 3: the divergence starts at Android 13
The brand gap is not constant across OS versions. On Android 10 to 12, Motorola and Redmi devices lose their APN configuration at 8 to 22 percent, unremarkable rates. At Android 13 both jump sharply: Motorola to 52.5 percent, peaking at 67.2 percent on Android 14; Redmi to 43.6 percent. Neither recovers on later versions. Samsung, measured across the same six OS versions on the largest samples in the study, stays between 10 and 21 percent with no cliff. OPPO moves in the opposite direction entirely, improving from 28.4 percent on Android 11 to under 10 percent from Android 13 onward. Whatever changed in Android 13's handling of carrier configuration, manufacturers absorbed it very differently, and the difference persists three OS generations later.
Finding 4: intensity compounds the gap
Affected devices are not affected equally. A Motorola device on Android 13 averages 12.8 configuration losses over the period, the highest intensity in the study, against a Samsung average of 2.7 on the same version. For an operator, this is the difference between a device that occasionally needs one correction and a device that generates a support contact pattern.
What this means for device strategy
Two devices at the same price point can differ five-fold in how reliably they hold the configuration that makes data work. That difference is invisible in spec sheets, invisible in lab homologation, and invisible to network-side monitoring, because the failures happen quietly on devices in the field and, on a managed fleet, are corrected before anyone reports them. It becomes visible only when configuration state is measured continuously at the device layer. For operators, the practical use is direct: configuration stability belongs in homologation criteria and device-mix planning, next to price and availability.
About the data
Rates are the share of observed devices per brand experiencing at least one steady-state internet APN configuration loss between July 2025 and June 2026, after deduplication, exclusion of no-op events, collapse of sub-minute repeats, exclusion of implausible reporting cohorts, and exclusion of each device's first seven days. Brand-version cells with fewer than 150 observed devices are not reported. Both deployments run on the same host network, so radio environment differences between brands are minimized by design. Measurement is performed on-device by SDKCX through Android Carrier Privileges.