5G Drive Test KPIs: The Complete Field Reference 2026

An RF engineer on a rural site at midnight. The 5G SA cluster was commissioned 3 hours ago. SS-RSRP looks fine at -85 dBm. But throughput is 12 Mbps instead of the expected 800. Why? Because RSRP alone tells you nothing about beam configuration, MIMO rank, or MCS. The answer is in the Layer 1 KPI stack, and knowing how to read it is the difference between a 3-hour troubleshoot and a 3-day vendor ticket.

This reference covers every 5G NR KPI that matters in the field: definitions, 3GPP sources, quality thresholds, what the Android API hides, and a direct mapping from KPI anomaly to root cause.

The 5G NR KPI Stack

5G NR introduces a layered KPI hierarchy that is fundamentally different from LTE. Understanding where each metric sits in the stack is the first step toward meaningful diagnosis.

SS-RSRP: the coverage anchor

SS-RSRP (Synchronization Signal Reference Signal Received Power) is defined in 3GPP TS 38.215 §5.1.1. It measures the linear average power of the resource elements carrying the SS/PBCH block (SSB) reference signals. Range: -44 to -156 dBm.

The critical distinction from LTE RSRP: SS-RSRP is beam-specific. Each SSB beam has its own RSRP value. A site transmitting 8 beams produces 8 SS-RSRP measurements. The UE reports the beam with the highest RSRP, but the other 7 matter for handover decisions and coverage continuity. Practical thresholds:

• Excellent: > -80 dBm (outdoor macro)
• Good: -80 to -95 dBm
• Marginal: -95 to -110 dBm
• Poor: < -110 dBm (service at risk)

Indoor environments apply a 10-15 dB offset, shifting the “good” threshold to > -95 dBm.

SS-RSRQ and SS-SINR: interference and capacity

SS-RSRQ (Synchronization Signal Reference Signal Received Quality) expresses the ratio of SS-RSRP to total received power including interference. Range: -43 to 20 dB. It reflects both interference level and cell load simultaneously, which makes it less precise for root-cause isolation but useful as a first-pass indicator.

SS-SINR (Synchronization Signal Signal-to-Interference-plus-Noise Ratio) is the most reliable KPI for capacity prediction. Range: -23 to 40 dB. It directly measures the quality of the received signal against noise and interference, with no dependency on cell load. Target > 20 dB for peak throughput exploitation. Below 0 dB, the cell is in a coverage hole or experiencing severe pilot pollution.

CSI-RSRP vs SS-RSRP: two different jobs

CSI-RSRP uses CSI-RS (Channel State Information Reference Signals), transmitted more frequently and on more resource elements than SSB. The two KPIs serve different diagnostic purposes:

• Use SS-RSRP for coverage assessment and initial deployment validation
• Use CSI-RSRP for beamforming performance analysis, beam tracking quality, and MIMO precoding evaluation

A gap between SS-RSRP and CSI-RSRP at the same location often points to a beam management issue: the SSB sweep covers the UE, but the CSI-RS beam refinement is failing.

Layer 1 KPIs: What the Android API Hides

The Android telephony API exposes a useful but deliberately limited subset of modem data. For drive testing, this limitation is not a minor inconvenience — it hides the entire physical layer diagnostic picture. The Qualcomm DIAG interface gives access to everything the modem actually processes.

KPI	Android API	Qualcomm DIAG
SS-RSRP	Yes (approximate)	Yes (precise, per SSB)
SS-SINR	Yes (since Android 11)	Yes
CQI	No	Yes (wideband + subband)
RI (Rank Indicator)	No	Yes
PMI	No	Yes
MCS (DL/UL)	No	Yes
BLER	No	Yes
Resource Blocks allocated	No	Yes
HARQ statistics	No	Yes
Timing Advance	No	Yes

The gap between API data and DIAG data is the gap between “I see 5G” and “I understand why throughput is 12 Mbps.”

CQI (Channel Quality Indicator): ranges 0-15. Reported by the UE to the scheduler to indicate the maximum MCS it can support. CQI < 7 means reduced modulation order, directly limiting throughput. Wideband CQI gives a single value for the entire bandwidth; subband CQI reveals frequency-selective fading patterns that wideband averaging conceals.

RI (Rank Indicator) and PMI (Precoding Matrix Indicator): RI indicates the number of spatial layers the channel can support. RI = 1 means single-stream transmission; RI = 4 means four simultaneous spatial streams. PMI tells the gNB which precoding matrix to apply. These two values together define MIMO operating point and are invisible to the Android API.

MCS (Modulation and Coding Scheme): directly determines the bits-per-resource-element. MCS 28 (256-QAM) delivers 8.64 bits per resource element; MCS 0 (QPSK, low code rate) delivers 0.23. The ratio explains why 12 Mbps on a site that should deliver 800 Mbps is a Layer 1 problem, not a core network problem.

BLER (Block Error Rate): the fraction of transport blocks received in error before HARQ retransmission. Target < 10%. Above 10%, the scheduler is operating in a region where HARQ retransmissions consume a significant share of capacity.

For drive test work requiring full Layer 1 visibility, a professional Android drive test tool with native Qualcomm DIAG access eliminates the API gap entirely.

EN-DC KPIs: NSA-Specific Measurements

In Non-Standalone (NSA) 5G, the UE maintains dual connectivity: an LTE anchor and an NR secondary node running simultaneously. This architecture requires a separate set of KPIs beyond the standard NR RF measurements.

MCG (Master Cell Group) handles the LTE anchor side: it manages the control plane, mobility decisions, and NAS signaling through the MME. SCG (Secondary Cell Group) is the NR bearer: all 5G throughput flows through it.

Critical NSA KPIs that have no equivalent in SA testing:

• SCG Add Success Rate: the percentage of successful NR secondary cell additions. Below 90% indicates a significant NSA quality issue. Each failure maps to a specific RRC cause and must be correlated with NR coverage at the failure location.
• SCG Failure Rate: captures RACH failures, radio link failures, and configuration errors. Per 3GPP TS 38.331, SCG failures are reported via SCGFailureInformationNR in the RRC message. Each failure type points to a different root cause.
• EN-DC Throughput Split: the ratio of MCG (LTE) contribution to SCG (NR) contribution. If the SCG never contributes meaningfully, the NSA deployment is misconfigured or NR coverage is insufficient at the measurement point.
• PDCP Reordering Delay: cross-layer latency introduced by splitting PDCP PDUs across MCG and SCG paths. Elevated reordering delay increases application-layer latency even when RF conditions look acceptable.

KPI-to-Root-Cause Mapping

This is the section that converts a measurement campaign into a diagnostic. For each KPI anomaly, there is a specific interpretation and a specific next action.

Observed KPI	Likely Root Cause	Next Action
SS-SINR < 0 dB	Severe interference or coverage hole	Check neighbor cell list, pilot pollution
MCS < 5 (QPSK)	Coverage limited, RSRP < -110 dBm	Power/tilt optimization, site qualification
RI = 1 consistently	High spatial correlation (single path)	Antenna cross-polarization check, reflection
BLER > 10%	Interference or link instability	Check adjacent channel, verify ICIC config
SCG Add Fail > 10%	NR coverage gap or configuration error	Verify NR cell plan, check X2/Xn interface
TA > 500 µs	UE very far from cell (> 75 km equivalent)	Coverage edge, check site plan
CQI < 7	Poor channel, reduced throughput expected	Correlate with RSRP, SINR, interference

The table above is a starting point, not a complete decision tree. In practice, KPI anomalies rarely appear in isolation. A field cluster with SINR < 0 dB, RI = 1, and BLER > 10% simultaneously points to a specific scenario: the UE is at the edge of two strong beams from different sectors, experiencing pilot pollution, and the antenna configuration needs review.

RRC State Machine and Protocol Events

RF KPIs explain what the radio layer is doing. Protocol events explain what the network is deciding. Both are necessary to close a diagnostic loop.

RRC state transitions

RRC Connected to RRC Idle transitions that occur too frequently signal a misconfigured inactivity timer. Each transition introduces a re-establishment delay of 50 to 200 ms, which compounds under mobile conditions. If measurement logs show dozens of Connected-to-Idle transitions per minute, the inactivity timer is too aggressive for the service traffic pattern.

Measurement report events

The network configures the UE with measurement event triggers defined in 3GPP TS 38.331:

• Event A3: a neighbor cell becomes stronger than the serving cell by a defined offset. Triggers intra-frequency handover.
• Event B1: an inter-RAT cell exceeds a configured threshold. Used for 5G-to-LTE fallback.
• Event B2: the serving cell falls below a threshold while a neighbor exceeds another. Used for inter-RAT handover when coverage is degrading.

All of these events are carried in MeasurementReport RRC messages. Capturing and decoding them is the only way to understand why a handover was triggered, whether it was appropriate, and whether the executed handover was the right target cell.

5G SA NAS events

In Standalone deployments, the NAS layer provides a second diagnostic dimension:

• Registration Request / Accept / Reject: a rejected registration with cause code #7 (5GS services not allowed) or #22 (congestion) points to AMF configuration or capacity issues that are completely invisible to RF-layer KPIs.
• PDU Session Establishment: failures here indicate SMF or UPF issues, not radio problems. Distinguishing these from RF failures saves significant diagnostic time.
• Authentication procedure: authentication failures point to AUSF or UDM configuration issues in the 5G core.

A 3GPP protocol decoder that renders NAS messages alongside the RF KPI timeline makes this correlation possible without manual log parsing.

Handover latency

Target handover latency (from MeasurementReport transmission to RRCReconfigurationComplete) should be below 50 ms for mobility-sensitive services. Latency above 100 ms during handover causes visible disruption on video and real-time applications. Capturing this metric requires timestamped L3 message logging, not just RF KPI sampling.

Conclusion

A drive test without Layer 1 KPIs is like diagnosing an engine with only a speedometer. SS-RSRP, SS-SINR, RI, MCS, BLER, and EN-DC events together form the complete diagnostic picture for 5G NR. Coverage looks fine; throughput is broken; the answer is in MCS and RI, not in RSRP.

The practical implication: any drive test tool that cannot access Qualcomm DIAG is structurally unable to close a Layer 1 diagnosis. The Android API gap is not a minor limitation — it is the gap between “we measured signal strength” and “we diagnosed the network.”

For more on 5G testing methodology and how to structure a complete measurement campaign, see the 5G network testing tool guide.

Which 5G KPI has been the most surprising to discover in the field — and what root cause did it reveal? Share your experience in the comments.