LTE RSRQ average
LTE signal quality ratio. Degraded RSRQ despite good RSRP = high interference.

RSRP min / avg ✕ / max
Floating bar shows full drive range. ✕ marker = average. Wider spread = more RF variability.

NR SS-SINR range
Min/avg/max SINR over full drive. Shows signal quality variation across route.Source: nr_cell_meas.nr_servingbeam_ss_sinr_1

NR SS-RSRQ average
Reference Signal Received Quality — combines RSRP and interference. Used for cell reselection.Source: nr_cell_meas.nr_servingbeam_ss_rsrq_1

NR measurement samples
Total rows in nr_cell_meas for this operator. More = better route coverage picture.Source: COUNT(*) from nr_cell_meas

LTE RSRP average
LTE anchor signal (5G NSA mode). Source: lte_cell_meas.lte_inst_rsrp_1.

LTE RSRQ average
LTE signal quality ratio. Degraded RSRQ despite good RSRP = high interference.

LTE RSRP range
Min/avg/max LTE RSRP over full drive. Shows LTE anchor signal variation.Source: lte_cell_meas.lte_inst_rsrp_1

NR active % — NR RSRP confirmed
% of GPS samples where NR serving beam RSRP was actively measured (nr_cell_meas.nr_band_1 IS NOT NULL AND nr_servingbeam_ss_rsrp_1 IS NOT NULL). Previous method using raw row count was wrong — nr_cell_meas has rows at every posid even when UE is LTE-only, inflating NR% to near 100%. on the route.

ENDC (EN-DC) Success Rate
% of LTE+NR (E-UTRA/NR Dual Connectivity) addition attempts that succeeded. Source: pp_events_no_pause — "ENDC SgNB Addition Success" / "ENDC SgNB Addition Attempt". Low ENDC SR = devices struggle to attach to 5G.

NR PCC average bandwidth
Average configured NR bandwidth in MHz for the Primary Component Carrier. Source: pp_nr_rb_stat.nr_bw_1. Berlin deployments typically use 80MHz on Band 78 (n78, 3.5GHz).

NR PCC band distribution
% of NR measurement rows per primary serving band. Source: GROUP BY nr_band_1. Berlin operators predominantly use n78 (3.5 GHz Sub-6 GHz). n3 (1800 MHz) used as SCC or fallback.

LTE PCC band distribution
% of LTE measurement rows per band, computed by mapping EARFCN → band number. DT uses B3 (1800 MHz) as primary anchor. O2 uses B3+B1. VF and 1&1 use B1 (2100 MHz).

NR average MIMO rank
Average Rank Indicator (1–4). Number of parallel spatial MIMO streams. Higher = better channel conditions.Source: nr_cell_meas.nr_ri_1

NR DL Resource Block utilisation %
Average % of available DL resource blocks actively scheduled. Source: pp_nr_rb_stat.nr_dl_rb_percent_1. Low % = network not congested. High % during test = heavy traffic demand or small cell.

NR UL Resource Block utilisation %
Average % of UL resource blocks used. Source: pp_nr_rb_stat.nr_ul_rb_percent_1. UL RB% typically lower than DL reflecting asymmetric mobile traffic patterns.

NR Channel Quality Indicator average
CQI (0–15) reported by UE. Higher = better channel. CQI 8 → typically 64QAM. CQI 12+ → 256QAM. Source: pp_statement_sum_nr_dl_modulation_changes.nr_cqi.

NR Rank Indicator average
Number of MIMO spatial streams (layers) used. RI=1 = single stream, RI=2 = 2-layer MIMO (doubles theoretical peak). Source: nr_dl_modulation_changes.nr_ri.

NR DL modulation distribution
% of modulation-change events per order. Source: pp_statement_sum_nr_dl_modulation_changes.nr_dl_modulation_order_1.
QPSK: 2 bits/sym — poor signal
16QAM: 4 bits/sym — medium signal
64QAM: 6 bits/sym — good signal
256QAM: 8 bits/sym — excellent signal
More 64QAM/256QAM = better spectral efficiency.

NR CA % vs LTE CA %
Side-by-side comparison of NR and LTE carrier aggregation usage. NR CA = nr_band_2 IS NOT NULL. LTE CA = lte_earfcn_2 IS NOT NULL.

TWAMP one-way delay
Outbound = device→server. Inbound = server→device. Asymmetry indicates routing differences. VF shows 401ms outbound vs 165ms inbound.

TWAMP first packet time
Time to receive first TWAMP response packet after session start.Source: twamp_sub — first row timestamp − session_start

TWAMP session success rate
% of TWAMP sessions completed without timeout or error.Source: statement_sum_twamp twamp_timing_tr19a_twamp_operation_success

DNS + server request processing
Time after TCP handshake until first data byte — DNS lookup and server processing delay.Source: statement_sum_http_download first_byte_received − server_connected (post-handshake)

HTTP DL transfer duration
Time from first byte to last byte. Pure data transfer time at network speed.Source: statement_sum_http_download last_byte_received − first_byte_received

HTTP UL Time To First Byte Sent
Time from connection start to first upload byte transmitted.Source: statement_sum_http_upload first_byte_sent_time − server_connecting_time

HTTP UL transfer duration
Time from first to last upload byte. Measures sustained UL throughput window.Source: statement_sum_http_upload last_byte_sent − first_byte_sent

OpenAI TTFT percentiles
P50=typical, P90=near worst-case, P95=worst-case AI response start time.Source: statement_sum_open_ai_prompt_test first_token_received_time − start_time → percentiles

HTTP DL peak throughput
Maximum download speed in any single session.Source: pp_statement_sum_http_download.app_tp_dl_mbps_max

HTTP DL 95th percentile
95% of sessions above this. Meritech Basic Speed 95% KPI. Computed: pct(session_avgs, 95).

DL vs UL average comparison
Side-by-side DL vs UL average throughput for this test variant.Source: pp_statement_sum_http_download.app_tp_dl_mbps_avg pp_statement_sum_http_upload.app_tp_ul_mbps_avg

DL Fixed max
HTTP DL peak — Fixed test sessions.pp_statement_sum_http_download app_tp_dl_mbps_max WHERE session_count=1

DL Fixed p95
HTTP DL 95th percentile — Fixed test.pp_statement_sum_http_download app_tp_dl_mbps_95th_percentile WHERE session_count=1

DL Fixed TTFB
Time To First Byte — Fixed test. TCP connect to first data byte.statement_sum_http_download first_byte_received − server_connecting_time

DL Capacity max
HTTP DL peak — Capacity test sessions.pp_statement_sum_http_download app_tp_dl_mbps_max WHERE session_count>1

DL Capacity p95
HTTP DL 95th percentile — Capacity test.pp_statement_sum_http_download app_tp_dl_mbps_95th_percentile WHERE session_count>1

DL Capacity TTFB
TTFB — Capacity test. Connect overhead on large file.statement_sum_http_download first_byte_received − server_connecting_time

HTTP UL peak throughput
Maximum upload speed in any single session.Source: pp_statement_sum_http_upload.app_tp_ul_mbps_max

HTTP UL 95th percentile
High-end consistent upload speed, excluding top 5% outliers.Source: pp_statement_sum_http_upload.app_tp_ul_mbps_95th_percentile

HTTP UL transfer duration
Time from first to last upload byte. Measures sustained UL throughput window.Source: statement_sum_http_upload last_byte_sent − first_byte_sent

UL Fixed max
HTTP UL peak — Fixed test sessions.pp_statement_sum_http_upload app_tp_ul_mbps_max WHERE session_count=1

UL Fixed p95
HTTP UL 95th percentile — Fixed test.pp_statement_sum_http_upload app_tp_ul_mbps_95th_percentile WHERE session_count=1

UL Fixed TTFB
First upload byte sent — Fixed test.statement_sum_http_upload http_upload_first_byte_sent − server_connecting_time

UL Capacity max
HTTP UL peak — Capacity test sessions.pp_statement_sum_http_upload app_tp_ul_mbps_max WHERE session_count>1

UL Capacity p95
HTTP UL 95th percentile — Capacity test.pp_statement_sum_http_upload app_tp_ul_mbps_95th_percentile WHERE session_count>1

UL Capacity TTFB
First upload byte sent — Capacity test.statement_sum_http_upload http_upload_first_byte_sent − server_connecting_time

YouTube VST range
Min/avg/max Video Start Time. Shows spread of startup experience.Source: pp_statement_sum_youtube.youtube_duration_to_first_play_secs × 1000

NR DL throughput during YouTube
Avg NR DL speed during YouTube sessions — actual network load under video streaming.Source: nr_throughput joined to youtube posids nr_p_plus_scell_nr_pdsch_tput_mbps

Browse page load 90th percentile
Worst-case load time for 90% of sessions.Source: pp_statement_sum_browse.duration_time_secs × 1000

Browse session throughput
Avg throughput during browse session (total bytes / duration).Source: statement_sum_browse total_bytes / load_duration

YouTube vs Browse NR throughput
NR DL throughput comparison during YouTube vs Browse. Shows how different apps load the network.Source: nr_throughput joined to posids during each test nr_p_plus_scell_nr_pdsch_tput_mbps

Pair Success Rate (Option A)
paired / (paired + mo_unpaired + mt_unpaired). Section 28.5 algorithm: an MO leg paired with its MT counterpart within 5s constitutes a successful pair. Unpaired legs are call attempts whose other side was never seen.

Pair Delta (seconds)
Time between MO call_init_time and MT call_init_time within a paired record. Smaller is better; expected ~1s on healthy networks.

MO Call Success Rate (Option B)
Fraction of paired records where call_result_mo == 'Success'. Distinct from Pair SR: this measures call completion within paired calls, not pairing success.

MT Call Success Rate (Option B)
Fraction of paired records where call_result_mt == 'Success'.

POLQA SWB MOS (MO)
ITU-T P.863 Super-Wideband POLQA score for the MO leg of each paired call. Range 1.0 to 5.0, higher is better.

Pair Success Rate (Option A)
paired / (paired + mo_unpaired + mt_unpaired). Section 28.5 algorithm (reciprocal phone match plus 5s time window) extended to WhatsApp per Section 32. Unpaired legs are call attempts whose other side was never seen.

MO Call Success Rate (Option B)
Fraction of MO records (paired plus unpaired_mo) where call_result_mo == 'Success'. Distinct from Pair SR: this measures call completion across all MO attempts, including unpaired ones.

MT Call Success Rate (Option B)
Fraction of MT records (paired plus unpaired_mt) where call_result_mt == 'Success'.

Pair Delta (seconds)
Time between MO call_init_time and MT call_init_time within a paired record. Smaller is better; sub-second is typical on healthy WhatsApp connections.

MO Conversation Duration
Time between call established and call end as observed by the MO leg. Source: paired[].mo_side.call_call_duration_ms.

MT Conversation Duration
Time between call established and call end as observed by the MT leg. Source: paired[].mt_side.call_call_duration_ms. MO and MT durations typically differ by a few hundred milliseconds due to state machine timing.

ViSQOL MOS · Downlink
Per-segment ViSQOL score for the receiving direction of each WebRTC OTT call. Range 1.0 to 5.0, higher is better. Aggregated across both cars (Section 33.1 replica-combine, weighted by n_segments_dl).

ViSQOL MOS · Uplink
Per-segment ViSQOL score for the transmitting direction. Weighted by n_segments_ul. UL typically scores lower than DL on cellular uplink-limited networks.

MOS DL 5th percentile
The worst 5% of DL segments. Recomputed from raw _raw.segments[] across both cars (not averaged from per-car p5). Reflects the bottom of the quality envelope.

DL Packet Loss percentage
Per-segment packet loss on the receiving direction, weighted by n_segments_dl. Healthy WebRTC connections show under 1%.

DL Jitter (ms)
Per-segment DL jitter weighted by n_segments_dl. WebRTC typically tolerates jitter up to about 30ms before audio quality degrades noticeably.

App-layer DL Throughput (Mbps)
Codec-level DL throughput observed during the OTT call. Source: pp_ott_voice_mos.ott_app_tp_dl_mbps_avg, weighted by n_segments_dl.

OpenAI TTFT percentiles
P50=typical, P90=near worst-case, P95=worst-case AI response start time.Source: statement_sum_open_ai_prompt_test first_token_received_time − start_time → percentiles

Total AI response time
First send → last token. Source: response_ms. Meritech AI Response Time KPI.≤1500ms=100 ≤2000ms=80 ≤2500ms=65 ≤3000ms=50

AI token generation time
first_token_received → completion. Time for model to stream all tokens after first. Separates server queue delay from generation throughput. Source: statement_sum timestamps.

OpenAI TTFT range
Min/avg/max Time To First Token across all sessions.Source: statement_sum_open_ai_prompt_test first_token_received_time − start_time

RF context during OpenAI sessions
NR RSRP/SINR during OpenAI sessions. Shows correlation between signal quality and AI latency.Source: nr_cell_meas joined via posid during openai sessions nr_servingbeam_ss_rsrp_1, nr_servingbeam_ss_sinr_1

ai_connection_test.aggregate.success_ratio
Aggregate ICMP success rate across all AI service hostnames. Section 24.2.4 Reliability KPI.

ai_connection_test.aggregate.basic_latency_p5
Best 5% of ICMP latencies across all AI hostnames. Section 24.2.4 Performance KPI.

ai_connection_test.aggregate.top_latency_p95
95th percentile of ICMP latencies. Worst-of-2 replica-combine fallback. Section 24.2.4 Performance KPI.

ai_connection_test.aggregate.jitter
Mean of |RTT[i]-RTT[i-1]| across per-packet samples. Section 24.2.4 Performance KPI.

web_browsing.aggregate.success_ratio
Aggregate success rate across all browse URLs. Section 24.2.3 Reliability KPI.

web_browsing.aggregate.session_time
Average page load time across all sessions. Section 24.2.3 Performance KPI.

web_browsing._raw.n_sessions
Total browse sessions executed across all URLs and bundles.

web_browsing._raw.n_urls
Distinct URLs tested. URL list comes from data (not hardcoded) per Section 24.4.2 Q4.

youtube_video.aggregate.success_ratio
YouTube on-demand video test success rate. Section 24.2.3 Reliability KPI.

youtube_video.aggregate.video_start_time
Loading-time to first-play. Derived from event timestamps when pp_statement_sum_youtube unavailable. Section 24.2.3 Performance KPI.

youtube_video.aggregate.vsqi
Video Service Quality Index (VMOS). Section 24.2.3 Performance KPI. Not derivable from current merged.db schema - emitted as null with _raw.vsqi_status flag.

youtube_live.aggregate.success_ratio
YouTube live-stream test success rate. Section 24.2.3 Reliability KPI.

youtube_live.aggregate.video_start_time
Loading-time to first-play for live streams. Section 24.2.3 Performance KPI.