On this page

What a drive test is for

Performance counters tell you that a cell is underperforming. They do not tell you where, or why. A drive test produces the geographic evidence — signal level, quality and serving cell against position — that turns a statistical symptom into a physical cause with a location attached to it.

That evidence is also what makes a recommendation defensible. A tilt change justified by a measured trace survives review; one justified by a prediction does not.

Designing the route

The route determines what the data can prove. Three route types cover most work, and they are not interchangeable.

  • Cluster route — a contiguous group of sites driven as one. Used to assess mobility and dominance between neighbours, because handover behaviour is only visible when the boundaries are crossed.
  • Benchmark route — a fixed, repeatable route used to compare operators, or the same network before and after a change. Its value comes entirely from being identical every time.
  • Single site verification — a route around one site to confirm it radiates as designed after installation or change. Checks sector orientation, coverage footprint and basic service.

Record the route with the data. A trace whose route is not documented cannot be repeated, and a measurement that cannot be repeated cannot demonstrate improvement.

Scanner and device-based collection

The two collection methods answer different questions, and serious work usually needs both.

Collection methods compared
Method What it sees Best for
Scanner Every detectable cell on the band, regardless of what the network would serve. Independent of device behaviour and network decisions. Interference and dominance analysis, PCI conflict detection, neighbour discovery, coverage footprints
Device based What an actual subscriber gets — serving cell, real handovers, real throughput, real failures, including the signalling behind them. Service verification, accessibility and retainability, throughput, call testing

A scanner will show a polluted area even where the device happened to hold a usable connection. A device will show a dropped session the scanner has no way to explain. Collect both and the two datasets explain each other.

The core diagnostic

Almost every drive test conclusion comes from reading level against quality. Plot RSRP and SINR together and the problem classifies itself.

Diagnosis from RSRP against SINR
SINR good SINR poor
RSRP good HealthyNothing to do. Verify the samples are representative of the area rather than of one stationary moment. InterferenceSignal is present but contested. Investigate overshoot, overlap, PCI conflict and external sources.
RSRP poor Clean edgeWeak but uncontested. Tolerable at a genuine boundary; a problem if it is a service area. Coverage holeNo usable server. Needs a physical answer — coverage change or new infrastructure.

What each SINR band costs

SINR is not an abstract quality score — it selects the modulation and coding scheme, which sets throughput directly. This is the translation between a radio measurement and a user-visible outcome.

SINR band, modulation and consequence
SINR Typical modulation Throughput Action
≥ 20 dB256QAM or 64QAMAt peakMaintain
10–20 dB64QAM / 16QAMGoodMonitor channel quality reporting
0–10 dBQPSK, low coding rateReducedInvestigate interference or coverage
< 0 dBMinimum schemePoor, approaching link failureCoverage or interference action required

From log to recommendation

Post-processing is where a drive test earns its cost. A log file is not a deliverable; a defensible recommendation is.

  1. Validate the data. Check position lock quality, sampling rate and coverage of the intended route. Discard stationary periods that would otherwise weight the statistics toward one location.
  2. Classify by area, not by sample. One bad sample is noise. A contiguous run of bad samples across a defined stretch is a finding.
  3. Attribute to a serving cell. A problem without an identified server cannot be actioned. Where no cell dominates, that absence is itself the finding.
  4. State the physical cause. Coverage, interference, configuration or capacity — the same elimination order used for counter analysis.
  5. Recommend one change, and say how it will be verified. Name the re-drive that will confirm it worked.

Before and after, on the same route

A change is only demonstrated if the verification drive follows the same route, at a comparable time of day, with the same collection setup. Anything else compares two different measurements and calls the difference an improvement.

Engineering support

Need this applied to your network?

These are the methods we use on live networks. Send through your site list, KPI definitions and reporting template and we will come back with scope, method and timeline.