diff --git a/BerlinMOD/benchmarks/batch/bench/bench_mduck.sh b/BerlinMOD/benchmarks/batch/bench/bench_mduck.sh
index d07ee28..2e63d3e 100755
--- a/BerlinMOD/benchmarks/batch/bench/bench_mduck.sh
+++ b/BerlinMOD/benchmarks/batch/bench/bench_mduck.sh
@@ -33,6 +33,7 @@
 set -euo pipefail
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
 BERLINMOD_DIR="$(cd "${SCRIPT_DIR}/.." && pwd)"
+QUERIES_SQL="${SCRIPT_DIR}/queries.sql"   # shared single-source query set
 
 # ── defaults ──────────────────────────────────────────────────────────────────
 DUCKDB="${DUCKDB:-duckdb}"
@@ -114,7 +115,7 @@ _duck_q() { "$DUCKDB" "$DBFILE" -noheader -list -c "$1" 2>/dev/null; }
 if $LOAD; then
   echo "=== Loading data into: $DBFILE ==="
   rm -f "$DBFILE"
-  LOAD_BODY="$(sed '/^SET VARIABLE DATADIR/d' "${BERLINMOD_DIR}/load_mduck.sql")"
+  LOAD_BODY="$(sed '/^SET VARIABLE DATADIR/d' "${SCRIPT_DIR}/load_mduck.sql")"
   LOAD_SQL="${MOBILITY_LOAD} SET VARIABLE DATADIR='${DATADIR}/'; ${LOAD_BODY}"
   "$DUCKDB" "$DBFILE" -c "$LOAD_SQL"
   echo "    done."
@@ -165,10 +166,31 @@ echo ""
 TIMEFILE=$(mktemp)
 trap 'rm -f "$TIMEFILE"' EXIT
 
+# ── extract each selected query from the shared queries.sql ──────────────────
+# queries.sql is the single canonical source; every query is delimited by a
+# `-- BerlinMOD Q<n>:` (or `-- BerlinMOD QRT:`) marker and runs to the next one.
+marker_for() {                       # q07 -> Q7 ; q13 -> Q13 ; qrt -> QRT
+  local q="$1"
+  if [[ "$q" == "qrt" ]]; then echo "QRT"; else echo "Q$((10#${q#q}))"; fi
+}
+extract_query() {                    # emit the SQL body for a marker
+  awk -v start="-- BerlinMOD ${1}:" '
+    index($0, start)==1 { grab=1; print; next }
+    grab && /^-- BerlinMOD Q/ { exit }
+    grab { print }
+  ' "$QUERIES_SQL" | grep -v '^\s*--' | tr '\n' ' '
+}
+
 for Q in "${QUERIES[@]}"; do
-  QFILE="${BERLINMOD_DIR}/${Q}.sql"
-  [[ -f "$QFILE" ]] || { echo "  [skip] ${Q} — SQL file not found"; continue; }
-  QSQL=$(grep -v '^\s*--' "$QFILE" | tr '\n' ' ')
+  QSQL=$(extract_query "$(marker_for "$Q")")
+  [[ -n "${QSQL// /}" ]] || { echo "  [skip] ${Q} — not found in queries.sql"; continue; }
+  # Validate once: a query the extension cannot run (missing function, etc.) is
+  # reported and EXCLUDED, never recorded as a phantom sub-second timing.
+  ERR=$("$DUCKDB" "$DBFILE" -c "${MOBILITY_LOAD} SET search_path='portable,main'; ${QSQL}" 2>&1 >/dev/null) || true
+  if echo "$ERR" | grep -qiE "error"; then
+    echo "  [FAIL] ${Q} — $(echo "$ERR" | grep -iE 'error' | head -1)"
+    continue
+  fi
   printf "  timing %-6s: " "$Q"
   for RUN in $(seq 1 "$RUNS"); do
     T0=$(date +%s%3N)
diff --git a/BerlinMOD/benchmarks/batch/bench/load_mduck.sql b/BerlinMOD/benchmarks/batch/bench/load_mduck.sql
new file mode 100644
index 0000000..5a9654c
--- /dev/null
+++ b/BerlinMOD/benchmarks/batch/bench/load_mduck.sql
@@ -0,0 +1,75 @@
+--------------------------------------------------------------------------------
+-- BerlinMOD Batch Load — MobilityDuck / DuckDB
+--------------------------------------------------------------------------------
+--
+-- This file is part of MobilityDB.
+-- Copyright(c) 2020-2026, Université libre de Bruxelles and MobilityDB
+-- contributors
+--
+--------------------------------------------------------------------------------
+--
+-- Loads the cross-platform BerlinMOD portability export (produced by
+-- berlinmod_portability_export() in berlinmod_export.sql) into the schema the
+-- shared batch queries (bench/queries.sql) expect — the SAME CSV set the
+-- MobilityDB and MobilitySpark runners consume, so all platforms benchmark the
+-- identical corpus.
+--
+-- Input CSVs (DATADIR is injected by bench_mduck.sh):
+--   vehicles.csv       : vehId, licence, type, model
+--   trips.csv          : tripId, vehId, trip        -- tgeompoint as hex-EWKB
+--   query_licences.csv : licenceId, licence
+--   query_instants.csv : instantId, instant
+--   query_points.csv   : pointId, geom             -- geometry as EWKT
+--   query_periods.csv  : periodId, period          -- tstzspan as text
+--   query_regions.csv  : regionId, geom            -- geometry as EWKT
+--
+-- SRID is carried by the serialization and resolved by MEOS parsers — never
+-- parsed or reprojected by hand: the trip comes as hex-EWKB
+-- (tgeompointFromHexEWKB) and the static geometries as SRID-tagged EWKT
+-- (ST_GeomFromText accepts the SRID= prefix).  The th3index prefilter column is
+-- built in-engine from the lat/lon trajectory, exactly as a real ingest would.
+--------------------------------------------------------------------------------
+
+-- Default for standalone use; bench_mduck.sh strips this line and injects the
+-- real data directory.
+SET VARIABLE DATADIR='./data/';
+
+-- The runner selects `SET search_path='portable,main'`; create the schema so that
+-- selection resolves (the tables live in main and are found via the fallback).
+CREATE SCHEMA IF NOT EXISTS portable;
+
+-- ── Vehicles ──────────────────────────────────────────────────────────────────
+CREATE OR REPLACE TABLE Vehicles AS
+  SELECT vehId, licence, type, model
+  FROM read_csv(getvariable('DATADIR') || 'vehicles.csv', header = true);
+
+-- ── Trips ─────────────────────────────────────────────────────────────────────
+-- trip arrives as hex-EWKB (SRID embedded); trip_h3 is computed here.
+CREATE OR REPLACE TABLE Trips AS
+  SELECT tripId, vehId, tgeompointFromHexEWKB(trip) AS trip
+  FROM read_csv(getvariable('DATADIR') || 'trips.csv', header = true, all_varchar = true);
+ALTER TABLE Trips ADD COLUMN trip_h3 TH3INDEX;
+UPDATE Trips SET trip_h3 = th3index(transform(trip, 4326), 7);
+
+-- ── Query parameter tables (100-row selectors used by queries.sql) ────────────
+CREATE OR REPLACE TABLE QueryLicences AS
+  SELECT licenceId, licence
+  FROM read_csv(getvariable('DATADIR') || 'query_licences.csv', header = true);
+
+CREATE OR REPLACE TABLE QueryInstants AS
+  SELECT instantId, instant::TIMESTAMPTZ AS instant
+  FROM read_csv(getvariable('DATADIR') || 'query_instants.csv', header = true, all_varchar = true);
+
+CREATE OR REPLACE TABLE QueryPeriods AS
+  SELECT periodId, period::TSTZSPAN AS period
+  FROM read_csv(getvariable('DATADIR') || 'query_periods.csv', header = true, all_varchar = true);
+
+-- Points carry both the parsed geometry and the SRID-tagged EWKT text
+-- (queries.sql surfaces p.geomWKT); geomWKT is the export text verbatim.
+CREATE OR REPLACE TABLE QueryPoints AS
+  SELECT pointId, ST_GeomFromText(geom) AS geom, geom AS geomWKT
+  FROM read_csv(getvariable('DATADIR') || 'query_points.csv', header = true, all_varchar = true);
+
+CREATE OR REPLACE TABLE QueryRegions AS
+  SELECT regionId, ST_GeomFromText(geom) AS geom
+  FROM read_csv(getvariable('DATADIR') || 'query_regions.csv', header = true, all_varchar = true);