diff --git a/formulas/run-user-test.toml b/formulas/run-user-test.toml
new file mode 100644
index 0000000..30176e1
--- /dev/null
+++ b/formulas/run-user-test.toml
@@ -0,0 +1,109 @@
+# formulas/run-user-test.toml
+#
+# Persona-based UX evaluation against the harb stack.
+#
+# Type: sense — produces UX metrics, changes no code or contracts.
+# The formula spins up a full self-contained stack, runs Playwright against
+# all 5 personas, collects structured reports, then tears the stack down.
+
+[formula]
+id          = "run-user-test"
+type        = "sense"
+description = "Persona-based UX evaluation against the harb stack"
+depends_on  = [973]   # evidence directory structure must exist
+
+# ── Stack management ─────────────────────────────────────────────────────────
+# The formula is self-contained: it starts and stops its own stack.
+
+[stack]
+start_cmd  = "./scripts/dev.sh start"
+health_cmd = "./scripts/dev.sh health"
+stop_cmd   = "./scripts/dev.sh stop"
+
+# ── Inputs ───────────────────────────────────────────────────────────────────
+# 5 personas across 2 funnels. Each persona has a dedicated Playwright spec
+# that simulates the full journey: connect wallet → mint ETH → buy KRK →
+# stake → verify position.
+
+[[inputs.funnels]]
+name = "passive-holder"
+
+  [[inputs.funnels.personas]]
+  name    = "tyler"
+  display = "Tyler 'Bags' Morrison"
+  spec    = "tests/e2e/usertest/tyler-retail-degen.spec.ts"
+
+  [[inputs.funnels.personas]]
+  name    = "alex"
+  display = "Alex Rivera"
+  spec    = "tests/e2e/usertest/alex-newcomer.spec.ts"
+
+  [[inputs.funnels.personas]]
+  name    = "sarah"
+  display = "Sarah Park"
+  spec    = "tests/e2e/usertest/sarah-yield-farmer.spec.ts"
+
+[[inputs.funnels]]
+name = "staker"
+
+  [[inputs.funnels.personas]]
+  name    = "priya"
+  display = "Dr. Priya Malhotra"
+  spec    = "tests/e2e/usertest/priya-institutional.spec.ts"
+
+  [[inputs.funnels.personas]]
+  name    = "marcus"
+  display = "Marcus 'Flash' Chen"
+  spec    = "tests/e2e/usertest/marcus-degen.spec.ts"
+
+# ── Steps ────────────────────────────────────────────────────────────────────
+
+[[steps]]
+id          = "stack-up"
+description = "Spin up full stack (boots Docker services, waits for health)"
+run         = "./scripts/dev.sh start"
+
+[[steps]]
+id          = "run-personas"
+description = "Run all 5 personas via Playwright (workers=1, sequential to avoid account conflicts)"
+run         = "./scripts/run-usertest.sh"
+after       = ["stack-up"]
+
+[[steps]]
+id          = "collect"
+description = "Aggregate per-persona JSON reports from tmp/usertest-results/ into evidence/user-test/{date}.json"
+output      = "evidence/user-test/{date}.json"
+schema      = "evidence/README.md#user-test"
+after       = ["run-personas"]
+
+[[steps]]
+id          = "stack-down"
+description = "Tear down stack"
+run         = "./scripts/dev.sh stop"
+after       = ["collect"]
+
+[[steps]]
+id          = "deliver"
+description = "Commit evidence file to main and post summary to issue as comment"
+after       = ["collect"]
+
+# ── Products ─────────────────────────────────────────────────────────────────
+# Three outputs following the standard evidence delivery pattern:
+#   1. evidence file  → committed to main
+#   2. screenshots    → referenced inside the evidence file
+#   3. summary        → posted as issue comment
+
+[[products]]
+type        = "evidence"
+path        = "evidence/user-test/{date}.json"
+schema      = "evidence/README.md#user-test"
+destination = "commit"
+
+[[products]]
+type        = "screenshots"
+path        = "test-results/usertest/"
+destination = "evidence-ref"   # paths recorded in the evidence file, not committed separately
+
+[[products]]
+type        = "summary"
+destination = "issue-comment"