Optimizing
KPI control tower
Proves: Simulated reinforcement learning — deterministic seeded bandit (real ThompsonBandit), synthetic traffic, no live users. Rewards are illustrative latent CTRs. — back to the demo →
⟳ Simulated RL
Master business KPIs · progress toward target Optimizing
acquisition
Bounce rate
43.8%
· 89% to target
base 58.0%
↓ target 42.0%
Hero CTR
4.4%
· 8% to target
base 4.0%
↑ target 9.0%
Lead → application
20.1%
· 81% to target
base 12.0%
↑ target 22.0%
activation
Dormant reactivation
7.8%
· 20% to target
base 6.0%
↑ target 15.0%
revenue
Paid conversion
5.5%
· 58% to target
base 3.5%
↑ target 7.0%
B2B demo requests
2.1
· 3% to target
base 2.0
↑ target 6.0 per 1k
Per-scenario learning · which variant the bandit picked Optimizing
ANew anonymous viewer
learned winner: A1 ✓ · share 89%
A1
Daypart adaptive theme ★
share 89% · post 0.14 · ↑bounce
A2
B2B peer-logo hero
share 3% · post 0.054 · ↑b2b_demo
A3
Local-proof badge
share 8% · post 0.089 · ↑ctr
Ax
Recognize-return (ungated)
would-be CTR 0.2 · selected 0×
BExisting customer (email match)
learned winner: B2 ✓ · share 58%
B1
Welcome-back resume
share 19% · post 0.116 · ↑reactivation
B2
Tailored-offer module ★
share 58% · post 0.118 · ↑paid_conv
B3
Prestige / peer hero
share 21% · post 0.108 · ↑lead_to_app
Bx
Income-recite (ungated)
would-be CTR 0.2 · selected 0×
CEmailed lead who clicked
learned winner: C1 ✓ · share 80%
C1
Click-continuity ★
share 80% · post 0.146 · ↑lead_to_app
C2
Resume / abandoned
share 12% · post 0.128 · ↑paid_conv
C3
Scarcity / urgency
share 7% · post 0.102 · ↑lead_to_app
Cx
Comparison-call-out (ungated)
would-be CTR 0.21 · selected 0×
Trust rails · monitored continuously Watched
Provenance
100%
coverage across 9 variants ·
21 facts, each with a source.
hold-in-copy: 0 · missing source: 0.
every shown fact carries a source; hold facts only steer, never recited
hold-in-copy: 0 · missing source: 0.
every shown fact carries a source; hold facts only steer, never recited
Drift
0
arms paused. 6/6 TTL-governed facts fresh.
bought/enrich facts are TTL-governed; on source change Drift pauses the arm (pool.pause).
bought/enrich facts are TTL-governed; on source change Drift pauses the arm (pool.pause).
Hallucination / overclaim
0
ungated arms selected, of 3 blocked.
Blocked by construction (structural).
ungated/creepy arms are kept out of the pool — the bandit cannot select them.
ungated/creepy arms are kept out of the pool — the bandit cannot select them.