delectable AI · OPERATOR STUDIO

The merchandiser & marketer's control surface for the agent. Tune the levers that shape what the agent recommends, surfaces, substitutes, and promotes. Run experiments. See attribution. Every change is auditable. Every effect is measurable.

Experiments & holdouts Seven tests running. Two with statistical winners.
Active experiments
7
across 5 audiences · 3 surfaces
Significant winners
2 / 7
ready to promote
Cumulative lift
+$1.84M
incremental Q3 · across winning variants
Experiment
Audience
n-shoppers
Basket lift
Conv. lift
p-value
Duration
Status
Attribution & measurement The full picture — who got credit, who actually moved the basket.
Q3 attributed revenue
$19.8M
across all channels · agent-routed
Incremental lift
+18%
vs control cohort · cumulative
Avg ROAS
3.2×
RMN placements · multi-touch model
Channel contribution · last 90 days

Where the revenue lands

Multi-touch attribution model · 30-day decay window · agent-mediated only. Channels are tagged via the channel field on the UCP request envelope.
Multi-touch path explorer · top 5

How shoppers actually arrive

ROAS matrix · audience × placement

Where to lean in. Where to pull back.

Cell shading: red = ROAS < 1.5 · gold = 1.5–3.0 · green = 3.0–5.0 · deep green = >5.0
Cumulative incremental lift · 12-week pilot window

Cohort B over Cohort A · the agent's incremental contribution

$25M $20M $15M $10M $5M W1 W3 W5 W7 W9 W11 W12 +$22M flat baseline
Cohort B · Delectable agent live Cohort A · keyword-search control