Lily Max
How Lily Max turns product intelligence into measurable lift.
Your team sets the goal. Lily Max runs the optimization loop: discovering gaps, running experiments, and proving the lift with the same controlled tests a finance team will accept.
You set the goal. Lily Max runs the optimization loop.
Agents do the work; humans stay in control at the two points that matter: the goal and the approval.
Humans own the goal
Your team sets the objective and the guardrails: target ROAS, margin floors, brand rules. Lily Max never invents the destination.
Agents execute the work
Agents discover gaps, draft enrichment, and stand up experiments at catalog scale: the repetitive work no team can do by hand.
Humans approve what ships
Nothing goes live without sign-off. You review proposed changes and the projected impact, then approve what reaches the feed.
The loop
Every test makes the next test smarter.
01
Discover gaps
Score the catalog against what each surface rewards and rank the highest-impact attribute gaps.
02
Generate hypotheses
Turn each gap into a testable enrichment hypothesis with a clear expected outcome.
03
Launch experiments
Roll changes into matched-spend tests so the effect is isolated from seasonality and budget shifts.
04
Measure lift
Read the result with statistical rigor, then feed the winners back into the next round.
Google ships. Lily shipped.
When Google introduced Conversational Attributes in Merchant Center, a new product-data type designed to make catalogs legible to AI Mode, Gemini, and Business Agent. Lily Max shipped support for all six attributes within days.
The window between Google announcing a new AI-surface capability and your team actually shipping it determines whether you show up in the next wave of AI commerce or you don't. Lily Max is built so that window is days, not quarters. A catalog that's always scored, content that's always testable, a pipeline that's always ready to ship.
Matched-spend test · treated vs. control
Proof designed for scrutiny.
Lift is only useful if it survives the finance review. Every test is built to hold up to it.
- Matched-spend A/BTest and control cohorts run at the same spend, so lift reflects the feed change, not a bigger budget.
- Difference-in-differenceWe compare the change in treated products against untreated ones to strip out market-wide movement.
- Statistical significanceResults ship with confidence intervals and significance: the same scrutiny a client's CFO will apply.
Quality scoring supports the performance engine.
Compliance
Every attribute respects Google policy and your brand rules before it ships.
Conversion
Scoring weights the attributes that actually move add-to-cart and ROAS.
Relevance
Products match more of the queries and audiences they should be eligible for.
Differentiation
Enrichment makes each product distinct, not a templated duplicate of the last.
Get started
Start with your GMC feed. Expand across every AI-mediated commerce surface.
Run a free GMC opportunity map and see where richer, AI-ready attributes can lift visibility, conversion, and ROAS.