HOW TRELLIS WORKS

Spend less time guessing.
Confidently manage your performance marketing campaigns.

very Catalyst Audit grounds its recommendations in your own order record, your real business economics (gross margin, AOV, COGS, target CPA and ROAS), your Changelog of the campaign changes you’ve recorded, and a statistical core that compounds with every audit you run. Each audit reflects what your specific account does, and sharpens as your account history accumulates.

The first-party spine FIG. 2.1 · Trellis Methodology

THE MATH BEHIND EVERY RECOMMENDATION

Every recommendation shows its work and names what would change it.

Each recommendation in your Catalyst Audit pairs an action with the evidence behind it, a credible alternative the same data supports, and a flip condition the next audit can verify. When the evidence is thin, the audit calls it out and holds the finding for monitoring rather than pushing a recommendation.

Recommendation Confident

Reduce target CPA on Brand | Exact by 18% over the next two weeks.

Evidence [BAYESIAN: 95% CI $46–$58]

Conversion rate steady at 6.4% across 412 conversions in the audit window. Cost-per-conversion has trended 11% above your target CPA for three consecutive weeks.

Depends on

Sustained conversion rate above 5.8% and no material change in landing-page bounce behavior.

Credible alternative

Hold target CPA flat and reinvest the savings into Brand | Phrase, where impression share is volume-capped and conversion quality is matching exact-match performance.

Flip condition

Conversion rate drops below 5.8% for two consecutive weeks, OR impression-share-lost-rank on Brand | Phrase exceeds 35%.

Understanding a Catalyst Audit, in four moves.

Each audit is calibrated to your account. It compounds with every run.

Every Catalyst Audit runs on a Bayesian statistical core. It reads your foundation and computes a posterior — the updated probability each finding holds true given what your account has actually produced. Recommendations surface only when that probability clears the action threshold; everything else waits. The posterior carries forward, so each audit starts where the last one ended.

IIt reads your foundation.

Orders, refunds, COGS, attribution lineage, and every change you’ve recorded since the last audit. Trellis stays read-only; you are the actor and the recorder.

IIIt grades the evidence.

Every finding is calculated by the statistical core against what your account has actually done — not generated by a prompt. The math travels; the data stays per-account.

IIIIt holds what the data can’t support.

Insufficient data is a valid result, not a failure. You see what’s worth watching alongside what’s ready to act on. Never a recommendation forced from thin data.

IVIt compounds with each run.

After three audits, the model can project budget impact. After six, it runs full inference calibrated to how your account responds, not to a generic benchmark.

YOUR DECISION TRAIL

You record the change. The next audit evaluates the outcome.

The Changelog connects each Catalyst Audit to the next. It stores every recommendation surfaced in a Catalyst Audit alongside every campaign change you record after making it on the ad platforms, so the next audit can evaluate what was acted on against what it actually produced. Trellis stays read-only across your accounts. You control which changes you make.

The compounding loop FIG. 4.1 · Trellis Methodology

THE TRELLIS HEALTH SCORE

A Health Score weighted toward margin, not revenue.

Every Catalyst Audit opens with a single Health composite, scored across five signals about your account. Margin pulls heavier than volume, so your audit will not congratulate you on growing top-line revenue while gross margin is moving the wrong way. Whichever signal is weakest becomes the next-best-action message at the top of your report.

Margin: Profitability against break-even targets.
Tracking: Platform-reported conversions reconciled against the first-party order record.
Cadence: Whether audits are recent enough that the data behind recommendations is still current.
Context: Whether the profile has the data the audit needs: COGS, attribution, business metrics.
Follow-through: Whether prior recommendations have been acted on, observed, and recorded.

CONFIDENCE, NAMED IN THREE TIERS

Some findings are ready to act on. Others are worth watching or holding for further review.

A single confidence score conflates two situations that should stay distinct: a finding with measured evidence behind it, and a finding the data is too thin to evaluate. From one number you cannot tell which is which. Your Catalyst Audit reports them in three buckets so you can. Confident findings clear the threshold to act on now. Directional findings are worth watching but not yet decisive. Held findings are recorded as open questions the next audit will return to when more data arrives.

The confidence buckets FIG. 3.1 · Trellis Methodology

Your first Catalyst Audit

Ready when you are.

Connect your accounts and your first Catalyst Audit runs in minutes. Every recommendation is grounded in your first-party data and your real business economics.