The homepage shows the loop. This page shows the parts. Every event captured, every model run, every dollar moved, every agent action, every forecast — the same corrected signal, end to end.
Capture every event, server‑side, deduplicated.
Five attribution models in parallel, never one lens.
Spend reallocates to what actually earned.
Agents that read the corrected signal, not the pixel.
Fourteen days ahead, with confidence bands.
Capture every event, server‑side, deduplicated.
Five attribution models in parallel, never one lens.
Spend reallocates to what actually earned.
Agents that read the corrected signal, not the pixel.
Fourteen days ahead, with confidence bands.
A single tag captures sessions, orders, and LTV across your storefront — server‑side where it counts. Then the same order claimed by four platforms gets reconciled to one truth.
Meta, Google, TikTok, Pinterest, Snapchat — all via Conversions API. No pixel‑only fallback.
First‑party cookies, hashed email, server‑sent events. 87% of sessions resolved to known users.
iOS 17 and cookieless visitors get probabilistic matching with confidence scoring per session.
One order, one credit. The four platforms above all claim Order #4781 — Omnitrail keeps one.
Every model has a blind spot. Pick one and you commit to its bias. Omnitrail runs all five against the same dedup'd events so you can see where they agree, where they diverge, and which channel just shifted under your feet.
| Model | Google Ads | Meta | TikTok | Direct | True ROAS | Conf. | ||
|---|---|---|---|---|---|---|---|---|
Linear Equal credit to every touchpoint. Honest but flat. | 28% | 22% | 16% | 14% | 12% | 8% | 2.91× | — |
First‑Click All credit to the discovery channel. Rewards top‑of‑funnel. | 42% | 31% | 15% | 5% | 4% | 3% | 2.47× | — |
Last‑Click All credit to checkout. Over‑rewards email and direct. | 18% | 14% | 8% | 38% | 18% | 4% | 2.18× | — |
Data‑Driven Shapley‑based credit from your conversion paths. | 31% | 24% | 18% | 17% | 7% | 3% | 3.08× | 91% |
AI‑Powered★ TRUTH Probabilistic + causal lift. Updated continuously. | 34% | 21% | 23% | 16% | 4% | 2% | 3.24× | 97% |
The corrected attribution signal feeds directly back into your ad accounts. Last month's reallocation, visualized — and what the AI agent recommends for next month.
Per‑campaign corrected ROAS pushed back to Meta, Google, and TikTok ad accounts hourly.
Approve a recommendation and budgets shift across ad accounts in a single move.
Set per‑channel floors, ceilings, and pacing rules. Recommendations stay inside them.
Every reallocation is logged with the model state and the input data that triggered it.
An AI agent that reads Omnitrail's corrected attribution — not the platform pixel — and rebalances inside your guardrails. The first seven actions go through you. After that, it earns trust.
Approve the first seven recommendations to teach the agent your bias bar. Once it has a track record, qualified actions auto‑apply with full audit trail.
First seven recommendations need explicit approval. After that, qualified actions auto‑apply.
Agents act on Omnitrail's deduplicated attribution, not the platform pixel they're acting on.
Per‑channel floors, ceilings, max daily move, and pacing rules. Agents never breach them.
Every auto‑applied action has a one‑click rollback inside Omnitrail.
True ROAS 1.84× · below floor (2.50×) for 6 days running.
Meta CAC up 18% week‑over‑week. TikTok confidence at 94%.
Brand campaign over‑credited; bid ceiling cut from $4.20 to $3.10.
$2,400 redistributed across 6 ad sets. No single move > 8% of budget.
Variant C of Meta DPA campaign · 48h sample · z = −2.3.
Continuous spend pacing · within per‑channel bands · no flag.
Awaiting your review · unlocks $4.8k/wk additional spend.
A continuous forecast of revenue, CAC, and ROAS by channel — with P10 / P50 / P90 bands, not a single‑point guess. Reforecasts nightly. Surfaces when reality drifts outside the 80% interval.
Simulation runs against your corrected attribution, not platform‑claimed numbers.
Confidence intervals, not single‑point estimates. Drift outside 80% triggers a flag.
New events, new model output, new bands. By morning the forecast accounts for yesterday.
When actuals fall outside the 80% interval, you get a flag before it shows up in a monthly review.
GA4, platform pixels, and MMM each solve part of the problem. Omnitrail is the layer above that ties them together — and acts on the corrected number.
★ Omnitrail · the layer above the others. We integrate with all three and reconcile the signals.
A 30‑minute call. We'll plug a read‑only connection into your ad accounts and one storefront, then walk through every capability on this page applied to your last 30 days. You leave with a working forecast and one budget recommendation worth acting on.