Methodology
Marketing data is full of gaps, overlaps and estimates. These are the rules the product follows so you always know what a number is, where it came from and how far to trust it.
When a platform does not report a value, or a source is not connected, Deltryx shows "—" (not available). It never fills the gap with 0, because a zero would pull down totals, averages and ratios that are actually unknown.
Store revenue, platform-reported revenue, attributed revenue and incremental revenue answer different questions. They sit side by side with their own labels and are never added together or swapped for one another.
Each session, touchpoint and order is tagged with how its source is known. You can open any attributed order and see the exact evidence behind it.
When live API data exists for a channel and day, aggregate rows imported from spreadsheets for the same channel and day are dropped. The two sources are never summed, so spend and conversions are not counted twice.
Rule-based models (last click, first click, linear, time decay, position based) and a data-driven model run on the same touchpoints, with windows you choose. View-through credit is only added through one capped path: it can never exceed what was left as direct or unattributed, so it cannot inflate total revenue.
The marketing mix model needs at least 56 days of daily data and is checked against days it was not trained on. If any validity check fails, Deltryx shows why and does not hand its numbers to reports or the assistant.
A geo or time holdout test reports a return only when spend actually moved during the test and the measured effect is outside the margin of error. Otherwise the result is reported as inconclusive, with no return figure.
Profit uses the margin measured from your order costs when they are available, and falls back to the margin you entered otherwise. The source of the margin is always shown next to the figure.
Each platform credits itself with its own window and rules, and several platforms often claim the same order. Deltryx keeps platform-reported revenue as reported and compares it to store revenue instead of forcing them to agree.
Yes. Every attributed order has an evidence trail: the sessions, touchpoints, click IDs or UTM tags behind it, and whether each one is verified or inferred.
Its numbers are hidden from reports and from the AI assistant, and the page explains which check failed and what data would fix it.
Start a free trial and connect a store and an ad account, or book a walkthrough with our team.