Connect the accounts you have. Choose one outcome to credit, then read that credit beside what each platform reported. The two are not the same number.
MetaNot connected yet
Google AdsAd spend
HubSpotLeads
StripePaymentsAdvertising, customers, and orders in a shared view. Keep the source and period attached to every result.
Read platform claims alongside your model. See the differences without combining overlapping totals.
One spend source is missing. Recover the records before comparing returns.
Turn the evidence into a useful explanation. Bring assumptions and gaps into the conversation.
Illustrative product concepts and sample data. The custom model is an estimate.
Monday’s budget review. Two platforms claim the same sale. A source is missing. Before the next decision, bring the evidence into the room.
Illustrative evidence view
Platform reports can tell different stories. Compare the same period and keep each attribution perspective distinct.
Inspect the source and matching rule. Separate what was observed from what was calculated or estimated.
A missing source becomes an action to investigate. A model becomes a method to review. The discussion has a shared basis.
A platform claim, a record match, and a model estimate answer different questions. Keep the distinction visible before making the decision.
Platform reportedClaimMeta’s attribution perspective
Source matchedDerivedOrder values matched under a stated rule
A separate modeled perspective
Platform reports use their own attribution rules and windows. Two platforms can claim the same order.
A matched result has a source and a rule. Trace it to the records, and read it within its coverage.
A model can offer another perspective. Its assumptions stay explicit, and its result stays separate from observed records.
See how a rule changes the distribution of credit. Platform-reported values sit alongside your model, with their differences made explicit.
Meta
Reddit
TikTokGoogle Ads
Criteo
Snapchat
DirectMove from a reported result to the recorded touchpoints behind it. Inspect the sequence, identifier, and outcome before interpreting the contribution.
Illustrative Meta click record with a campaign identifier. A recorded touch does not establish causation.
Illustrative journey · unobserved interactions may exist
Advertising recordsAvailableCustomer recordsAvailableOrder recordsAvailableThe missing source is excluded from the known total. It never quietly becomes zero.
Example source states, not verified integration availability
Keep missing sources, incomplete periods, and modeled values visible. Let the evidence set the strength of the conclusion.
Explore the evidence statesAsk for an explanation, examine a discrepancy, or start a review brief. The reasoning belongs next to the answer.
Try the AI examplesWhy are the platform totals different?
Meta reports $56,000. Google Ads reports $52,000. The custom model estimates $54,000. These are separate attribution perspectives; their totals may overlap.
Are the date ranges, attribution windows, and matching rules aligned?
Illustrative response · scripted concept · no live AI connection
Identify the sources, period, currencies, and records available for the question.
Review platform claims alongside a stated model. Keep estimates and matched records distinct.
Carry the sources, method, assumptions, and gaps into the conversation about what comes next.
Platforms can use different windows and credit rules, and may claim overlapping conversions. Comparing their perspectives under a stated scope makes those differences easier to examine.
No. In the example, the $300 order stays the same. The selected rule changes how credit is allocated across recorded touchpoints.
The figures on this page are an example. In the workspace, Google Ads, HubSpot, and Stripe connect, and the model you select splits the outcome.
Credit allocation alone does not establish causation or incremental lift. Those questions require appropriate additional evidence and methods.
A shared evidence layer for the way you work. Explore the AI coding ecosystem, from the terminal to the editor.
Aider
GooseEcosystem preview. Tool logos identify their respective products; weive connections are not demonstrated in this preview.
Define who can access the evidence, what a harness can use, and where human review belongs.
Proposed product controlsChoose the sources and records each harness may access.
Separate access to read, analyze, and prepare actions for review.
Make the source, request, and decision available for inspection.
Set approval points before consequential changes are made.