Path evidence converging on approved ordersRibbon width is normalized credit; opacity reflects evidence confidence
Marketing & Advertising
Shopping Campaign Attribution Calculator
Allocate descriptive shopping-campaign credit across commercial paths while preserving the distinction between observed orders and causal lift. The model removes cancellations, estimates overlap and unresolved orders, weights each path by participation, evidence confidence, and recency, and reconciles all allocated order and revenue credit back to one approved total.
Allocable approved orders—
Approved order total—
Unresolved share—
Shopping-ad credit—
Allocated revenue—
Reconciliation difference—
Commerce evidence braid
Reconcile overlapping shopping paths without presenting descriptive credit as causal lift
| Path | Participation | Confidence | Recency | Normalized credit | Orders | Revenue |
|---|
How to reconcile shopping paths
Begin with approved commerce outcomes
- Choose one order window and export observed orders from the commerce system.
- Remove canceled, rejected, fraudulent, or otherwise non-retained orders.
- Set aside approved orders whose identity or path evidence is insufficient.
- Measure participation for shopping ads, organic listings, brand search, and direct or assisted paths.
- Score evidence confidence and recency independently for every path.
- Normalize the path weights and verify allocated orders return exactly to the allocable pool.
Attribution fundamentals
Observation, allocation, and causality are different claims
Observed orders
All recorded orders before commercial approval.
Approved orders
Outcomes retained after rejection and cancellation.
Unresolved pool
Approved outcomes without sufficient path evidence.
Participation
How often a path appears in qualifying journeys.
Confidence
Strength of identity, tagging, and source evidence.
Recency
Explicit preference for temporally closer touches.
Commerce truth set
Platform conversion columns do not replace backend approval
Use the order-management or payment system to determine retained orders. Platform and analytics records can describe paths, but cancellations and rejections must be removed before allocating commercial credit.
Overlap control
Multiple paths can participate without creating extra orders
Participation percentages can sum above 100% because journeys overlap. The normalization step converts those overlapping signals into shares of one allocable pool, so the calculation never invents outcomes.
Causal boundary
Descriptive credit should not be labeled incremental lift
A path can appear in many successful journeys without causing the order. Use randomized holdouts, geo tests, or other credible designs when the decision requires an incremental claim.
Detailed calculation process
Approve, set aside, weight, normalize, and reconcile
Approved ordersOa = Oobs × (1 − z)
Allocable poolOp = max(Oa − U, 0)
Path weight and creditWᵢ = Pᵢ × Cᵢ × Rᵢ; Sᵢ = Wᵢ ÷ ΣW
Allocated outcomesOᵢ = Op × Sᵢ; Vᵢ = Oᵢ × AOV
- Oobs
- observed orders; orders
- z
- cancellation or rejection rate; decimal
- Oa
- approved orders; orders
- U
- unresolved approved orders; orders
- Pᵢ
- path participation; decimal
- Cᵢ
- evidence confidence; decimal
- Rᵢ
- recency factor; decimal
- Sᵢ
- normalized path share; decimal
- AOV
- average approved order value; currency/order
Default substitution
Approved orders = 1,850 × (1 − 0.09) = 1,683.5. Allocable orders = 1,683.5 − 170 = 1,513.5. Shopping-ad weight = 0.72 × 0.92 × 0.95 = 0.62928. The same operation is repeated for the other three paths before normalization.
Reconciliation: normalized shares total 100%; allocated path orders total 1,513.5; allocated revenue equals 1,513.5 × $112.
Decision use
Spend against evidence quality, not credit alone
High credit with weak confidence is a measurement priority, not a budget mandate. A large unresolved pool means the first investment may be identity and tagging repair rather than more media.
Model limitations
What the allocation leaves unresolved
The model excludes causal lift, cross-device error, privacy loss, view-through disputes, position effects, path interactions, delayed returns, margin variation, and uncertainty. Confidence and recency are entered judgments.
Evidence register
Records that support a defensible allocation
- Backend order, rejection, and cancellation export.
- Analytics path data with consent and identity coverage.
- Platform click and campaign identifiers.
- Documented window, confidence rubric, and unresolved-order policy.
Key terminology
Shopping attribution glossary
- Allocable pool
- Approved orders with sufficient path evidence.
- Unresolved order
- Approved outcome intentionally kept outside allocation.
- Participation
- Presence of a path in qualifying journeys.
- Evidence confidence
- Entered strength of path identification.
- Recency factor
- Entered time-proximity weight.
- Normalized credit
- Path weight divided by total path weights.
- Causal lift
- Outcome difference caused by an intervention.
Practical examples
Shopping Campaign Attribution Calculator in real planning situations
- Reconcile platform-reported shopping orders with analytics paths and backend approvals.
- Reduce brand-search credit when its participation is high but evidence confidence is weak.
- Keep unresolved and direct orders outside the allocable pool instead of forcing every order into paid media.
Important note
Before relying on this result
This model is descriptive. It excludes causal incrementality, identity error, privacy loss, cross-device uncertainty, view-through disputes, delayed returns, margin variation, and sampling uncertainty.
Additional Shopping Campaign Attribution Calculator questions
Why can path participation exceed 100% in total?
One order can contain several touches; normalization converts overlapping participation weights into one allocable credit total.
Is allocated credit incremental lift?
No. It is a descriptive reconciliation of entered evidence, not a causal estimate.
Why remove cancellations first?
Commercial attribution should reconcile to approved orders when canceled or rejected orders do not retain value.
What belongs in unresolved orders?
Orders with insufficient identity, consent, tagging, or path evidence to support allocation.