Portfolio as Weighted Contribution Vector
A multi-store portfolio is not a collection of independent P&Ls. It is a weighted contribution vector where each store’s true profit is a scalar component. Portfolio decisions — ad budget allocation, SKU prioritization, store investment sequencing — are capital allocation problems, not operational ones.
The portfolio health signal is margin compression at the aggregate level while individual store revenue grows. This occurs when high-growth stores carry structurally lower margins that dilute the portfolio average. Revenue-level reporting masks this entirely.
Portfolio composition metrics:
| Metric | Signal Layer | Portfolio Use |
|---|---|---|
| True Profit per store | P&L | Contribution ranking |
| Margin % per store | Efficiency | Budget reallocation trigger |
| True ROAS per store | Ad efficiency | Marginal profit efficiency |
| Fee % per store | Cost structure | Structural vs operational classification |
| Margin slope (8-week) | Trajectory | Stability classification |
Store Contribution Formula
Store Contribution Weight = Store True Profit / Portfolio Total True Profit
Portfolio Weighted Margin = Sum(Store True Profit * Store Margin%) / Portfolio Total True Profit
Rank stores by absolute contribution weight, not margin %. A store at 42% margin contributing 4% of portfolio profit is less capital-relevant than a store at 22% margin contributing 61% of portfolio profit. Margin % optimizes the wrong objective at the portfolio level.
Margin Trajectory Telemetry
Track weekly margin % per store over a rolling 8-week window. Classification thresholds:
| Class | Definition | Response |
|---|---|---|
| Stable | Week-over-week variance σ < 2pp, no directional trend | No intervention required |
| Declining | Linear slope < −0.5pp per week sustained over 4+ weeks | Root cause investigation required before budget increase |
| Volatile | σ ≥ 2pp with no consistent direction | Likely ramp phase or inconsistent ad cadence — compare against own prior trajectory only |
A Declining classification on any store contributing more than 40% of portfolio profit requires immediate root cause analysis. Do not increase ad spend on a Declining store — additional spend accelerates margin erosion, not recovery.
Root cause taxonomy for Declining classification:
- Fee rate change — TikTok adjusts category referral rates periodically; check settlement fee breakdown for period-over-period delta
- Ad efficiency degradation — ad-to-revenue ratio increasing without corresponding margin improvement
- Refund rate increase — return fees compound against gross margin; check
settlement_transaction_skus.refund_amounttrend - Price compression — ASP declining without corresponding COGS reduction
Fee Structure Variance Decomposition
Not all fee differences between stores are actionable. Decompose into two components:
Total Fee Delta = Structural Fee Delta + Operational Fee Delta
Structural fee delta — driven by category referral rate differences. A beauty store at 5% referral vs an electronics store at 8% referral carries a 3pp structural disadvantage. This is not fixable without changing category.
Operational fee delta — driven by choices:
- Affiliate commission rate (creator-heavy vs direct ad model)
- Fulfillment model (FBT vs self-ship; FBT adds fulfillment fees but reduces return-shipping costs)
- Promotional discount depth (affects net settlement amounts)
Only operational fee delta is an optimization target. Structural delta informs store-level margin ceiling — the maximum achievable margin given category constraints.
Marginal Profit Efficiency: Budget Reallocation Signal
Marginal Profit Efficiency (MPE) = True ROAS * Margin %
MPE is the composite signal for ad budget reallocation. A store with high True ROAS but low margin (high revenue, low profit return) produces less portfolio profit per ad dollar than a store with moderate True ROAS and high margin.
Reallocation decision rule:
| Condition | Action |
|---|---|
| MPE(Store A) > MPE(Store B) by > 15% | Shift incremental budget toward Store A |
| MPE(Store B) is Declining class | Freeze Store B budget; do not reallocate until trajectory reverses |
| Both stores MPE within 10% | No reallocation signal; optimize within each store independently |
Budget reallocation is not a permanent shift — re-evaluate MPE monthly. A store in ramp phase will have suppressed MPE that recovers as it exits Volatile classification.
Concentration Risk Threshold
Concentration Risk Flag = Store Contribution Weight > 0.70
Any single store exceeding 70% contribution weight is a concentration risk. A single bad settlement cycle, TikTok policy change, or product recall against that store will compress total portfolio profit by more than 70%. This is not a soft advisory — it is a structural portfolio risk that requires active mitigation.
Mitigation vectors (ordered by capital efficiency):
- Improve margin on secondary stores — even +3pp margin on a 15% contributing store increases its absolute profit and reduces concentration ratio
- Reduce single-SKU dependency within the dominant store — SKU-level concentration compounds store-level concentration risk
- Cap ad spend on low-MPE stores at a floor threshold until they reach minimum viable margin (typically 15%+) before accepting incremental budget
Monthly Portfolio Reconciliation Protocol
Execute in sequence. Each step gates the next.
- Pull 30-day true profit by store — from settlement data layer, not order data
- Recompute contribution weights — has the ranking changed month-over-month?
- Classify each store’s margin trajectory — Stable / Declining / Volatile per 8-week slope
- Decompose fee variance — identify any new structural or operational fee deltas vs prior month
- Compute MPE per store — flag any store where MPE shifted > 15% vs prior month
- Check concentration risk — any store crossing 70% contribution threshold?
- Set next-month ad budget allocation — based on MPE ranking and trajectory classification
- Document root causes for any Declining stores — without documented root cause, do not allocate incremental budget
This protocol produces a capital allocation decision, not a performance narrative. Output is: budget deltas per store, stores requiring investigation, and concentration risk status.