Every example and amount on this page is illustrative. It shows the method and is not a benchmark, a quote or advice.
What Decision Intelligence Actually Measures
Decision intelligence measures the quality of a decision under uncertainty, not a single dashboard metric. A revenue figure describes what happened. A decision-intelligence view asks whether the figure is reliable enough for the choice in front of you, what assumptions shape it, which alternatives exist, and how the result will be checked afterward.
The process connects four parts. Data establishes the starting position: delivered revenue, product cost, advertising spend, returns, inventory and customer behaviour. Assumptions make definitions explicit, such as whether revenue excludes cancelled and returned orders or whether ad performance uses attributed ROAS or total-store MER. Alternatives define the available actions, such as scale, reduce, fix economics, test or stop. Outcomes then show whether the decision produced the expected commercial result, without treating a forecast as a promise.
Different measures answer different questions. Revenue can support a demand or cash-planning decision, but it cannot establish profitability. Gross margin tests product economics before advertising; contribution after advertising is more relevant when deciding whether acquisition spend is sustainable. MER describes total-store efficiency, while attributed ROAS describes the performance assigned by an advertising platform. Neither alone proves that an incremental budget increase will be profitable.
This is why a number can be mathematically correct and still be unsuitable for a decision. A collection may have strong ROAS but lose money after product cost and returns. A profitable product may have limited remaining stock. A forecast may be useful for capacity planning while remaining inappropriate as evidence that a proposed action will cause a particular result. ORVX’s ecommerce financial analytics connects these measures while keeping the merchant’s trading currency consistent.
| Measure | What it answers | When it is relevant |
|---|---|---|
| Revenue | How much was realised from delivered orders | Demand, cash and sales reporting |
| Gross profit or margin | Whether sales cover product cost | Product pricing and assortment decisions |
| Contribution after advertising | What remains after product cost and ad spend | Acquisition and product-level budget decisions |
| MER | How total advertising spend relates to total store revenue | Blended channel efficiency |
| Attributed ROAS | How an ad platform assigns revenue to spend | Campaign and platform reporting |
Match the Measure to the Decision
A measure is useful only when it matches the decision’s time horizon and downside. Revenue and conversion rate describe demand; ROAS describes paid-media efficiency; contribution margin tests whether an order creates economic value; customer lifetime value extends that view across repeat purchases; expected value compares uncertain choices by weighting possible outcomes.
| Measure | What it answers | When to act on it |
|---|---|---|
| Revenue | How much value was realized from delivered orders. For illustration only, 100 orders at $50 produces $5,000; cancelled and returned orders should not remain in realized revenue. | Use it for sales pacing, cash planning and period comparisons. Do not use it alone to increase spend: $5,000 in sales can still lose money. |
| Conversion rate | What share of visits became orders: 50 orders ÷ 2,000 sessions = 2.5%. | Use it to investigate traffic quality, merchandising and checkout friction. Act when the denominator and tracking are consistent; a rate change without comparable traffic can mislead. |
| ROAS | How much attributed revenue came from advertising: $4,000 ÷ $1,000 ad spend = 4.0x. | Use it to compare paid campaigns within the same attribution view. Act on it only with margin context, because attributed revenue is not profit and attribution does not equal incremental demand. |
| Contribution margin | What remains after variable product, fulfilment and advertising costs. A $100 order less $35 product cost, $10 fulfilment and $30 advertising leaves $25, or 25%. | Use it for scaling, reducing or stopping activity. This is the right measure when the decision is whether incremental orders improve the business. |
| Customer lifetime value | The expected or observed economic value of a customer across purchases, after relevant costs. A $60 first order followed by $40 of contribution from repeat orders gives $100 cumulative contribution. | Use it for acquisition payback, retention and cohort decisions. It requires a defined horizon and should not be treated as certain value for a new customer. |
| Expected value | The probability-weighted value of uncertain outcomes. A test with a 60% chance of $1,000 gain and a 40% chance of a $500 loss has expected value of $400: (0.6 × $1,000) − (0.4 × $500). | Use it when choosing among actions with different probabilities, payoffs and downside, rather than reacting to one observed result. |
The Formula in Plain English
A practical expected decision value formula is: **expected decision value = probability of the outcome ×
A Worked Example: Should the Merchant Increase Spend?
Assume a merchant currently spends $20,000 per month on paid media and receives 1,000 delivered orders. Each order produces $80 of realized revenue and $32 of product cost, leaving a $48 contribution before advertising. Holding spend steady therefore gives $48,000 before the current ad cost, or $28,000 after it.
The proposed change is an additional $8,000 in spend. The merchant’s own evidence suggests two plausible outcomes: a 60% chance of 500 incremental delivered orders, or a 40% chance of only 100. The figures below treat $48 as the contribution from each incremental order and measure value against the hold-steady option.
The increase has positive expected value, but the downside is visible: in the weaker outcome, the extra orders generate $4,800 of contribution against $8,000 of spend, producing a $3,200 loss. The arithmetic is 0.60 × ($24,000 − $8,000) + 0.40 × ($4,800 − $8,000) = $8,320. That is the expected incremental value, not a promise that the campaign will generate it. If the merchant cannot defend the probabilities or margin assumption from its own numbers, the answer should be treated as uncertain rather than precise.
| Decision | Delivered orders | Contribution margin | Uncertainty | Expected incremental value |
|---|---|---|---|---|
| Hold spend steady | 1,000 | $48 per order; $28,000 after the existing $20,000 ad spend | Reference case; no incremental demand assumed | $0 |
| Increase spend by $8,000 | 1,340 expected total orders | $48 per incremental order; $16,320 expected incremental contribution before added ad spend | 60% chance of +500 orders; 40% chance of +100 | +$8,320 |
The Mistakes That Make the Number Wrong
Attribution is not incrementality: platform credit is not causation. Use attributed ROAS for reported channel credit and MER for total advertising spend versus total store revenue; neither is causal evidence.
Blended averages hide economics. A $40 average order contribution can combine 100 orders at $70 with 100 orders at $10. Scaling the average also scales weak orders. Use product-, collection-, cohort- and channel-level cuts; a financial analytics view should preserve them.
Treat a $5,000 development bill already paid as sunk, but include costs that change. If a $60 sale has $24 of product cost, $8 of advertising and $6 of other incremental costs, its contribution is $22, not $52.
Set baselines with their conditions, accounting for seasonality, stock, price and customer mix. Reconcile revenue at order level: adding Shopify revenue to platform-reported conversion value can double-count orders; exclude returned and cancelled orders. ORVX calculates realized revenue from delivered orders and reports refund value and return rate against revenue.
Match revenue, advertising spend, product cost and refunds to the decision window, then revisit it as the window closes. Show probability assumptions, range and baseline; do not give uncertain outcomes false precision.
| Measure | Useful for | Not sufficient for |
|---|---|---|
| Attributed ROAS | Comparing reported credit across campaigns | Proving that added spend caused added demand |
| MER | Assessing total advertising efficiency at store level | Explaining which product or campaign created the result |
| Realized contribution | Deciding whether additional delivered orders are economically worthwhile | Forecasting demand without a defensible baseline |
| Product or collection profitability | Choosing what to promote, reduce or investigate | Replacing a time-aligned store-level cash or revenue view |
What to Do With the Answer
Define the threshold that would change the decision: for paid media, expected incremental contribution above $0 or a required return compensating for cash and inventory risk; for a discount, added delivered orders covering the margin given up. Set it before the result, or a forecast may move the goalposts.
Test assumptions that could change it. An added $6,000 campaign expected to produce 250 delivered orders at $80 realised revenue and $32 product cost gives $48 contribution before advertising: 250 × $48 − $6,000 = $6,000. Break-even is $6,000 ÷ $48 = 125 delivered orders; if returns reduce contribution to $36, it is 167. That range may change scale to test or stop.
Vary downside assumptions: delivered orders, product cost, refund value, advertising cost and cash-collection timing. Ask which crosses the threshold. ROAS measures media efficiency; contribution after advertising tests profit where margin and returns matter. Revenue checks demand, not profitability.
Record date, action, baseline, threshold, assumptions, expected range, owner and review date. State scale, reduce, fix economics, test or stop, and the evidence that would reverse it. ORVX can surface a prioritised recommendation from store numbers, but the operator remains responsible; it does not place bids, change prices, reorder stock or fulfil orders.
At review, compare the same measures and definitions. Reconcile delivered orders, refunds, product cost, advertising spend and contribution; do not compare a realised-revenue forecast with actual gross bookings. Separate forecast error from execution variance: demand may differ, or spending may not follow plan. Check a coming-period revenue forecast against actual revenue and backtested error, not causation. Update the assumption, threshold or experiment design after documenting what changed.
| Decision | Primary check | Action threshold |
|---|---|---|
| Increase advertising | Incremental contribution after advertising | Expected contribution exceeds the required hurdle |
| Run a discount | Realized contribution after product cost and refunds | Added delivered orders cover the margin given up |
| Reduce or stop a product | Product-level profitability and inventory position | Loss or dead stock exceeds the cost of continuing |
Decision intelligence turns a metric into a decision by making the baseline, economics, uncertainty, and action explicit. The useful answer is not the most impressive number; it is the number that shows whether a specific change is expected to create enough incremental contribution to justify its cost and risk.