AI for Ecommerce Unit Economics Decisions

Define the Unit Economics Stack

Unit economics in DTC ecommerce rest on three pillars: contribution margin (CM), merchandise return economics (MER), and the LTV:CAC ratio. Each requires different data sources and decision rules.

Contribution margin is revenue minus cost of goods sold, payment processing fees, and fulfillment cost per unit. For a $60 item with $18 COGS, 2.9% payment fee ($1.74), and $4 fulfillment cost, CM is $36.26 per unit. This baseline shifts with volume discounts, regional shipping costs, and seasonal fulfillment labor.

Merchandise return economics measure the cost of processing a return, restocking, and potential markdown or disposal. A 25% return rate on a $60 item with $8 return processing cost and 15% markdown loss means MER reduces effective CM by $19 per unit sold (0.25 × ($8 + $9 markdown)). High-return categories (apparel, footwear) require separate CM thresholds than low-return categories (supplements, home goods).

LTV:CAC ratio divides customer lifetime value by the fully loaded cost to acquire that customer. If CAC is $35 and average customer LTV is $140 (2 purchases, $70 AOV, 40% repeat rate), the ratio is 4:1. Ratios below 3:1 signal unprofitable acquisition; above 5:1 signal room to increase spend.

Connect the Data Sources

AI cannot propose unit economics decisions without real - time feeds from four systems: the ecommerce platform (transaction data, AOV, repeat purchase windows), the payment processor (fee schedules by region and payment method), the fulfillment provider (cost per unit, regional surcharges, return processing fees), and the ad platform (spend by channel, cohort, and creative).

The ecommerce platform should export daily: units sold, revenue, COGS (by SKU or category), return count and reason, and customer ID with purchase history. Payment processor data must include fee percentage and per - transaction cost by region. Fulfillment data must break out base cost, weight - based surcharges, and return handling cost.

Ad platform data must include spend, impressions, clicks, conversions, and customer ID so that acquisition cost can be attributed to cohorts. Without customer ID linkage, CAC cannot be tied to LTV. This is the most common gap in DTC unit economics stacks.

Ecommerce platform: daily transaction, return, and customer data
Payment processor: fee schedules by region and payment method
Fulfillment provider: unit cost, surcharges, return processing cost
Ad platform: spend, conversions, and customer ID for cohort attribution
Data warehouse or CSV export: unified daily refresh, no manual reconciliation

Automate CM and MER Calculation

Once data is connected, AI should calculate CM and MER daily by product, category, and acquisition channel. The formula is straightforward but the execution requires discipline.

CM per unit = Revenue - COGS - Payment Fee - Fulfillment Cost. MER per unit = CM - (Return Rate × (Return Processing Cost + Markdown Loss)). Net CM = CM - MER.

The decision rule is: if Net CM falls below a threshold (typically 25% of revenue for DTC), flag the product or channel for review. If Net CM is above 45%, signal that pricing or acquisition spend can increase. AI should propose specific actions: raise price by 5%, reduce fulfillment cost by negotiating with the provider, or pause acquisition in a low - LTV:CAC channel.

The human operator reviews the proposal, checks for context (seasonal demand, competitive pressure, inventory position), and approves or rejects. AI does not change prices or pause channels without approval.

Calculate CM daily by product, category, and channel
Deduct MER (return cost + markdown loss) from CM to get Net CM
Flag products or channels where Net CM < 25% of revenue
Propose price increases, cost reductions, or acquisition pauses
Operator approves before any change is live

Monitor and Optimize LTV:CAC Ratios by Cohort

LTV:CAC is the most dynamic metric because both LTV and CAC shift with seasonality, creative performance, and repeat purchase behavior. AI should calculate the ratio weekly by acquisition channel, creative, and customer cohort.

LTV calculation requires a lookback window (typically 90 or 180 days) and a repeat purchase model. For a new customer acquired on Day 1, LTV is the sum of all purchases within the window, minus any returns. If the brand has 12 - month repeat purchase data, AI can project 12 - month LTV from 90 - day cohort data using a decay curve.

CAC is total ad spend divided by conversions in a cohort. If a Facebook campaign spent $5,000 and generated 100 customers, CAC is $50. If those 100 customers generated $15,000 in LTV (repeat purchases included), the ratio is 3:1.

The decision rule: pause or reduce spend in channels or creatives where LTV:CAC < 2.5:1. Increase spend in channels where LTV:CAC > 4:1 and inventory supports it. Test new creatives in low - LTV:CAC channels to improve the ratio before scaling.

Calculate LTV for each acquisition cohort using 90 - or 180 - day lookback
Divide total CAC by LTV to get the ratio
Monitor ratio weekly by channel, creative, and customer segment
Pause or reduce spend where ratio < 2.5:1
Scale spend where ratio > 4:1 and inventory is sufficient

Set Decision Thresholds and Approval Workflows

AI proposals should be tiered by impact. Small changes (price increase under 3%, pause a low - spend creative) can be auto - approved if the brand has set thresholds. Large changes (price increase over 10%, pause a major channel) require operator review.

Thresholds depend on brand risk tolerance and operational capacity. A brand with high inventory risk and low operational bandwidth should require approval for any price change or channel pause. A brand with lean inventory and a dedicated growth team can auto - approve price increases under 5% if Net CM is above 40%.

The approval workflow should include: AI calculates metric, AI proposes action with reasoning, operator reviews in under 24 hours, operator approves or rejects with notes, action executes or is logged for manual follow - up. If an operator rejects a proposal, AI should log the reason and adjust future proposals accordingly.

Define auto - approval thresholds for price, spend, and product changes
Require human review for changes above thresholds
Log all rejections and reasons to improve future proposals
Review thresholds quarterly as business scales

What Stays Human

Unit economics AI is a calculator and advisor, not a decision - maker. The operator owns three decisions that AI cannot make: strategic pricing (brand positioning, competitive response), inventory allocation (which products to stock and in what quantity), and customer acquisition strategy (which channels to invest in and why).

AI can propose a price increase because Net CM is low, but the operator must decide if the brand can absorb a price increase without losing market share. AI can flag a high - return product, but the operator must decide if the product is a loss leader for customer acquisition or a genuine problem.

The operator also owns the data quality check. If COGS data is stale, if fulfillment costs are not updated, or if return reasons are miscoded, the unit economics calculation is wrong. AI should alert the operator to data gaps, but the operator must fix them.

Operator sets strategic pricing and positioning
Operator decides inventory allocation and product mix
Operator chooses acquisition channels and budgets
Operator validates data quality and updates cost assumptions
AI proposes; operator decides

Reporting and Feedback Loop

Unit economics AI should produce a weekly or daily report showing: Net CM by product and category, LTV:CAC by channel and cohort, flagged products or channels, and proposed actions with approval status. The report should be accessible to the operator and the finance team.

The feedback loop is critical. If AI proposes a price increase and the operator approves it, AI should track the impact on conversion rate, return rate, and repeat purchase rate over the next 30 days. If conversion rate drops 15% and LTV:CAC falls below 3:1, AI should flag the change as unsuccessful and propose a rollback.

Over time, this feedback loop trains the AI model to understand the brand's elasticity and customer behavior. The operator learns which proposals are reliable and which require skepticism.

Questions

FAQ

What is the minimum data maturity required to start using AI for unit economics?

The brand must have: (1) daily transaction data from the ecommerce platform with SKU, revenue, and COGS, (2) payment processor fee schedules, (3) fulfillment cost per unit from the provider, and (4) customer ID linkage between the ad platform and the ecommerce platform. Without customer ID linkage, LTV:CAC cannot be calculated. Without COGS data, CM cannot be calculated. Start with these four data sources; add return reason codes and regional cost breakdowns as maturity increases.

How often should unit economics metrics be recalculated?

CM and MER should be calculated daily because they are sensitive to daily sales volume and return rates. LTV:CAC should be calculated weekly because cohort LTV requires a lookback window (90 or 180 days) and weekly updates are sufficient to detect trends. Daily LTV:CAC calculations are noisy and can trigger false alarms. Monthly reviews of thresholds and decision rules are recommended.

What is a healthy LTV:CAC ratio for a DTC ecommerce brand?

A ratio of 3:1 or higher is generally considered healthy; 4:1 or higher signals strong unit economics and room to scale acquisition. Ratios below 2.5:1 indicate unprofitable customer acquisition and warrant investigation. However, the threshold depends on gross margin, repeat purchase rate, and business stage. A brand with 60% gross margin can sustain a 2:1 ratio; a brand with 40% gross margin needs 4:1. Use the ratio as a decision trigger, not an absolute benchmark.

How should AI handle seasonal fluctuations in unit economics?

AI should calculate unit economics separately for seasonal cohorts (e.g., Q4 holiday customers vs. Q1 customers) because LTV and return rates differ by season. A customer acquired in November may have higher LTV due to gift - giving behavior, but also higher return rates due to sizing uncertainty. Use a 12 - month lookback window to smooth seasonal noise, and flag when a metric deviates more than 2 standard deviations from the seasonal average. Operator should review seasonal thresholds quarterly.

Want this on your account?

Thirty minutes. Bring the number that keeps you up.

Book a demo