An actuary's job is to stare at a population of uncertain outcomes until a number falls out. The number has to be right — not approximately right, not directionally right, but right enough that the company pricing on it doesn't go broke paying claims. That's a different kind of pressure than "we think our margins are around 35%."
I spent years doing actuarial work before moving into e-commerce data. The shift felt enormous at first. Insurance is centuries old. E-commerce is a teenager. But the deeper I got into e-commerce pricing, the more I kept seeing the same fundamental error: businesses pricing products based on a cost-plus logic that looks right in a spreadsheet and quietly destroys margin at scale.
The actuarial framework solves this. It doesn't require an insurance background to apply. It requires treating your customer base as a portfolio of risks, and pricing accordingly. Here's how it works.
PART 01What actuaries actually do¶
When an insurer prices a policy, they don't ask "what does it cost us to administer this policy?" They ask "what is the expected loss we'll pay out on this risk, and what premium covers that loss plus our operating costs, plus a margin for profit?"
The loss ratio is the key metric: claims paid divided by premiums earned. A loss ratio of 70% means for every £100 in premium, £70 went out in claims. The remaining £30 covers expenses and profit. The insurer's job is to set premiums so that the loss ratio stays within a target range — not too high (you go broke paying claims), not too low (you're overcharging and losing customers to competitors).
The important nuance: they don't price individual policies in isolation. They price risk cohorts. A 35-year-old non-smoker in a low-crime postcode who has never claimed is a different risk from a 35-year-old non-smoker in the same postcode who made two claims last year. Same demographic. Completely different expected loss. The pricing reflects the cohort behaviour, not just the visible characteristics.
Most e-commerce founders know their blended margin. Almost none know their margin by customer acquisition cohort. That gap is where the money hides.
PART 02The margin percentage problem¶
Standard e-commerce margin calculation: take revenue, subtract COGS, divide by revenue. You get a percentage. Call it 40%. Looks fine on the dashboard. What it doesn't capture:
- Returns and refunds for this product or category, as a proportion of sales
- The acquisition cost of the customer who bought it
- Whether that customer bought again, and at what margin on the subsequent purchase
- Chargeback rates by product type or acquisition channel
- The actual shipping cost versus the quoted shipping cost for this specific product weight and destination
When you factor all of these in, your 40% product margin often looks more like 18% at the customer relationship level — and sometimes negative, for customers acquired through expensive paid channels who buy once and never return.
This is the same problem actuaries solved decades ago in insurance. The product margin equivalent (the premium) looks fine until you look at the actual claims coming out the other side. In e-commerce, the "claims" are your returns, refund requests, chargebacks, and the cost of serving a customer who was never going to be profitable.
PART 03Cohort-based pricing¶
Here's where the actuarial model diverges from standard e-commerce thinking. Instead of pricing a product, you're pricing the expected value of a customer relationship that begins with the purchase of that product.
Take a nutrition subscription brand. Their core product is a £29.99 monthly supplement box. Their cost-plus margin on the box is 42%. Looks healthy. But when you segment the customer base by acquisition channel:
| Acquisition channel | CAC | Avg. months retained | Return rate | Relationship margin |
|---|---|---|---|---|
| Organic search | £8 | 9.2 | 4% | 38% |
| Email referral | £12 | 11.4 | 3% | 41% |
| Meta paid | £47 | 3.1 | 12% | -6% |
| Google paid | £31 | 4.8 | 9% | 7% |
| Influencer | £22 | 2.4 | 18% | -14% |
The product is the same. The price is the same. The effective relationship margin is wildly different depending on how the customer arrived. Meta and influencer channels are actively destroying margin at scale, while organic and email are the profitable engine — and most brands know this vaguely, but don't have the numbers to act on it.
The actuarial approach forces the calculation. You segment your customer base into acquisition cohorts, calculate the expected lifetime value and expected cost (including returns, support, and chargeback rates) for each cohort, and derive a cohort-level loss ratio. Any cohort with a loss ratio above your target gets a different pricing or acquisition strategy — not the same price applied uniformly.
PART 04Building your loss ratio model¶
You don't need an actuary or specialist software to build this. You need clean transaction data, attribution data, and a willingness to do the uncomfortable calculation. Here's the structure:
- Pull all orders for the past 12–24 months, with acquisition channel, first-purchase date, all subsequent purchase dates, and return/refund flags. If you're on Shopify, this is an orders export plus a returns export, joined on customer ID.
- Calculate earned revenue per customer: total revenue minus returns and refunds, minus chargebacks. This is your equivalent of "premiums earned net of claims."
- Calculate total cost per customer relationship: COGS across all orders, plus your CAC for first acquisition (from your attribution model — warehouse-verified, not platform-reported), plus an allocated share of ongoing retention spend.
- Derive relationship margin by cohort: (earned revenue − total cost) / earned revenue, segmented by acquisition channel, first-purchase product category, and customer tenure band. This is your loss ratio equivalent.
- Identify your floor cohorts: any cohort with relationship margin below your target (typically 15–20% after all costs for a scaling e-commerce brand) is a pricing or acquisition problem, not a product problem.
One brand I audited discovered their most profitable customer segment — organic search, first purchase via a specific blog post — had a 44% relationship margin and an average retention of 14 months. Their least profitable segment — Meta retargeting, first purchase via a discount code — had a -22% relationship margin. They were actively advertising their way to a smaller business.
PART 05The CAC-adjusted pricing floor¶
Standard pricing floor logic: price must cover COGS plus a minimum margin. The actuarial floor is different: price must cover COGS plus amortised acquisition cost for the expected customer lifetime of this cohort, plus operating overhead, plus your target profit margin.
The formula looks like this:
Where expected_orders is the average number of orders for customers in this acquisition cohort, and expected_return_rate is the historical return rate for this product × this cohort type.
This formula produces a different floor price for the same product depending on how the customer was acquired. A customer acquired via organic search, who you expect to order 8 times, has a much lower pricing floor than a customer acquired via a paid campaign at £45 CAC who you expect to order twice. For the paid customer, either the price needs to be higher or the acquisition cost needs to come down. Usually both.
In practice, you can't charge different prices to different acquisition cohorts for the same product. What you can do is:
- Stop running acquisition channels where the CAC makes the relationship economics unviable
- Use the cohort floor to inform which channels get more investment versus which get cut
- Use discount and promotion strategy deliberately — discounts only for cohorts where the lifetime value justifies the short-term margin hit
- Re-price products in categories where return rates are consistently eroding relationship margin below target
PART 06Quarterly re-pricing in practice¶
Actuaries don't set rates once and forget them. They review the loss experience every quarter. Claims experience changes. The risk profile of the portfolio shifts. Economic conditions affect claim frequency. The rates update to reflect reality.
E-commerce brands should run the same process. Quarterly, at minimum, update your cohort-level loss ratio analysis. Ask three questions:
- Which product categories have seen return rate deterioration in the past quarter, and does that shift any pricing floors?
- Have any acquisition channels become significantly cheaper or more expensive, changing the CAC-adjusted floor for products in their typical first-purchase mix?
- Have any cohorts shifted in lifetime retention patterns, changing the amortisation denominator in the floor calculation?
This quarterly cadence turns pricing from a gut-feel annual exercise into a data-driven operational process. It's the difference between setting sail once and adjusting course continuously based on what the instruments actually say.
Where to start if your data is messy
The actuarial model requires clean attribution data. If your Shopify numbers don't reconcile with your ad platform numbers, if you don't have a reliable CAC by channel, or if your return data lives in a separate system that's never been joined to your order data — start there. The model is only as good as the data feeding it.
As I cover in 5 Revenue Leaks in Your E-commerce Data, the attribution layer is usually the first thing that needs fixing in a mid-market e-commerce business. Once that's clean, the cohort analysis is straightforward SQL on a single joined table. The calculation itself is not the hard part. Getting the data to the point where the calculation is trustworthy usually takes three to four weeks.
If you want to run a cohort-level loss ratio analysis on your own catalogue, the Data Audit I offer starts with exactly this — mapping where your relationship economics are working and where they're quietly destroying margin. Book a 20-minute call and I'll show you what the numbers look like before we scope anything.
Sources. Author's field notes from e-commerce data audits, 2023–2025; cohort retention benchmarks from Klaviyo and Recharge subscription data reports; CAC benchmarking from Triple Whale e-commerce benchmarks report 2024.