Customer Lifetime Value for Ecommerce: Start with Realized Cohorts Before Forecasts
Customer lifetime value in ecommerce is most useful when it helps you make a unit economics decision. Start with realized LTV from closed cohorts, then use trailing LTV for cohorts still active, and only then layer in predicted LTV. That order helps founders and retention teams judge acquisition, pricing, and retention using actual customer behavior, not a forecast disguised as fact.
If you are deciding whether to scale a channel, adjust an offer, or invest in retention, begin with cohort revenue, contribution profit, repeat rate, and time horizon. Those four inputs tell you whether growth is improving profitability or just increasing order volume.
Define LTV in a way that supports decisions
In ecommerce, customer lifetime value (LTV) should mean the economic value a customer generates over a defined period. The definition changes depending on what you include, so write it down before you compare channels or cohorts.
A practical unit economics definition is:
LTV = cumulative revenue or contribution profit from a customer cohort over a specified time horizon
For most operating decisions, contribution profit is more useful than revenue because it accounts for variable costs. A working formula is:
Contribution profit LTV = net revenue − cost of goods sold − fulfilment and payment costs − variable service or incentive costs
That is not the same as gross margin, and it is not the same as attributed revenue from an ad platform. Attribution assigns credit to a touchpoint; it does not tell you whether the customer was incremental or profitable.
Shopify’s ecommerce KPI guidance stresses choosing metrics based on the decision you need to make, rather than tracking everything at once, and its analytics guidance recommends combining quantitative data with customer feedback (Shopify KPIs, Shopify ecommerce analytics tools).
Build cohort views before you forecast
The cleanest way to measure LTV is by cohort. A cohort is a group of customers who share a starting event, usually first purchase month, first order date, or first subscription start date.
A cohort view helps you answer practical questions:
- How much revenue did customers acquired in a given period generate over time?
- What repeat rate did they show?
- How quickly did they become profitable?
- Which acquisition source produced better customers, not just more orders?
The minimum fields to track in each cohort
Track these fields for every cohort:
- Cohort size: number of first-time customers in the cohort
- Cohort revenue: revenue generated by that cohort over time
- Margin or contribution profit: revenue minus variable costs
- Repeat rate: share of customers who made a second purchase, third purchase, and so on
- Time horizon: 30, 60, 90, 180, 365 days, or another period relevant to your buying cycle
- Forecast uncertainty: the degree of confidence in any projection beyond realized data
Shopify’s ecommerce metrics guidance recommends connecting funnel, customer, inventory, and marketing data on daily, weekly, and monthly cadences (Shopify ecommerce metrics). That cadence matters: daily for monitoring, weekly for optimization, monthly for cohort and profitability review.
Use realized, trailing, and predicted LTV separately
A three-part model keeps the analysis honest:
| LTV type | What it means | When to use it | Main risk |
|---|---|---|---|
| Realized LTV | Actual revenue or contribution profit already earned from a closed cohort | Historical evaluation, board reporting, model validation | None, as long as the data is complete |
| Trailing LTV | Observed value so far from an open cohort that is still purchasing | Active cohort management, retention tracking | Mistaking partial data for full lifetime value |
| Predicted LTV | Model-based estimate of future value beyond what has been observed | Budgeting, acquisition planning, scenario analysis | Overconfidence, especially with short histories |
This separation matters because a 90-day cohort can look weak or strong depending on the category, reorder cycle, and subscription behavior. Forecasts are useful, but they should not overwrite observed data.
Use contribution profit, not just revenue
Revenue LTV can overstate customer quality if the customers are expensive to fulfill or heavily discounted. For ecommerce, contribution profit usually gives the better answer.
A simple contribution profit LTV formula
For each customer or cohort:
Contribution profit LTV = total revenue
minus COGS
minus fulfilment and shipping subsidies
minus payment processing fees
minus discounts and credits
minus variable support or retention costs
Label every assumption clearly. For example:
- COGS based on standard landed cost
- Fulfilment based on average pick-pack-ship cost per order
- Discounts based on realized net order value
- Support costs included only if variable and material
This is where measurement discipline matters. If your team needs consistent definitions, a Unit Economics measurement guide should define which costs are included and who owns each data source. For profit-focused decisions, see also Ecommerce profit analytics.
Decision rule: ask what changes if the cohort is weaker
The point of contribution profit LTV is not to create a nicer dashboard. It is to change a decision.
Ask:
- Would we still buy this traffic if first-order margin is low but repeat behavior is strong?
- Should we reduce discounting if customers are profitable without it?
- Are we acquiring customers who repurchase, or customers who churn after the first order?
If the answer changes when you include variable costs, contribution profit is the right lens.
Compare acquisition sources by customer quality, not just attribution
Attribution tells you which channel or campaign gets credit. It does not prove which source created the customer, or whether that customer was incrementally acquired.
That distinction matters. Market positioning across analytics tools shows different approaches: attribution, BI, creative analysis, MMM, incrementality, and financial command-center use cases (Triple Whale vs Northbeam, ThoughtMetric attribution tools, Nummbas comparison, ShelfMerge comparison). For LTV analysis, the better question is usually about customer quality by source, not channel credit alone.
What to compare by acquisition source
Compare each source on:
- First-order contribution profit
- 60-day and 180-day realized or trailing LTV
- Repeat rate
- Average order value over time
- Refund or cancellation rate
- Discount dependency
- Payback period, if you can measure it reliably
How to interpret differences between sources
If two sources show similar attributed revenue but different realized LTV, do not assume one source is better until you check:
- Whether tracking is implemented correctly
- Whether the cohorts are comparable by customer profile
- Whether one source is acting more like retargeting than acquisition
- Whether the difference is incremental or just attributable
Google Analytics’ ecommerce documentation is explicit that ecommerce reporting depends on correctly implemented ecommerce events and required parameters (GA4 ecommerce purchases). Instrumentation quality comes first; interpretation comes second.
A worked example: cohort revenue and profit over 180 days
Suppose a January acquisition cohort has 1,000 first-time customers.
Assumptions:
- Average first-order net revenue: $60
- Average repeat revenue over 180 days: $45 per customer
- Net revenue therefore totals: $105 per customer
- COGS plus variable fulfilment plus fees average 65% of net revenue
- Contribution profit margin: 35%
Realized 180-day result
- Revenue LTV = $105 per customer
- Contribution profit LTV = $105 × 35% = $36.75 per customer
- Cohort contribution profit = 1,000 × $36.75 = $36,750
Now compare that with acquisition cost.
If the cohort cost $28 per customer to acquire:
- Customer acquisition cost = $28
- Contribution profit before fixed overhead = $36.75 − $28 = $8.75 per customer
That is a very different decision from looking at revenue alone.
Why the same cohort can be misread
If you only observe 30 days of data, the same cohort may appear to have weak LTV because repeat orders have not had time to emerge. If a retargeting-heavy source captures customers who would have bought anyway, attributed revenue may look strong while incrementality is weak.
That is why realized cohorts should anchor the analysis. Forecasts should extend them, not replace them.
Common LTV mistakes that distort ecommerce decisions
Here are the errors that most often lead teams astray.
1) Mixing attribution with LTV
Attribution answers: “Which touchpoint gets credit?”
LTV answers: “What is the economic value of the customer over time?”
Those are related, but they are not interchangeable.
2) Using revenue instead of contribution profit for scaling decisions
Revenue can hide high fulfilment costs, heavy discounting, or expensive customer service. A source with lower revenue can still produce better profit.
3) Forecasting too early
If your cohort has only a few purchase cycles, predicted LTV will carry high uncertainty. State the horizon explicitly and avoid presenting the forecast as certainty.
4) Comparing cohorts with different time windows
A 30-day cohort and a 365-day cohort are not directly comparable. Use the same horizon whenever possible.
5) Ignoring instrumentation quality
If ecommerce events, parameters, or order values are misconfigured, the cohort view will be wrong before analysis even starts. Validate your tracking setup first in line with GA4 documentation (GA4 ecommerce purchases).
A practical decision framework for founders and retention teams
Use this sequence when reviewing LTV:
-
Check data integrity
- Are ecommerce events implemented correctly?
- Are order values, refunds, and currency handling consistent? -
Review realized cohorts
- Which cohorts have enough history to be trusted?
- What are their realized revenue and contribution profit? -
Inspect repeat behavior
- What is the repeat rate by cohort?
- How fast does repeat purchasing happen? -
Compare acquisition sources
- Which sources create higher-quality customers?
- Are differences likely to be attributable or incremental? -
Forecast cautiously
- Extend from observed cohort curves.
- Show assumptions and uncertainty.
- Separate predicted from realized value. -
Change one decision
- Shift budget, pricing, offer structure, or retention investment based on the result.
What to measure this week
If you only build one view, make it this:
- Cohort start date
- Cohort size
- Revenue by age of cohort
- Contribution profit by age of cohort
- Repeat rate by order number
- Acquisition source
- Forecasted LTV, shown separately from realized and trailing values
That single table can support smarter acquisition, retention, and margin decisions without overstating confidence.
For broader operating context, connect this view to your UE-01 framework and the wider Unit Economics hub.
Create your LTV cohort view
If you are building customer lifetime value ecommerce reporting for the first time, start with a realized cohort table before you add predictions. Then layer in contribution profit and acquisition source comparisons so the view answers a real business question: which customers are worth acquiring, retaining, and scaling?
Create an LTV cohort view with realized, trailing, and predicted LTV separated clearly, and use it to decide where your next growth dollar should go.