LTV by Cohort: The Math Most DTC Dashboards Get Wrong

LTV by cohort is the only honest way to measure DTC customer lifetime value, and most of the dashboards we audit get it wrong in one of three specific ways. Andrew Chen's framing of cohort retention curves that flatten as a product/market-fit signal is the conceptual foundation the rest of this article builds on (via Andrew Chen). The wrong version makes a brand look healthier than it is and feeds bad scaling decisions on paid spend. The right version takes more effort to produce, surfaces uncomfortable truths, and pays for itself the first time a founder catches a deteriorating retention trend before it shows up in revenue.

This is the cohort retention math we use on DTC reporting engagements, written for an operator who wants to understand it well enough to interrogate their own dashboard, not for a data scientist.

Why blended LTV is the wrong starting point

The single most common founder ask is "what's my LTV." The single most common dashboard answer is "total revenue from all customers divided by total customers, blended." That number is mathematically valid and operationally useless.

Blended LTV mixes a customer who first bought yesterday with a customer who first bought 18 months ago. Yesterday's customer has had one day to repurchase. The 18-month customer has had 18 months. Lumping them produces an "average LTV" that doesn't describe either of them. Worse: when the brand is growing — adding more new customers each month — blended LTV tilts toward the new customers and looks like a decline, even if cohort retention is healthy or improving.

LTV by cohort solves this by holding time-since-acquisition constant. You compare January 2025's first-time buyers at 90 days to February 2025's first-time buyers at 90 days to March 2025's first-time buyers at 90 days. Apples to apples. The trend you see is real.

How to actually build the cohort table

The minimum viable cohort LTV table:

• Rows: acquisition month (Jan 2025, Feb 2025, Mar 2025, ...)
• Columns: cumulative revenue per customer at 30 days, 60 days, 90 days, 180 days, 365 days
• Values: cumulative revenue per customer in that cohort at that time point

So the cell at row "March 2025" and column "90 days" is: total revenue from customers whose first purchase was in March 2025, in their first 90 days post-acquisition, divided by the number of customers in that cohort.

Hypothetical table structure (cells shown as $X placeholders — this is a layout reference, not data from any specific brand):

The cells past the time horizon (Apr 2025 at 365 days, e.g., in March 2026) are marked "incomplete" because the data doesn't exist yet. Most dashboards extrapolate. The right move is to mark them incomplete and use the partial data only when it's complete.

The three common mistakes

Mistake 1: Reporting 365-day LTV based on extrapolation

The 365-day cohort revenue number is what most founders want to see and what most dashboards estimate by projecting forward from 90- or 180-day data. The projection is usually wrong.

Real cohort decay curves are not linear. They follow a logarithmic-ish shape: lots of revenue in the first 30 days, decreasing rates of additional revenue at 60, 90, 180, then a slower plateau toward 365 — the "flattening" Andrew Chen flags as the actual sign of stickiness. Projecting a 90-day cohort linearly to 365 days typically overstates LTV by a meaningful margin. DTC brands that make paid-spend decisions based on projected 365-day LTV often find the actual number arriving a year later well below the projection.

The right move: use only completed cohort data. If you started measuring in January 2025, the first 365-day complete cohort is January 2025 measured at January 2026. Plan the dashboard so the 365-day column is honestly incomplete for the most recent 11 months, and use the 90-day or 180-day column for newer cohorts.

Mistake 2: Using cumulative revenue instead of unit revenue

The cumulative revenue number per customer mixes two things: how often customers buy and how much they spend each time. A cohort with $94 cumulative revenue at 365 days could be one customer buying $94 once, or four customers averaging $23.50 per order.

Decomposing the cumulative number tells you what's actually happening:

• Repeat rate at 30/60/90/180/365 days (percent of cohort who made a second purchase)
• Average order value by order number (1st order, 2nd order, 3rd order...)
• Inter-order interval (how many days between order 1 and order 2 for the average repeat buyer)

A cohort with 30% repeat rate and $80 AOV looks very different from a cohort with 60% repeat rate and $40 AOV, even if both produce the same cumulative LTV. The first depends on AOV-driven retention (often big-ticket items). The second depends on frequency (often consumables). Operator-reported DTC repeat-purchase benchmarks tend to sit around 25-30% blended, with consumable categories (supplements, coffee, replenishment skincare) running materially higher — use the category-appropriate band, not the blended average (via bsandco.us). Paid spend strategy differs for both shapes.

Mistake 3: Ignoring the decay curve shape

The decay curve is the rate at which a cohort adds revenue over time. Two cohorts with identical 365-day LTV can have completely different decay shapes:

• Front-loaded: $52 by day 30, $94 by day 365. Customers bought heavily early and tapered.
• Back-loaded: $25 by day 30, $94 by day 365. Customers ramped up over time.

The front-loaded curve is healthier for short-payback DTC economics. The back-loaded curve is harder to model paid spend against because the revenue arrives long after the acquisition cost.

Most dashboards report only the cumulative endpoint. The decay shape lives in the overlay chart that most agencies skip.

The 30/60/90/180/365 minimum viable dashboard

For most DTC operators, the viable cohort LTV dashboard has four panels:

1. Cohort revenue table. Rows = acquisition month for last 18 months. Columns = 30/60/90/180/365 days. Values = cumulative revenue per customer. Cells marked "incomplete" when the time horizon hasn't passed.
2. Repeat-rate cohort table. Same row/column structure. Values = percent of cohort who made a 2nd purchase by that time point. This panel is where retention quality shows up most clearly.
3. Decay curve overlay. Line chart with one line per cohort month, X-axis is days since acquisition (0 to 365), Y-axis is cumulative revenue per customer. Lines bunch up if cohorts are stable; lines fan out if retention is changing.
4. AOV by order number. Bar chart showing avg order value for 1st order, 2nd order, 3rd order, 4th+ order. Tells you whether customers spend more or less as they repeat.

A cohort LTV dashboard worth building includes these four panels. They're the minimum honest view. Anything less hides one of the three mistake patterns above. The dashboard work itself can live in spreadsheets, BigQuery + Looker Studio, or a dedicated reporting tool like our Level cohort views — what matters is the math being correct.

What good looks like 12 months in

The value of cohort LTV done correctly shows up the first time a retention deterioration surfaces in the cohort table before it shows up in blended numbers. A common pattern: per-customer repeat rate drops quarter-over-quarter in a way blended LTV hides because new customer adds are still growing. The operator team catches it from the cohort view, investigates, and finds an upstream cause (product reformulation, fulfillment quality issue, post-purchase email regression) months before it would have hit blended revenue.

That's the value of cohort LTV done correctly. Not a vanity metric. A leading indicator that beats blended numbers by months.

Glossary: the cohort vocabulary that matters

Operators frequently confuse adjacent terms. Quick definitions:

• Cohort — a group of customers grouped by a shared characteristic, usually acquisition month or acquisition week.
• LTV (Lifetime Value) — total revenue from a customer over their full lifetime as a buyer. For an 18-month-old DTC brand, "lifetime" is at most 18 months. Use the time-bounded version (90-day, 365-day) until the brand is old enough for true lifetime.
• Repeat rate — percentage of a cohort that made a second purchase by a given time point. A cohort with 30% repeat rate at 90 days means 30 of every 100 first-time buyers came back inside 90 days.
• AOV (Average Order Value) — total revenue divided by total orders. Important: AOV is different at 1st order vs 2nd order vs 3rd, and the trend matters more than the headline number.
• Inter-order interval — average days between two consecutive orders for a repeating customer. For consumables, 30-60 days is healthy; for high-AOV considered purchases, 120+ days is normal.
• Decay curve — the shape of how a cohort's cumulative revenue grows over time. Front-loaded (most revenue early) vs back-loaded (revenue ramps over time) matter for paid-spend payback timing.
• Payback period — how long it takes for a cohort's cumulative contribution margin to recover the CAC. A cohort with $40 CAC and front-loaded $60 contribution margin at 60 days has a 60-day payback.

The cohort decisions you should make (and the ones you shouldn't)

Cohort dashboards generate a temptation that costs more than they earn: founders see month-over-month cohort variance and start making operational decisions on every wobble in the data. Most of those decisions are noise-chasing. The honest framing is that cohort data is high-value for a specific set of decisions and actively misleading for another set. Knowing the difference is what separates a useful cohort dashboard from an expensive distraction.

Decisions cohort data should drive:

1. CAC ceiling per cohort segment — the 90-day cumulative contribution margin minus COGS gives you the maximum CAC you can pay for that cohort segment and still break even at 90 days. This is a policy decision (set quarterly), not a tactical one (set weekly). Use the trailing 6 cohorts averaged, not the most recent one.
2. Channel-mix shifts when a channel's cohort quality degrades — if Meta-acquired cohorts from the last 3 months show materially worse 90-day repeat rates than Google-acquired cohorts at the same CAC, the spend mix should shift. This is the cohort decision that pays for the dashboard most often.
3. Product or fulfillment investigation triggers — when 30-day repeat rate drops two cohorts in a row by more than ~5 percentage points and CAC source has held constant, the cause is product-side (reformulation, packaging, fulfillment quality, post-purchase email regression). Cohort data is the leading indicator that triggers the investigation.
4. Retention program prioritization — front-loaded vs back-loaded decay shape tells you whether the retention investment should target order-2 acceleration (front-loaded — speed up the second purchase) or order-3+ depth (back-loaded — extend the tail). Different programs entirely.

Decisions cohort data should NOT drive:

1. Weekly ad spend adjustments — a single weekly cohort is too small to read; the variance is mostly noise. Operators who cut Meta spend on one bad-looking week's cohort data are reacting to statistical wobble, not signal.
2. Individual creative kill decisions — creative performance lives in the ad account, not the cohort dashboard. Cohort impact of a single creative is measurable only with very large sample sizes most DTC brands don't have.
3. Pricing changes — pricing experiments need clean A/B controls, not cross-cohort comparison. A cohort acquired at the new price vs a cohort acquired at the old price differs in 14 other variables (channel mix, seasonality, creative). Cohort comparison cannot isolate the pricing effect.
4. Forecasting next quarter revenue — the temptation to project cohort revenue forward into a revenue forecast is strong and almost always wrong. Cohort data is descriptive of acquired customers, not predictive of acquisition volume. Treat the two separately.

The pattern we see consistently: brands that use cohort data for the first four decisions and explicitly ignore it for the second four out-perform brands that try to use it for everything. The dashboard is a leading indicator on policy, not a tactical instrument. Operators who confuse the two end up paying for the data twice — once in the dashboard subscription, once in the spend they reallocated incorrectly chasing noise.

Three signals your current cohort dashboard is misleading you

Pattern recognition:

If two of these are true on the current dashboard, the cohort view isn't reliable enough to drive spend decisions yet.

Decision framework: which cohort metric to pay attention to first

Founders ask which cohort number matters most. The honest answer depends on the stage of the brand:

• Pre-Product-Market-Fit (< $500K ARR): 30-day repeat rate. If first-time buyers don't come back in 30 days, the product isn't sticky enough yet.
• Early growth ($500K-$2M ARR): 90-day cumulative revenue per cohort. This is where paid-spend payback math gets real; if the 90-day per-cohort revenue doesn't cover blended CAC plus COGS, scaling Meta will burn cash.
• Scaling ($2M-$10M ARR): Cohort decay curve shape over 180 days. The flattening shape tells you whether retention is improving or degrading even as the brand grows.
• Mature (> $10M ARR): 365-day cumulative revenue plus retention curve overlay. The brand has enough completed cohorts to measure full LTV; lookback at 365-day cohort revenue tells you if customer quality is improving over time.

Most founders skip to 365-day LTV because it's the most exciting number. The honest answer is whatever metric the brand has 6+ months of completed data for.

Where to next

If you want the broader cross-channel reporting view that the cohort dashboard sits inside, our cross-channel marketing dashboard comparison covers the surrounding tooling. Cohort math is only as honest as the data feeding it — our marketing data analytics guide covers getting to one source of truth before the numbers start arguing. If you want our LTV reporting deployed on your own data, our scoped engagement page walks through implementation. The agency cohort dashboard version is also available as part of Level if you run an agency serving multiple DTC clients. For the fuller KPI picture beyond LTV specifically, our marketing KPI dashboard guide covers what else belongs on the panel. And if you're comparing which platform can actually build these cohort views for you, our marketing reporting software roundup covers the options.