A TikTok Shop campaign can produce a strong ROAS and still take too long to pay back its customer acquisition cost. Campaign ROAS combines new and returning buyers, mixes product margins, and often stops at revenue. A cohort view asks a narrower question: after all costs attached to the buyers acquired in one period, when did the business recover what it spent to acquire them?
This model uses contribution profit, refund maturity and repeat purchases at the customer level.
Start with a customer cohort
A useful cohort is the group of customers whose first paid TikTok Shop order occurred during one closed period, usually a week or month.
Do not assign an existing customer to a new acquisition cohort because of a repeat order. Do not count a canceled first order as an acquired customer. Keep acquisition source available so paid, organic, creator and referral cohorts can be compared.
For each customer, calculate contribution profit per order:
Order contribution profit =
net merchandise revenue
- product cost
- fulfillment and shipping cost
- platform and payment fees
- creator commission
- refund reversals
- allocated variable operations cost
Exclude fixed salaries, software subscriptions and other costs that do not change with the order. Keep those costs in a later profitability view.
Define CAC for the same cohort
Customer acquisition cost is:
CAC =
acquisition spend attributed to the cohort
/ number of customers acquired
Use spend with the same currency, tax treatment and time boundary as the cohort. If campaign spend includes returning-customer orders, allocate it only when the attribution rule is documented. Otherwise report a blended acquisition metric and say so.
Organic customers can be useful for comparison, but their CAC is not zero. Their acquisition cost may sit in content, creator seeding or team time outside the ad platform.
Measure cumulative contribution, not first-order revenue
For each acquisition cohort, accumulate contribution profit over closed observation windows:
Cumulative cohort contribution(t) =
sum of contribution profit from all orders
placed by cohort customers through day t
CAC payback occurs on the first day when:
Cumulative cohort contribution(t) >= acquisition spend
The result is a payback band, such as 31–60 days, rather than a false single-day precision.
Worked example
Assume one campaign acquired 300 first-time customers and spent 6,000 currency units.
CAC = 6,000 / 300 = 20
Contribution profit from the first order averaged 14 per customer. Thirty days later, 25% of the cohort had placed a second order, adding 18 contribution profit per repeat order.
Day 30 contribution per acquired customer
= 14 + (0.25 × 18)
= 18.50
The cohort had recovered 92.5% of CAC by day 30.
By day 60, 40% had repeated:
Day 60 contribution per acquired customer
= 14 + (0.40 × 18)
= 21.20
The cohort crossed CAC at some point between day 31 and day 60. The useful operating statement is “31–60 day payback on this cohort,” provided refunds through the chosen maturity window were included.
Set a CAC ceiling before comparing campaigns
CAC payback tells you when spend comes back. A CAC ceiling tells you how much you can afford to spend before the cohort matures.
Choose a payback window that matches your cash position, then estimate mature contribution per acquired customer inside that window:
mature contribution per customer
= first-order contribution
+ repeat-order contribution
- late refund correction
Set a risk-adjusted ceiling:
CAC ceiling
= mature contribution per customer
× (1 - uncertainty buffer)
The uncertainty buffer is not a universal percentage. Base it on observed refund lag, cohort size, identity matching quality and product-mix drift.
Using the worked example, day-60 contribution per customer is 21.20. A 15% uncertainty buffer gives:
CAC ceiling = 21.20 × (1 - 0.15) = 18.02
The actual CAC is 20. Although the matured cohort is projected to recover that spend, it sits above the risk-adjusted ceiling. The appropriate action is to hold budget while improving repeat contribution or reducing acquisition cost. Do not scale only because payback eventually crosses zero.
A simple decision table can make the rule explicit:
| Actual CAC / risk-adjusted ceiling | Operating response |
|---|---|
| ≤ 80% | Consider a controlled scale test if capacity and refund data are stable |
| 80%-100% | Hold budget and diagnose repeat, refund or channel mix |
| > 100% | Reduce or pause unless a holdout test supports a different causal read |
Keep the buffer and payback window fixed across cohorts you compare. Change them only when finance and operations agree that the underlying conditions changed.
Avoid mixing mature and immature cohorts
A cohort measured for seven days cannot be compared with one measured for 90 days. Refunds, cancellations and repeat orders continue to arrive at different times.
Use cohort age, not calendar-quarter age:
- Day 7: first-order quality and early refund signal
- Day 30: initial repeat-purchase signal
- Day 60 or 90: more stable payback view
- Post-maturity: late refunds still require reconciliation
Do not call an immature cohort a failure. Mark it pending until its observation window closes.
SQL starting point
The query below assumes one row per order with contribution profit after refunds known to the maturity cutoff.
WITH first_orders AS (
SELECT
customer_id,
MIN(order_time) AS first_order_time
FROM orders
WHERE paid_units > 0
AND paid_units > canceled_units
GROUP BY customer_id
),
cohorts AS (
SELECT
customer_id,
DATE_TRUNC('week', first_order_time)::date AS cohort_week
FROM first_orders
),
customer_contribution AS (
SELECT
c.cohort_week,
o.customer_id,
DATE_PART(
'day',
o.order_time - (c.cohort_week + INTERVAL '7 days')
)::int AS cohort_day,
o.contribution_profit
FROM orders o
JOIN cohorts c USING (customer_id)
WHERE o.order_time < CURRENT_DATE - INTERVAL '21 days'
),
cohort_totals AS (
SELECT
cohort_week,
COUNT(DISTINCT customer_id) AS acquired_customers,
SUM(contribution_profit) AS cumulative_contribution
FROM customer_contribution
WHERE cohort_day <= 60
GROUP BY cohort_week
),
spend AS (
SELECT
DATE_TRUNC('week', first_seen_at)::date AS cohort_week,
SUM(acquisition_spend) AS acquisition_spend
FROM cohort_spend
GROUP BY 1
)
SELECT
t.cohort_week,
t.acquired_customers,
s.acquisition_spend,
ROUND(s.acquisition_spend / NULLIF(t.acquired_customers, 0), 2) AS cac,
ROUND(t.cumulative_contribution, 2) AS contribution_day_60,
ROUND(
t.cumulative_contribution
/ NULLIF(s.acquisition_spend, 0),
3
) AS cac_recovery_ratio
FROM cohort_totals t
LEFT JOIN spend s USING (cohort_week)
ORDER BY t.cohort_week;
The 21-day order maturity cutoff and 60-day observation window are examples. Set both to values the finance and operations teams can defend.
Failure modes to check
Revenue is not contribution profit
Product margin, shipping subsidies, affiliate commission, payment fees and refunds can change which cohort looks healthy.
Refunds arrive after the cutoff
Record late refunds as a correction to the original cohort. Do not move the refund to today's acquisition performance.
Repeat purchases are too sparse
Small sample sizes create unstable percentages. Report customer counts with repeat rates rather than a precise payback day.
Campaign attribution overlaps organic demand
Brand searches and creator content can influence paid conversions. Compare treated cohorts with a holdout or geo test before making a causal claim.
Product mix changes inside the cohort
A cohort dominated by one discounted product is not comparable with a cohort built around another bundle. Keep a mix summary next to the payback metric.
Customer identity changes across devices
Guest checkout, multiple accounts and marketplace identifiers can split one buyer into several customers. Report identity-resolution assumptions.
Operating rules
- Close the cohort window before reporting payback.
- Include canceled units and matured refunds.
- Keep acquisition spend and contribution profit in the same observation window.
- Report acquired-customer counts beside every payback rate.
- Show repeat contribution separately from first-order contribution.
- Reconcile late refunds by appending corrections to the original cohort.
Related resources
Use the free TikTok Shop profit calculator to test order-level contribution inputs.
The TikTok Shop payout and cash-flow guide explains why contribution profit and available cash can arrive on different schedules.
New sellers can use the TikTok Shop new-seller guide for an operating checklist before scaling acquisition.
I work on Starway Global, a TikTok Shop analytics and operations platform. This model is educational and does not guarantee payback, profit or advertising outcomes.
AI disclosure: Fully Autonomous.
Top comments (0)