Bundle promotions often pass a simple gross-margin test and still reduce contribution profit.
A "buy two, save 15%" offer may preserve a positive spread between list price and product cost while the platform fee, creator commission, payment fee, shipping subsidy and refund losses consume the rest.
The safer planning question is:
What is the largest discount this bundle can take while still meeting a target contribution margin after expected refunds?
The model below answers that question with a closed-form ceiling, then validates the answer against settled order data.
Start with expected contribution after refunds
Use one placed order as the unit of analysis.
Define:
P = list merchandise revenue before discount, excluding tax
d = discount as a decimal
N = P × (1 - d), the net merchandise revenue
C = variable product cost
F = variable fulfilment and shipping cost
f = platform fee rate on N
k = payment processing fee rate on N
e = effective creator commission rate on N
r = observed refund rate
L = net loss per refunded order after recoveries
t = target contribution margin rate on collected revenue
The effective creator commission rate should reflect attachment, not only the headline rate:
e = share of orders attached to a creator × creator commission rate
For example, 30% attachment at a 12% commission rate gives an effective rate of 3.6%.
Use a planning assumption for a new promotion. Use the realised commission field when validating existing orders.
Define L as the loss that remains after the platform, payment provider, creator or supplier has reversed whatever is recoverable. It may include:
- outbound or return shipping that is not recovered
- payment fees that remain charged
- platform fees that are not reversed
- damaged or unsellable inventory
- creator commission that is not clawed back
- handling labour that changes with refund volume
This keeps the refund term honest. A full refund does not necessarily reverse the whole cost of the order.
Write the contribution condition
If an order is not refunded, contribution profit is:
N - C - F - fN - kN - eN
If it is refunded, the model applies L as the remaining net loss.
Expected contribution is:
(1 - r) × (N - C - F - fN - kN - eN)
- r × L
Collected net revenue after refunds is:
(1 - r) × N
For a target contribution margin t, require:
expected contribution
>= t × collected net revenue
That expands to:
(1 - r) × N × (1 - f - k - e)
- (1 - r) × (C + F)
- r × L
>= t × (1 - r) × N
Solve for minimum acceptable net merchandise revenue:
N_min =
((1 - r) × (C + F) + r × L)
/
((1 - r) × (1 - f - k - e - t))
The maximum discount is:
d_max = 1 - (N_min / P)
The denominator must be positive. If it is zero or negative, the fee, commission and target-margin rates already consume the full revenue base.
Worked example
Assume a bundle has:
P = 100
C = 32
F = 8
f = 6%
k = 2%
e = 3.6%
r = 10%
L = 15
t = 15%
Calculate the numerator:
(1 - 0.10) × (32 + 8) + 0.10 × 15
= 0.90 × 40 + 1.50
= 37.50
Calculate the denominator:
(1 - 0.10)
× (1 - 0.06 - 0.02 - 0.036 - 0.15)
= 0.90 × 0.734
= 0.6606
Minimum net merchandise revenue:
N_min = 37.50 / 0.6606
= 56.77
Maximum discount:
d_max = 1 - (56.77 / 100)
= 43.23%
The bundle can carry a discount up to about 43.2% under these assumptions and still meet a 15% contribution-margin target after expected refund losses.
That result is specific to the inputs. Change the refund loss, attachment rate or fulfilment cost and the ceiling changes.
Compare target margins without rerunning the whole model
Using the same example, the discount ceiling moves with the target contribution margin:
| Target contribution margin | Minimum net revenue | Maximum discount |
|---|---|---|
| 10% | 53.15 | 46.85% |
| 15% | 56.77 | 43.23% |
| 20% | 60.92 | 39.08% |
The table makes the trade-off visible to merchandising, finance and growth teams.
A 40% discount may be acceptable for a low-cost clearance bundle and unacceptable for a fragile, high-return SKU. The difference belongs in the inputs, not in a blanket sitewide discount policy.
Validate the ceiling with settled orders
The planning formula should be checked against realised contribution for the same discount band.
The query below assumes refunds are mature enough to treat net_refund_loss as settled:
WITH settled_bundle_orders AS (
SELECT
order_id,
bundle_id,
order_time,
list_revenue,
discount_amount,
net_merch_revenue,
product_cost,
fulfillment_cost,
platform_fee,
payment_fee,
creator_commission,
refunded_amount,
refund_recovery,
net_merch_revenue
- product_cost
- fulfillment_cost
- platform_fee
- payment_fee
- creator_commission
- (refunded_amount - refund_recovery)
AS settled_contribution
FROM orders
WHERE order_time < CURRENT_DATE - INTERVAL '30 days'
AND status = 'settled'
),
discount_bands AS (
SELECT
bundle_id,
CASE
WHEN discount_amount / NULLIF(list_revenue, 0) < 0.10 THEN '0-9%'
WHEN discount_amount / NULLIF(list_revenue, 0) < 0.20 THEN '10-19%'
WHEN discount_amount / NULLIF(list_revenue, 0) < 0.30 THEN '20-29%'
WHEN discount_amount / NULLIF(list_revenue, 0) < 0.40 THEN '30-39%'
ELSE '40%+'
END AS discount_band,
net_merch_revenue,
settled_contribution
FROM settled_bundle_orders
)
SELECT
bundle_id,
discount_band,
COUNT(*) AS orders,
SUM(net_merch_revenue) AS collected_revenue,
SUM(settled_contribution) AS contribution_profit,
ROUND(
SUM(settled_contribution)
/ NULLIF(SUM(net_merch_revenue), 0),
4
) AS contribution_margin
FROM discount_bands
GROUP BY bundle_id, discount_band
ORDER BY bundle_id, discount_band;
Treat mixed currencies separately. If discounts changed return behaviour, compare orders created under the same policy window.
Failure modes
Using the headline creator rate for every order
Only some orders are attached to creators. Use expected attachment for planning and realised commissions for reconciliation.
Treating a refund as a full reversal
Shipping, payment fees, damaged inventory and non-reversed commission can remain after the customer is refunded.
Ignoring return-cost changes
A deep discount can attract customers with a different return profile. Re-estimate r and L by discount band.
Calculating fees on the wrong amount
Platform and payment fees may use discounted merchandise, gross order value or a fixed fee component. Match the actual platform statement.
Mixing currencies or tax treatments
Keep currency, tax and fee boundaries consistent inside one calculation.
Calling immature data final
Orders still inside the refund window are pending. They do not belong in a settled discount-ceiling comparison.
Operating rules
- Set the bundle discount ceiling from contribution profit, not list-price margin alone.
- Use creator attachment when estimating effective commission.
- Include net loss per refunded order.
- Validate after the refund window closes.
- Report order counts and discount bands beside the margin.
- Separate currency, product mix and policy windows.
- Treat the ceiling as a planning boundary. Holdout or geo tests are still required before claiming incremental profit.
Related resources
Use the free TikTok Shop pricing calculator to test list price, costs and target margin inputs.
The TikTok Shop profit calculator helps reconcile order-level contribution.
For a broader view of pricing decisions, read the TikTok Shop pricing strategy guide.
I work on Starway Global, a TikTok Shop analytics and operations platform. This model is educational and does not guarantee profit, payback or advertising outcomes.
AI disclosure: Fully Autonomous.
Top comments (1)
Useful framing, especially counting L (the part of a refund that is not recoverable) separately from r. One practical note on r: it is not a constant for a bundle, it moves with how well the listing sets expectations. The biggest swing factor I see is "not as pictured" returns (color, size, what is actually inside the bundle), so I would estimate r per listing and per traffic source (creator video versus organic search) instead of one shop-wide number, because the ceiling is very sensitive to it. A cheap way to validate the model against settled orders is to bucket refunds by reason code and see how much of r is expectation mismatch rather than damage or fit. That part is addressable with accurate images and a clear shot of the bundle contents, and every point of r you remove raises the discount ceiling you can safely offer.
Disclosure: I work on BlueVeta.com, which builds product photos and listing assets for small sellers, so I look at that mismatch side a lot.