Marketplace fee change: the response plan
A marketplace changes a fee or a penalty in the middle of your season. Here is how to detect it before the settlement does, which numbers to redo and in what order, and how to decide per SKU instead of panicking per category.
- Track fee charged divided by order value by fee type each week, because a step change in that ratio detects a fee revision even when the announcement was missed.
- Recompute in order from per-order charge to per-SKU contribution to advertising payback, since break-even ROAS moves the moment contribution moves.
- Tiered percentage charges step rather than curve, so orders sitting just above a band boundary need a deliberate decision and discounts can push orders across boundaries.
- Delist decisions belong at SKU level using contribution, structural role and inventory position, because a negative SKU with deep cover is a liquidation plan rather than a delist.
Marketplaces change fees and penalties on their own schedule, and the shape of those changes has shifted: charges that were once a flat amount are now a percentage of order value with tiers, and dispatch and cancellation penalties sit in bands. That structural shift matters more than any single revision, because it turns a number you could hard-code into a function of basket value. What follows is the change management routine, not an explanation of the fee lines themselves.
Find the change before your settlement does
Platforms announce through a banner in the seller panel, a policy page carrying an effective date, and an email to the registered address, which in most Indian seller accounts is a shared mailbox nobody owns. The practical result is that the first person to notice a fee change is whoever opens a settlement report three weeks later and finds it short.
Two layers fix this. The first is a change register with a named owner per platform, a fifteen-minute weekly sweep of the announcements tab and the fee schedule page for every account you run, and a dated entry recording effective date, scope and categories. Save the fee schedule page each month so you have something to diff against.
The second layer catches what the first one misses. Build a monitor that divides fee charged by order value, split by fee type, by week. A step change in that ratio is a fee change whether or not anybody told you. It is a detection signal, not a reconciliation exercise, and it belongs on the ops dashboard rather than in a finance close.
The recompute, in order
When a change lands, most teams open the channel P and L. That is the wrong place to start, because it is the last number to move and the least useful one for a decision.
Recompute in this sequence. Start at the order level, since most changes are per-order rather than per-SKU and a per-order charge lands very differently on a single-unit order than on a four-unit basket. Then take per-SKU contribution at your actual current selling price and your actual mix of order sizes, not at list price. Then isolate the SKUs that cross zero, because those are the only ones requiring a decision this week. Then blend up to channel contribution. Then, and this is the step almost everyone skips, redo your advertising payback, because break-even ROAS moves the moment contribution moves and every bid rule keyed to a stale target is now wrong.
Run all of it against your last sixty days of real orders rather than a modelled basket. Tiered percentage charges behave according to your actual order-value distribution, and no spreadsheet assumption reproduces that distribution accurately.
The band boundary problem
When a charge is a percentage of order value with tiers, the effective rate does not rise smoothly. It steps. Two orders a few rupees apart can sit either side of a boundary and carry materially different charges, and the difference is not proportional to the price gap.
Plot your order values as a histogram with the boundaries drawn on it. Any mass sitting just above a line is a decision. You have three options: move the price just below the boundary if the resulting price is defensible and does not break parity with your other channels, change bundle or pack composition so multi-unit baskets land inside a better band, or accept the rate because the volume at that price point is worth more than the rate saving.
The trap is treating the band as a curve. The second is forgetting that discounts move orders across boundaries. A promotion built on pre-change arithmetic can be healthy at full price and loss-making at the discounted price, purely because the discount pushed the order into a different band.
Delist or absorb, per SKU and not per category
Category-level responses destroy value in both directions, because a single category holds hero SKUs at thirty percent contribution and tail SKUs at three. Make the call per SKU on three inputs: contribution after the new charge, the SKU’s structural role, and its inventory position.
A negative-contribution SKU sitting on nine hundred units of cover is not a delist, it is a liquidation plan. A negative-contribution SKU that is the entry point into a range you make money on is a marketing cost you decide to keep, with a written review date. Only a SKU that is negative, structurally isolated and thin on stock is a genuine delist, and delisting properly means no stock, no ad spend and enough cover elsewhere, plus an honest acceptance that you will pay the rank cost again when you relist.
Reprice in waves, not all at once
Two reasons to stagger. The first is attribution. Move four hundred prices in one night, watch units fall twelve percent, and you cannot separate the fee change from the price change from the week’s seasonality. The second is reaction. Competitor repricers and platform parity checks respond, and a simultaneous catalogue-wide move triggers those responses everywhere at once.
Sequence it. Wave one is ten to twenty SKUs where the arithmetic is unambiguous and the strategic risk is lowest, held for a full sales cycle before you read anything. Read units and category share, not revenue. Wave two applies what you learned to the mid tier. Heroes move last and by the smallest increment, because that is where a wrong move is most expensive, and never in the same week as a deal event.
When the change lands weeks before peak
This is the hard version. Purchase orders are placed, stock is already inside the fulfilment centre, ad budgets are committed. The instinct is to declare everything locked and absorb the hit.
Separate what is genuinely locked from what you have merely been treating as locked. Inventory is locked, and pulling it back out of a fulfilment centre costs money. Ad budget is not locked, it reallocates daily and it is your fastest lever. Neither is price, nor which SKUs get the deal slots. So the question is not whether to sell the committed stock but at what price and with how much ad support, which means rebuilding the event plan around post-change contribution.
Take your ask to the category manager while the change is fresh and every seller is raising it, and make it specific, such as a deal slot or a co-funded discount, rather than a general complaint. And write down the pre-change baseline before you touch anything, because otherwise the post-event review becomes an argument about what the change actually cost you.