The split-scorecard problem: on a marketplace, each function hits its own number while the platform loses value on the actual buyer-seller transactions underneath. Shared accountability is the fix.
Every marketplace operator knows the flywheel: buyers draw sellers, selection draws buyers, and liquidity compounds. That part is settled. The harder problem sits one level down, in the operating model.
Most marketplaces already measure both the buyer and seller sides well. They still reward each team for optimizing its own side. What no one is accountable for is the exchange itself, meaning the buyer-seller transaction and everything behind it, from price and availability to fulfillment and service. That missing layer is where value quietly leaks.
The split-scorecard problem
A Head of Digital is measured on conversion and repeat. A Head of Marketplace Sales is measured on seller acquisition, GMV, and ad revenue. Operations owns fulfillment defects. Finance sees the margin at the end. Each one can hit the assigned number while the exchange itself becomes less valuable.
A ranking change lifts conversion by piling demand onto sellers who cannot fulfill it. A seller incentive brings in more listings, but many come from sellers who cannot reliably stock or ship them, so cancellations and returns climb. Retail media books attributed revenue against products nobody buys twice. Each of these is a reasonable call inside one function, and a win on that function's scorecard. The loss is systemwide, and no scorecard shows it.
This is the split-scorecard problem: every function can see the whole marketplace, but none of them owns the exchange itself. The damage only surfaces in the P&L, after it has already spread across all of them.
Incisiv's 2026 research, in conjunction with Anaplan, surveyed 298 retail supply chain and merchandise planning executives. It found that 95% operate without unified cross-functional incentives. Aligned incentives show up in only 5% of the broader field, but in 24% of the operational leaders.
The dashboards already exist. What is missing is one view of the buyer-seller exchange that someone actually owns.
The economics hidden by marketplace averages
GMV, take rate, conversion, and active sellers can all move in the right direction while the individual exchanges underneath get less profitable. The averages hide which buyer cohorts create repeatable demand, which sellers serve that demand profitably, and which combinations quietly generate returns, support costs, and churn.
The fix is to change the unit of analysis from the marketplace to the exchange. Tie every buyer outcome to the seller, offer, and fulfillment path that produced it, and growth stops being a single number. You can see which exchanges produce repeat purchases and viable seller economics, and which ones reverse into cancellations and churn once the full cost is counted.
That shift changes the two headline metrics. Buyer LTV stops being a marketing number and becomes a marketplace number. Seller economics stop being fee yield and become a measure of the value a seller creates for the exchange.
That is where the questions worth asking get answered. Does a promotion create demand that lasts, or demand that spikes and fades? Does a take-rate change protect platform value or erode it? Does a high-volume seller strengthen liquidity or consume it?
Act on the exchange while the signal still matters
Seeing the exchange clearly is worth little if you see it weeks late. A monthly or quarterly business review tells you what already happened, and a marketplace moves faster than that. A stockout or a fulfillment delay can move discovery and conversion within hours. The damage to repeat behavior and margin stays buried for weeks.
This is where the shared decision model has to become operational. It needs current data from both sides, connected at the point where decisions get made, so a leader can see how one move, like a ranking slot or a fulfillment change, affects buyer value and seller contribution at the same time.
The cost of not having it is measurable. Incisiv estimates that slow decisions cost the average retailer five cents on every dollar earned, against roughly two cents for the operational leaders. The typical signal-to-response cycle runs four to twelve weeks by the time a change is sensed, decided, executed, and attributed.
In a marketplace that decision lag hurts more, because a single signal usually needs several parties to move. When cancellation risk climbs, the fix can mean resetting inventory promises, dropping the ranking, and pausing media at once, with the affected sellers looped in. By the time each team has spotted its own piece and the reports are reconciled, the customer’s bad experience has already happened.
That is where near-real-time infrastructure earns its place. It lets teams and sellers catch a change, price its impact, and coordinate a response while the outcome is still recoverable.
Shared data, separate views
In Incisiv’s 2026 inventory intelligence research, retailers reported strong real-time visibility into 88% of inventory held in their own distribution centers. Visibility fell to 14% for inventory held by marketplace sellers, and only 17% could use marketplace-seller inventory to fulfill orders. The supply is sitting right there. The decision system cannot see it, or cannot use it, when it counts.
Connecting the data is not the same as exposing it. The marketplace operator holds customer records, seller performance, and platform economics. A seller needs the signals to run its own business, its own demand, inventory, and service performance, and nothing more. It does not need another seller's performance or economics, and it never needs customer PII.
A shared model with separate views does both. Each participant sees the signals a specific decision requires and no more:
- Each seller sees its own demand, inventory risk, and service gaps, and nothing from another seller
- The operator's commercial teams see how each seller's actions move buyer retention and platform contribution
- The operator's finance team sees the full economics of every exchange
Access controls and aggregation rules keep the private data private while everyone works from the same facts. Too little information and nobody moves fast enough. Too much and you have a privacy and competitive-exposure problem. The goal is a shared truth, not a shared free-for-all.
Someone has to own the tradeoff
Technology connects the signals, and increasingly it can act on them on its own, repricing, reranking, or pausing a campaign the moment a trigger fires. What it cannot do is decide which outcome should win when buyer value and seller growth pull against each other. That call, and the accountability for it, stays with people. You can put a flawless shared view of the exchange in front of every team and still leave the old incentives untouched underneath it. Now everyone sees the same numbers, but each team is still rewarded for a different one. The fix here is cross-functional incentives rather than more technology. Whoever is given the authority to act on the tradeoff should be measured on the whole exchange, not on one side of it.
Start narrow. Pick one category where the marketplace averages are hiding a fight. Name the buyer cohorts that matter, the sellers serving them, and the costs riding on those exchanges. Connect buyer LTV, seller contribution, and platform revenue for that slice. Then take one recurring decision, say ranking or promotions, and judge it against the whole economics of the exchange instead of the single metric it usually optimizes.
A pilot like that should answer four questions:
- Which buyer-seller interactions create value worth repeating
- Which local wins are quietly creating downstream cost
- Which signals need a response before the next reporting cycle
- Which teams and sellers need real authority to act
Build it so teams make decisions with it, not so executives get one more dashboard to admire.
Make the exchange the unit of accountability
Marketplace operators connected buyers and sellers long ago. What they still run separately is the economics, the incentives, and the information behind those connections. The next edge goes to whoever manages the exchange as one system instead of two.
Most operators already have the data somewhere in the stack. The work now is unglamorous. Connect it, govern it, and put it in front of the people and sellers who have to act. Start with one category and one decision, and prove that the exchange, not the silo, is the thing worth optimizing.
