Several product lines run through one acquisition funnel and one shared entry point, each with a dedicated budget and a dedicated target. A single blended number cannot say which line a customer belongs to, so a campaign built for one line can quietly move numbers for another, and the credit lands wherever the reporting happens to fall. Splitting budget between the internal owners of each line then falls to whoever owns the account, with no way to prove who earned what. Every split defended to those owners rests on judgement alone, not a number they can check.
A model per line that carries the other lines' spend as a known input makes a campaign's effect on a line it was not built for visible instead of absorbed. The resulting split traces to a cause rather than a share picked by feel. The allocation conversation starts from a number, not an opinion.
The instinct is to start with the market carrying the most data and the fewest open questions, because a narrow pilot feels safer than asking for full budget up front. But consumer-side spend already moves numbers on the business side for reasons nobody disputes, so a narrow scope does not narrow what the model has to account for. Committing to the full scope costs more than the quarter can obviously defend, and the person who proposed it becomes the one graded on whatever number comes back, a verdict on judgment rather than on the method.
A starting scope sized to what the quarter can defend now, with the other side's spend already built in as context rather than excluded and hoped away, resolves the mismatch. An early result stands up under internal challenge, because nothing material was left out to make it look cleaner. The case for wider scope builds on evidence, not on who sponsored it.
Paid spend runs on both sides of a two-sided business, but the number that growth gets judged on moves mostly with the network itself. A large merchant coming on board can send new users up sharply, and those users then convert more easily on merchants already there, a loop unrelated to any campaign run that week. A merchant leaving does the same in reverse, cutting new and returning users at once, and even good weeks are hard to read because the same tracking issue runs both ways. Accountability for growth lands on whoever owns the media plan, even though its real engine is the merchant network, with no way to say how much of a given week came from the plan itself.
Spend on both sides read against the network's state attributes a spike or a drop to its actual cause. Separating what the network did from what the campaigns did stops being a guess. And the resulting number holds up as owned, not borrowed.
Enrollment numbers in the founding market still dwarf every other market in operation, and this year's growth target assumes some of it comes from somewhere else. For the newer markets only total attributed revenue is available, not what a dollar of student recruitment spend produced there, because reporting groups whole regions into one block by language rather than by market. The annual enrollment goal is already ambitious, and finding growth in the least understood markets means a named author stands behind a reallocation with no prior evidence if it goes wrong.
A per-market read instead of a blended regional one stops a smaller market's real recruitment return from hiding inside a bigger neighbor's number. The next move gets sized by what a market has actually returned, not by which market has the longest track record. The resulting recommendation traces back to a cause when leadership asks why.
Spillover is when spend aimed at one product line, market, or side of a two-sided business measurably affects another. A campaign built for one line can lift or suppress numbers on a line it was never meant to touch, and a blended report absorbs that effect into whichever line happened to report it.
Each product line or market gets modeled with the others' spend included as a known input rather than left out. That makes it possible to see how much of a line's result came from its own campaigns and how much arrived as an effect of spend aimed somewhere else, instead of guessing at the split.
The same logic applies, with one addition specific to fintech: product lines such as wealth, lending, or payments often share one acquisition funnel and one entry point, so the model has to separate lines that were never built to be measured apart in the first place.
When a product line or market is too small to separate from the rest on its own spend and history, or when the decision needed is a day-to-day tactical call rather than a strategic split. In those cases the honest answer is that the line cannot yet be read apart, not a number dressed up to look precise.
The allocation conversation stops being an argument between opinions and becomes a review of a traceable result. Each internal owner sees what their line actually produced and what it received from elsewhere, so the next budget split is defended with a number rather than negotiated by seniority.