Budget allocation and the media mix

Your best channel isn’t where you should spend every dollar — it’s where the next dollar earns the most.

The idea

Every channel gets less efficient the more you pour into it. Search runs out of high-intent queries; a social audience gets saturated. So the question is never “which channel is best?” — it’s where does the next dollar buy the cheapest conversion?

Keep moving the next dollar to whichever channel is cheapest at the margin, and the channels’ marginal costs slowly converge. When they’re equal, you’re done: no reshuffle can buy another conversion. That last-dollar logic is why you cap even a great channel long before it runs dry.

How it works

Each channel’s conversions saturate — they climb fast at first, then flatten as the audience is used up. The extra conversions from one more dollar (the marginal return) keep shrinking, so the cost of the next conversion keeps rising.

conversions from a channel:   conv(s) = ceiling · (1 − e^(−s / scale))
next conversions per $1 more:  m(s)    = (ceiling / scale) · e^(−s / scale)
marginal CAC (cost of next):   1 / m(s)   <- RISES the more you spend

Rule: move the next dollar to the channel with the LOWEST marginal CAC.
Repeat until every funded channel shows the SAME marginal CAC — the optimum.

Our three channels, $18,000 total:
                 spend     conversions   marginal CAC (next conv)
  even split   6k/6k/6k        1,211      $55  /  $19  /  $16   <- lopsided
  all-in search  18k/0/0         399      $3,026 (search is tapped out)
  optimize     2.7k/5.5k/9.8k     1,298      $18  /  $18  /  $18   <- equal

Search has the cheapest first dollar, so “all-in on the winner” looks tempting — but its pool caps near 400. Dumping $18k there buys almost nothing after saturation. The optimum funds search only to $2,700, then spends where the margin is still cheap. Same budget, +87 conversions over an even split and 3× the all-in plan.

When to use it

Reach for marginal allocation when…	The trade-off / limit
You have a fixed budget and channels with measurable, saturating returns.	You need a decent read on each channel’s response curve — noisy attribution makes the curves fuzzy.
The goal is efficiency — more conversions for the same spend.	Pure last-click math starves upper-funnel brand that pays back later; balance the horizon (below).
You’re rebalancing an existing mix, not launching cold.	New channels have no curve yet — reserve a small test budget to learn one.

Brand vs performance. Performance channels convert demand that already exists; brand creates demand that shows up weeks or quarters later. Short-horizon or cash-tight goals tilt the split toward performance; a growth mandate with runway funds more brand, because today’s brand is next quarter’s cheaper performance.

Watch out for

• Ranking channels by average CAC. The channel with the best average CAC is usually already saturated. Decisions live at the margin, not the average.
• Going all-in on the winner. Concentrating spend drives that channel up its own diminishing-returns curve until the last dollars buy nothing. Cap it and diversify.
• Trusting last-click attribution alone. It systematically under-credits brand and top-of-funnel, so the “curve” you optimize against is biased toward the closer. Sanity-check with holdouts or a media-mix model.
• Optimizing a single week. Curves shift with seasonality, creative fatigue, and competitor bidding. Re-fit and rebalance on a cadence, not once.
• Zeroing out brand to hit a CAC target. It flatters this quarter’s efficiency and quietly raises next quarter’s CAC as demand dries up.

Worked example

An interviewer says: “You have $18k a month. Search returns a $12 CAC, social $16, brand $22. Where does the money go?” The trap is to pour it into search. A strong answer: those are average CACs at some current spend — I’d ask for each channel’s marginal CAC and shift budget from the highest marginal CAC to the lowest until they equalize. Search’s cheap average hides a small, quickly-saturated pool, so I’d likely cap it and fund social and brand further out. Then I’d name the horizon: if we need conversions this month, tilt to performance; if we’re building the category, protect a brand line, because it lowers everyone’s CAC later.