Most teams try to reduce customer acquisition cost by negotiating better rates or pausing the worst-performing campaign. That moves CAC by a few percent. The reductions that actually matter — 30% and more, without losing volume — come from changing who you acquire and what happens after they arrive.
CAC is a ratio, and teams almost always attack the numerator. Spend less, pay less per click, squeeze the agency. But the denominator is where the leverage is: how many of the people you paid for become users worth having.
Across the acquisition work I've done, the pattern repeats. The channel that looks cheapest on a cost-per-install basis is usually the one producing the least valuable users, and the channel that looks expensive is often subsidising the whole portfolio. You cannot see this without attribution at the channel level, which is why that is almost always the first thing to build.
Why Blended CAC Hides the Problem
Blended CAC — total spend divided by total new users — is the number most teams report and the least useful number they have. It averages a channel delivering users worth ₹500 with a channel delivering users worth ₹50, and tells you neither.
The fix is unglamorous: attribute spend and downstream revenue to the channel that produced each user, then look at cost against value rather than cost alone. When we built channel-level attribution from scratch on one consumer product and reallocated spend on what it showed, blended CAC fell 30% while acquisition volume held. D30 ARPU came out at ₹295 against an LTV/CAC ratio of 2.8×. Nothing about the media buying got cleverer; the money simply stopped going to users who were never going to pay back.
That work is written up in full, including what the attribution missed, in how we cut user acquisition cost by 30% without cutting volume.
Four Levers, In The Order They Pay Off
1. Stop paying for users who never convert
Before optimising anything, segment acquired users by whether they reached the action that makes them valuable. Most portfolios contain at least one channel where that share is close to zero. Cutting it releases budget immediately and costs nothing in real volume, because the volume was never real.
Using ML to cluster audiences by behaviour rather than demographics on one product cut monthly acquisition spend 31% and raised the profitable share of acquired users 67%, taking blended LTV:CAC to 3.8× and releasing ₹1.8 crore of monthly budget. The detail is in using ML to reduce CAC via smart audience segmentation.
2. Predict value before you pay for it
If you can identify which behavioural signals in the first sessions predict long-term value, you can bid differently for users who show them. On a behavioural model built over 100K+ sessions, precision reached 81%, which was enough to shift targeting: revenue per acquired user rose 61%, blended CAC fell 34%, and acquisition of high-LTV users went up 2.7×. See predicting high-LTV users using behavioural ML.
The trap here is chasing accuracy. A model at 81% precision that changes bidding today beats a model at 91% that ships next quarter — the cost of a false positive in acquisition is one wasted acquisition, which is cheap.
3. Fix the funnel before you refill it
Every point of conversion you recover divides CAC across the whole spend. This is the cheapest lever available and it is consistently deprioritised because it belongs to product rather than marketing. A funnel that converts at 4% instead of 3% cuts effective CAC by a quarter with no change in spend at all.
4. Build acquisition that does not cost per user
Paid acquisition has a floor: you pay for every user, forever. Organic and referral do not. On one product, programmatic organic pages brought acquisition cost to ₹31 per user and eventually carried 61% of acquisition — a channel whose marginal cost approaches zero as it compounds. On another, referral reached 22% of installs.
This is the lever with the longest lead time and the largest ceiling. It is also the reason CAC reduction and growth loop design are the same conversation.
What Does Not Work
Cutting spend across the board. This reduces total cost and total users proportionally. CAC does not move. It feels like action and changes nothing.
Optimising to cost-per-install. Optimising to the cheapest install reliably buys the least valuable users. The metric to optimise is cost per user who reaches value, which requires knowing what value means for your product.
Adding a channel. A new channel spreads the same problem across more surfaces. If the funnel leaks and the targeting is wrong, a fifth channel produces expensive users in a new place.
Where To Start
In order: build channel-level attribution so you can see cost against value; cut the segments that never convert; fix the largest funnel drop before activation; then invest in the acquisition that does not charge per user. Each step funds the next, which matters, because the last one takes the longest and returns the most.
Paying too much for users who don't stay?
Acquisition cost is usually a targeting and funnel problem wearing a media-buying costume. I've taken blended CAC down 30% and more on consumer products without losing volume. Let's look at where your spend is actually going.
Work with Yogesh →