How To Reallocate Paid Spend Using Channel-Level Retention Curves
A decision framework for shifting the next €10K of paid spend using channel-level retention curves, contribution margin, and a CAC crossover threshold.
Reallocating Paid Spend Using Channel-Level Retention Curves
A framework for shifting paid budget between channels using 90/180/365-day retention curves and contribution-margin-adjusted LTV, not blended CAC.
Retention-weighted reallocation is a decision procedure for moving the next slice of paid budget — typically the marginal €5-15K per month — from channels with steep drop-off curves to channels with flatter ones, even when the flatter channel has a worse first-order CAC. It converts each channel's 90-, 180-, and 365-day repeat-purchase curve into a contribution-margin-adjusted LTV, compares that LTV to the channel's true blended CAC, and calculates the crossover point at which a high-CAC channel becomes the better home for the next euro.
The framework assumes you already have channel-tagged cohort data and a reliable contribution margin per order — it is a reallocation tool, not an attribution one.
Most paid-media reallocation still runs on blended CAC and last-click ROAS. That works when every channel acquires roughly the same customer. It fails the moment your TikTok cohort churns twice as fast as your Google non-brand cohort — because the channel that looks cheaper on day 7 is often the more expensive channel by day 180.
This page walks through the three phases: build a contribution-margin-adjusted LTV per channel from your retention curves, compute the crossover CAC at which a higher-CAC channel wins, and execute the shift without breaking platform learning phases. It assumes retention by acquisition channel is already tagged in your data — if it isn't, that's the prerequisite.
Phase 1 — Turn each retention curve into a CM-adjusted LTV
For every acquisition channel, pull the cohort of customers acquired in a given month and calculate the share still purchasing at day 90, 180, and 365. Multiply each retention point by average order value and repeat-purchase frequency to get gross revenue per acquired customer, then multiply by contribution margin (revenue minus COGS, fulfilment, payment fees, and returns) to get CM-adjusted LTV.
Two guardrails matter here. First, only trust curves built on cohorts above your minimum cohort size — usually 400-500 acquisitions per channel per month for a stable 365-day curve. Second, exclude Q4-acquired cohorts from the annualised view when your category is gift-heavy; seasonality-contaminated curves make paid social look far more loyal than it actually is.
Phase 2 — Compute the CM-adjusted crossover CAC
The crossover is the CAC at which a high-retention, high-CAC channel produces the same 12-month margin as a low-retention, low-CAC channel. Formally: crossover_CAC(A) = CM_LTV(A) / CM_LTV(B) × CAC(B). If your channel's actual CAC is below its crossover, the next euro belongs there — even if it looks expensive on a blended-ROAS dashboard.
Worked example: Meta prospecting acquires at €28 CAC with €62 CM-adjusted 12-month LTV. TikTok acquires at €41 CAC but its cohort's flatter curve produces €104 CM-adjusted LTV. Crossover CAC for TikTok is 104/62 × 28 = €47. TikTok's actual €41 sits under €47, so the next €10K goes to TikTok — even though its day-30 ROAS is worse.
Three curves that will lie to you
Google brand-search curves overstate incrementality — you're often paying to reacquire customers who would have returned organically. Attribution windows that don't match your retention window (7-day click vs 365-day repeat) will produce disagreeing CAC and LTV figures for the same channel. And any channel below the minimum cohort size will show a curve dominated by noise, not signal — treat it as directional, not decisive.
Phase 3 — Execute the shift without breaking learning
Move budget in 15-20% weekly steps, not in one shot. Meta and TikTok both have platform minimum spend floors below which their optimisation stops learning; if your reallocation would take a campaign under roughly 50 conversions per week, restructure rather than cut. Pause channels that trip the 90-day retention tripwire mid-quarter — a sudden 10-point drop in day-90 retention on new cohorts almost always precedes a full-curve collapse.
Organic and referral often win the CM-adjusted comparison outright but can't absorb the reallocated budget. Treat those as a spending ceiling, not a scaling channel: fund the content and referral programme up to their absorption limit, then send the residual to the best-scoring paid channel. Re-run the full reallocation monthly; retention curves drift with creative fatigue, promo cadence, and audience saturation.
Front-loaded vs late-flattening retention curves (same day-1 acquisition)
Meta prospecting (front-loaded)
TikTok creator-led (late-flattening)
Google non-brand
Frequently asked questions
Monthly for the crossover calculation, weekly for tripwire monitoring. Retention curves drift with creative fatigue and promo cadence, so a quarterly cadence is too slow — you'll be reallocating into a channel whose curve has already deteriorated.
Aggregate two or three consecutive months into a single cohort, or treat the curve as directional only. Below roughly 400 acquisitions per month per channel, day-180 and day-365 retention points swing wildly on a handful of customers and shouldn't drive a reallocation decision.
Channel-level CAC, not blended MER. The whole point of the framework is to compare channels against each other, and MER collapses that signal. Use a consistent attribution window across all channels — otherwise the CAC and retention curve for the same channel will disagree.
Only with an incrementality adjustment. Google brand curves look phenomenal because they capture customers who would have returned regardless. Discount brand-search LTV by your measured incrementality rate — typically 20-50% — before running the crossover.
Exclude them from the annualised curve, or run a separate seasonal curve. Q4 gift buyers show a false loyalty spike at day 30 and then churn hard by day 90. If you don't segment them out, paid social channels that spike in Q4 will look better than they are.
No — reduce by 15-20% per week. Platform algorithms need consistent conversion volume to hold their learning, and cutting a channel to zero often forces you to rebuild targeting from scratch when you return. Reallocate the marginal budget, not the base.
They usually win the CM-adjusted comparison but can't absorb reallocated paid spend directly. Fund content, SEO, and referral programmes up to their absorption ceiling, then route residual budget to the best-scoring scalable paid channel.
Revenue net of COGS, fulfilment, payment fees, and returns — the actual gross margin per repeat order, not first-order margin. First-order margin often includes acquisition discounts that don't repeat, which inflates apparent LTV for channels that lean on welcome offers.
Yes for channels above the minimum cohort size where you have a validated pLTV model, but validate the prediction against actual 90-day retention every quarter. Predicted LTV models tend to overweight day-1 signal and underweight the curve shape past day 180.
Roughly €5K per month or 10% of total paid budget, whichever is larger. Below that, platform noise and creative variance drown out the retention signal, and you can't distinguish a real curve improvement from a two-week performance blip.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.