A multi-category snack and pantry subscription brand selling curated monthly boxes alongside à la carte refills replaced first-box bidding with predicted Month-6 retention value using AdZeta's ValueBid™. The result: a 44% ROAS lift, a 31% reduction in 60-day churn, and a 2.1x increase in the share of new subscribers active at the Month-6 milestone.

A growth-stage subscription commerce brand selling curated monthly snack and pantry boxes alongside an à la carte refill catalog. AOV ranges from $39 (single-box trial) to $135 (premium-tier monthly + add-ons). Subscriber LTV is heavily skewed toward customers who reach Month 6, the brand's primary retention milestone.
Roughly 42% of new subscribers churned before their second full delivery. Of the 58% who continued, fewer than half reached Month 6. The auction had no visibility into either churn cliff. Google and Meta were optimizing toward first-box conversion, and the bidder was systematically pulling in trial-driven, discount-led conversions that never reached subscription profitability.
Campaigns ran on Google Performance Max with tROAS configured against AOV, and Meta Advantage+ Shopping with first-event optimization. The team had built sophisticated lifecycle email flows in Klaviyo and an offline retention model in their warehouse. The retention model never flowed back into Google or Meta. The bidder was making allocation decisions on first-box value, not Month-6 retention probability.
AdZeta deployed ValueBid™ to predict each visitor's Month-6 retention probability multiplied by 12-month subscription revenue, from non-PII first-party signals (taste-preference quiz responses, browsing path, plan-tier exploration, à la carte engagement, household composition indicators). Predictions flowed into Google via OCI and Meta via CAPI as the conversion value the bidder optimized against.
Integration ran from Recharge, Shopify, Klaviyo, and the brand's data warehouse into the AdZeta pipeline in Week 1. Model training completed in Week 2 with AUC 0.86 on Month-6 retention prediction. ValueBid™ went live in Week 3 with a paired control on 50% of media spend for six weeks. Following the relearning period, ValueBid™ became the default conversion signal across both platforms.
At Week 16 post-relearn, blended ROAS was up 44%, 60-day churn was down 31%, Month-6 active subscriber share was up 2.1x, and average new-customer 12-month subscription revenue was up 47%. The brand expanded prospecting on Meta into previously deprioritized creative concepts that ValueBid™ identified as high-pLTV despite under-indexing on first-box ROAS.
Subscription commerce economics are concentrated at retention milestones, not at first-box conversion. Month 6 is where the bulk of LTV concentration happens. The bidder cannot see Month 6 from Week 1. Without a forward-looking signal, the auction over-invests in trial-driven discount-seekers and under-invests in customers most likely to reach the milestone that actually drives profit.
In subscription commerce, the first box is where the brand pays the highest CAC and earns the lowest margin. The auction is being optimized against the least profitable transaction in the entire customer relationship. Optimizing for more first boxes just multiplies the brand's exposure to the 60-day churn cliff.
Heavy first-box discounting (40 to 60% off is industry standard) pulls in promo-driven subscribers whose pLTV is dramatically lower than full-price acquisitions. In an AOV-bidding world, the auction cannot tell them apart at conversion. Discount-aware feature engineering in the ValueBid™ model was a prerequisite for honest LTV prediction.
The brand had built a competent Month-6 retention model in their warehouse. It informed lifecycle email cadence, offer targeting, and concierge intervention. It never flowed into Google or Meta's bidder. The most predictive signal in the business was sitting outside the system that allocated the most spend.
ValueBid™ takes the brand's quiz, lifecycle, and subscription data and predicts Month-6 retention probability multiplied by 12-month subscription revenue per visitor in real time. The score becomes the conversion value sent to Google OCI and Meta CAPI. The auction relearns on the new objective and starts winning impressions for the customer profile that actually crosses Month 6.
AdZeta's brand-specific model jointly predicts Month-6 retention probability and 12-month subscription revenue using taste-preference quiz responses, plan-tier exploration, à la carte engagement, browsing depth, and household composition signals. The composite pLTV score is the conversion value sent to the auction.
The model includes explicit features for first-box discount tier, promo source, and price sensitivity. This prevented systematic LTV over-prediction on heavily discounted cohorts that historically churn at twice the rate of full-price acquisitions, a common failure mode in pLTV models built without subscription commerce expertise.
Predictions are delivered as conversion values to Google's Offline Conversion Imports and Meta's Conversions API in real time. Google Performance Max and Meta Advantage+ Shopping campaigns kept their existing structure. Only the value the bidder is optimizing against changed.
AdZeta's reporting layer breaks Month-6 retention forecasts out by acquisition cohort, channel, and creative concept. The brand's CFO uses the cohort forecast as the primary input to monthly subscription revenue planning and inventory forecasting.
Google Performance Max with tROAS configured against AOV. Meta Advantage+ Shopping with first-event optimization. The default configuration the brand had been running for 24 months.
Same campaign structure, same audiences, same creative. ValueBid™ Month-6 retention probability multiplied by 12-month subscription revenue replaced AOV as the conversion value sent to Google OCI and Meta CAPI.
The brand had treated retention as a downstream lifecycle problem for years: better email cadence, better concierge, better re-engagement offers. All of that work mattered. None of it addressed the customer mix entering the funnel. ValueBid™ shifted the mix at acquisition, and the lifecycle team's existing infrastructure compounded against a structurally better cohort. The 60-day churn improvement was the headline metric. The 12-month revenue lift was the actual business outcome.
If your subscription brand's profit lives at Month 6 and beyond, but your auction only sees Box 1, AdZeta's ValueBid™ closes that gap. Predicted retention and subscription revenue delivered to Google OCI and Meta CAPI in real time.