Customer acquisition costs in e-commerce have followed a consistent trajectory over the last decade: upward, with periodic steepening that correlates with iOS privacy changes, auction market dynamics in paid social, and the maturation of the consumer cohorts that were cheaply acquired during earlier platform growth phases. The response from the majority of e-commerce operators has been to treat this as a unit economics problem — work the CAC/LTV ratio, optimise paid channels, invest in organic where margins permit. That framing is correct but insufficient. The companies that are building durable commerce positions are increasingly doing so by shifting the centre of gravity from acquisition to retention, and loyalty infrastructure is the architecture that makes retention economics legible and optimisable.
The legacy loyalty model — points-per-dollar, tiered status, redemption for discounts — has a structural problem that most retailers haven't fully reckoned with. It creates a financial liability (outstanding points balances), it rewards transaction volume without regard to margin, and it teaches customers to defer purchases until they can redeem a discount, which trains price sensitivity rather than brand affinity. That's a bad programme design problem masquerading as a loyalty problem, and it explains a lot of the scepticism about loyalty ROI that you hear from retail CFOs. What that scepticism misses is that the failure is in programme design and data architecture, not in the underlying proposition that retained customers are more valuable than acquired ones — which is consistently and empirically true across retail categories.
Clutch, which we backed in 2024, is building loyalty infrastructure at the platform level rather than as an end-to-end loyalty programme. The distinction matters: a loyalty infrastructure layer is API-accessible and allows commerce platforms to build their own programme logic on top of clean primitives for points issuance, tier management, reward fulfilment, and loyalty-linked segmentation. That's a different product than a white-label loyalty programme — it's infrastructure that makes the commerce platform the operator of their own loyalty experience rather than a licensee of someone else's. The implication is that the programme logic can be deeply integrated with the platform's existing transaction data, inventory state, and customer profile — which is where the genuinely interesting loyalty personalisation lives.
The APAC-specific dimension of loyalty infrastructure is significant. Consumer loyalty behaviour varies materially across APAC markets in ways that aren't obvious from a Western-market baseline. Loyalty programmes that drive high engagement in Australia through digital wallet integration and real-time reward visibility may perform very differently in markets where WeChat ecosystem integration or alipay-linked rewards are the channels consumers actually use. Building loyalty infrastructure that is flexible enough to handle that kind of channel and consumer behaviour variation — while maintaining a consistent underlying data model — is a meaningfully harder product problem than building for a single-market assumption.
The metric we watch most closely in loyalty is not redemption rate — which is a lagging indicator of programme engagement — but the degree to which loyalty programme membership correlates with purchasing frequency and basket size changes over time. A loyalty programme that is genuinely working for the retailer should show an increasing divergence between the behaviour of programme members and non-members over a 12–24 month cohort horizon. Retailers who can show that divergence have built something that compounds. Retailers who can't are running an expensive discount mechanism. The infrastructure question is what data architecture and programme design enables the former.