Two advertising surfaces around payment

On 8 October, Asda and Rokt announced the expansion of personalized offers across two adjacent stages of purchase. Rokt Pay+ operates on the payment page and Rokt Thanks on the post-payment order-confirmation page; the products are used in George’s online clothing and home store and Asda’s online grocery business. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

Asda began with Rokt Thanks at George in January 2025, then added Rokt Pay+ and took both placements into grocery. An algorithm selects an offer in real time; the payment page can sell third-party media while also distributing Asda’s own financial products. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

The companies say adding Pay+ more than doubled revenue from Rokt products at George. Pay+ now supplies 51% of Rokt revenue at George and 36% of combined Rokt revenue across George and grocery. This is not growth in total George or Asda revenue, but a narrow revenue line from Rokt’s commerce-media system. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

Conversion guardrails are a strength, not a complete causal report

Rokt’s case study describes a controlled experiment, pre-agreed guardrails and a gradual traffic ramp. Cart conversion was monitored during expansion; the companies say user conversion rate, or uCCR, remained steady for more than a year in both businesses. [2 · Rokt · Asda case study with measurement window and experiment description] [5 · arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018]

The public evidence is insufficient to reproduce that conclusion. It omits shoppers and transaction counts, randomization method, control share, absolute conversion, minimum detectable effect and confidence interval. These details matter for personalized strategies because dynamic user qualification can break group comparability. [2 · Rokt · Asda case study with measurement window and experiment description] [5 · arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018]

Stable conversion answers only whether fewer measured orders were completed. It does not capture satisfaction, complaints, returns, repeat purchases, retention or customer value. Media revenue must also be compared with Rokt fees, integration cost, possible cannibalization of owned offers and gross profit. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

72% is not Asda growth; it is reported growth in Rokt revenue

Rokt says its revenue in Asda’s grocery business rose 72% year on year over the 30 days from 22 August to 21 September. The vendor attributes this to greater volume and revenue per transaction after both products reached scale. No absolute amount, prior-year base or Asda profit is published. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

The word “incremental” needs care. A new line may not have existed before Pay+, but that does not prove every pound is net-new: responses could have shifted from other offers and long-term customer relationships were not measured. A persistent holdout should cover the order and subsequent behavior. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description] [5 · arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018]

The sources also contain a checkable discrepancy: the joint release and current Rokt case say 72%, while Retail Technology Innovation Hub published 54%. Without the underlying series, readers cannot tell whether this is an editorial error, an earlier estimate or another cut. The 72% therefore remains an attributed company claim. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description] [3 · Retail Technology Innovation Hub · independent report, 8 October 2026]

Expert commentary

The mechanism has a clear advantage: it uses high-intent traffic that already exists. Asda creates media inventory immediately before and after payment instead of buying another visitor. If conversion truly stays unchanged, every additional pound of net margin improves order economics without increasing customer acquisition cost. That matters especially in grocery, where picking and delivery costs leave limited margin headroom. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

The “more than doubled” headline is larger than its denominator. It refers only to Rokt revenue at George, not sales, profit or customer lifetime value. A management decision needs absolute revenue per thousand transactions, Asda’s share after vendor fees, integration costs and the return from the page’s best alternative use. Without those values, its materiality to the whole channel cannot be judged. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description]

A controlled experiment is encouraging: gradual ramp-up and a conversion guardrail reduce the risk of damaging checkout. Yet causal assessment requires known traffic allocation, a stable holdout and adequate statistical power. Those parameters are absent, so “no harm” remains the companies’ internal conclusion. The 72% versus 54% discrepancy also shows why metric definitions and dataset versions should be published. [2 · Rokt · Asda case study with measurement window and experiment description] [5 · arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018]

For customer relationships, the acceptability of personalization is pivotal. Experiments on ad transparency find that acceptable data use on a trusted platform can support response, whereas unexpected transfers or inferences can reduce effectiveness. Checkout needs a clear reason for the offer, an identifiable advertiser and an easy way to dismiss it. That turns trust from an abstract value into a testable interface property. [2 · Rokt · Asda case study with measurement window and experiment description] [4 · Journal of Consumer Research · ad transparency and effectiveness, 2019]

Financial offers deserve special caution because payment combines high intent with limited attention. Short-term revenue can coexist with complaints, accidental consent or perceived pressure even when an order completes. Opt-outs, support contacts, cancellations, returns and differences for vulnerable groups should be measured separately. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description] [4 · Journal of Consumer Research · ad transparency and effectiveness, 2019]

Over the next 6–12 months, the persuasive outcome would be a metric bundle: incremental gross profit per transaction, conversion and payment errors, repeat purchase, retention, complaints, adoption of owned financial products and a persistent holdout effect. Until then, this is a promising monetization channel, not proof of a better customer relationship. [1 · Asda and Rokt · joint results announcement, 8 October 2026] [2 · Rokt · Asda case study with measurement window and experiment description] [4 · Journal of Consumer Research · ad transparency and effectiveness, 2019] [5 · arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018]

Sources

  1. Asda and Rokt · joint results announcement, 8 October 2026 — Primary announcement covering Rokt Pay+, Rokt Thanks, Rokt revenue growth, Pay+ share and the claim of no conversion decline. It is a Rokt press release containing Asda comments.
  2. Rokt · Asda case study with measurement window and experiment description — The vendor case specifies 22 August–21 September, more than a year of conversion monitoring and a controlled experiment with gradual traffic ramp-up. It omits absolute values, sample size, group allocation and intervals.
  3. Retail Technology Innovation Hub · independent report, 8 October 2026 — The trade report corroborates the deployment but gives grocery Rokt revenue growth as 54%, while current primary materials say 72%; the discrepancy cannot be resolved without source data.
  4. Journal of Consumer Research · ad transparency and effectiveness, 2019 — Peer-reviewed experiments show that response to personalization depends on the acceptability of information flows and trust in the platform. The study does not test Asda or Rokt.
  5. arXiv · Online Controlled Experiments for Personalised E-Commerce Strategies, 2018 — Methods paper on causal evaluation of personalized strategies and pitfalls from dynamic user qualification. It is not an audit of the Asda experiment.