Why retail’s most powerful commercial asset is being used as a media channel — and what the leaders are doing instead
For retail CEOs, CMOs, and CTOs · May 2026
IN BRIEF
- Most retail CRM operates as a media channel — pushing campaigns, promotions, and supplier-funded deals — when it should function as a business intelligence system with a communication layer on top.
- The gap is not a technology failure or a data failure. It is structural: CRM sits in marketing, merchandising sits elsewhere, supplier co-funding distorts incentives, and short-term metrics make the damage invisible.
- The commercial cost is measurable. The media channel model delivers a net ROI of 0.8–1.4× once co-funding and baseline margin erosion are stripped out. The BI system model delivers 2.8–5× at maturity, with +80 to +200 bps margin impact over three years.
- Break-even on the BI model takes 14–36 months depending on data maturity — not the years most assume. Retailers with strong data infrastructure already in place can reach break-even in 14–18 months.
- Four retailers illustrate the spectrum: Kroger (completed transition, $1B+ data revenue), Tesco (original blueprint, cautionary tale included), Amazon (the asymmetric benchmark), and Carrefour (live transition — the most relevant mirror for most legacy retailers).
- Three structural decisions separate the leaders: separating the BI engine from campaign execution, giving merchandising teams formal data access, and monetizing the data externally.
- After 3–5 years, the data asset compounds into a moat that a competitor cannot close with investment alone. The time to act is before the gap becomes visible — by then it is already too late.
- The diagnostic question every executive team should ask: are we using CRM to change what we know about our customers — or only to send them messages?
CONTENTS
01 The paradox
02 Why the inferior model persists
03 The quantified cost
04 The ROI timeline: three horizons, not one
05 What the leaders actually do
06 The three decisions that separate them
07 The compounding moat
08 The leadership question
09 A practical decision framework: the quick win model
10 The conclusion
THE PARADOX
There is a contradiction at the heart of modern retail CRM. Every chief executive intuitively understands that their loyalty data is a strategic asset. Every chief marketing officer can articulate the theory: identify at-risk customers before they leave, engage shoppers in categories under pressure, develop high-value customers to spend incrementally more. Every chief technology officer has signed off on significant platform investment to make this possible.
And yet, in practice, the overwhelming majority of CRM effort in retail is consumed by something entirely different: launching products, pushing promotions, serving supplier-funded deals, and measuring redemption rates.
CRM is being used as a media channel when it should be operating as a business intelligence system with a communication layer on top.
This is not a technology failure. The platforms exist. It is not a data failure. The transaction data is there. It is a structural and organizational failure — one with a measurable commercial cost that rarely appears on any dashboard, because the damage accumulates slowly while the short-term metrics look reassuring.
This article diagnoses the paradox, quantifies what it costs, and shows what the retailers who have escaped it have done differently.
WHY THE INFERIOR MODEL PERSISTS
The drift toward campaigns, deals, and launches is not accidental. It has four structural causes that reinforce each other:
- Organizational ownership. CRM sits in marketing. Merchandising sits elsewhere. The two functions rarely share KPIs, and the data does not formally flow between them. The intelligence that should be governing range decisions, space allocation, and pricing strategy stays inside a campaign tool.
- Short-term earnings pressure. Weekly and quarterly targets push CRM teams to activate volume now. A churn-prevention program that delivers its return over two years cannot compete for budget against a promotional campaign that moves the dial next Tuesday.
- Supplier co-funding. A significant share of CRM budgets is funded by brands. Brands want activation and visibility, not churn analytics. Co-funding makes the media channel model look profitable at the campaign level, masking the true cost at the portfolio level.
- Measurement failure. Redemption rates and short-term attributed sales are easy to report. True incrementality — what the customer would have done anyway, stripped from what the CRM actually caused — is harder to calculate and uncomfortable to present. The result: CRM dashboards systematically overstate ROI.
THE QUANTIFIED COST
The commercial gap between the two operating models is large, and it widens over time. The table below summarises benchmarked ranges drawn from four primary sources: dunnhumby’s loyalty economics research (2019–2023); McKinsey’s Retail Practice publications on customer lifetime value and CRM incrementality (2020–2024); BCG’s “Personalisation at Scale” series (2021–2023); and Bain & Company’s retail loyalty and pricing research (2018–2023). Case-level outcomes for Tesco, Kroger, Amazon, and Carrefour are drawn from published annual reports, investor presentations, and documented analyst commentary for the periods stated.
| Metric | CRM as media channel | CRM as BI system |
| Net ROI multiple | 0.8–1.4× | 2.8–5× |
| True incrementality | 20–35% | 55–75% |
| Margin impact (3 years) | −150 to −300 bps | +80 to +200 bps |
| Customer lifetime value trend | Flat or declining | +15–30% over 3 years |
| Churn rate trajectory | Flat or worsening | −15% to −30% |
| Supplier funding dependency | High — masks true ROI | Low — fully independent |
| Time to first visible ROI | 2–6 weeks | 9–18 months |
SOURCE NOTES
Net ROI multiple & true incrementality: McKinsey & Company, “The value of getting personalisation right — or wrong — is multiplying” (2021); dunnhumby, “The Retailer Preference Index” annual series (2019–2023).
Margin impact & baseline erosion: Bain & Company, “Loyalty Insights” retail series (2018, 2022); Bain & Company, “The future of retail promotions” (2023). The 150–300 bps estimate reflects grocery retail across Western European and North American markets over a 3-year promotional conditioning window.
Customer lifetime value trend: BCG, “Personalisation at Scale” (2021, 2023); McKinsey, “Next in Personalisation 2021”. CLV uplift ranges reflect retailers operating mature BI-led CRM models for 3+ years versus campaign-led peers in the same market.
Churn reduction: dunnhumby, “Customer Centricity” white paper series (2020–2022); Harvard Business Review, “The Value of Keeping the Right Customers” (Reichheld). The 5–7x cost-to-retain vs. cost-to-acquire ratio is a widely replicated finding across grocery, DIY, and general merchandise verticals.
Supplier co-funding dependency: IGD, “Retail Media & Shopper Marketing” (2022–2023); PwC, “Retail & Consumer Report” (2023). Co-funding recovery estimates of 25–40% of gross CRM cost are based on disclosed trade terms structures across major European and US grocery operators.
Time to ROI: Directional estimates based on Hypertrade deployment experience across 15+ markets and corroborated by BCG “Personalisation at Scale” (2023) implementation timelines. Individual outcomes will vary by data maturity, organisational readiness, and market context.
The most consequential number in that table is not the ROI multiple. It is the margin impact over three years. Systematic discounting through a media-channel CRM trains the customer base to wait for deals. Bain’s retail loyalty research estimates this erodes baseline gross margin by 150 to 300 basis points over a three-year promotional conditioning window in grocery retail — a finding replicated across Western European and North American markets. This damage never appears in a campaign ROI report. It accumulates invisibly, in the gap between the price customers are willing to pay and the price the retailer has conditioned them to expect.
THE ROI TIMELINE: THREE HORIZONS, NOT ONE
The most common internal objection to the BI model is the payback period. The media channel model appears to return value in weeks; the BI model takes years. This comparison is structurally false — but it is also incomplete. The BI model does not produce a single ROI event. It produces three distinct return horizons, each from a different part of the system, each building on the last.
| Horizon | Period | What is happening | Net ROI | The implication |
| Horizon 1 Quick wins | Months 6–12 | Churn identification & first interventions. Category reactivation of lapsed high-margin buyers. Suppression of offers to customers who would have purchased anyway. | 1.5–2× on specific programs running | Not yet transformational — but sufficient to fund the next phase internally and demonstrate proof of concept to the board. |
| Horizon 2 Structural return | Months 18–36 | CLV improvement measurable across cohorts. Merchandising decisions informed by CRM data showing +80 to +150 bps margin improvement. True incrementality rising to 55–65%. | 2–3× on full CRM investment | The point at which the investment case becomes self-evident. Requires patience — and an organizational structure that does not harvest the short-term gains before the BI layer is built. |
| Horizon 3 Compounding return | Years 3–5 | Data asset 3+ years deep. Models trained on real intervention outcomes. Merchandising integration embedded in commercial calendar. External monetization through retail media generating incremental high-margin revenue. | 3–5× net, with trajectory still improving | The structural moat. Marginal cost of each intelligence cycle falls while output quality rises. A competitor starting from zero at this point cannot close the gap with investment alone. |
| Break-even by starting point | Data maturity | Cumulative break-even |
| Strong infrastructure already in place | CDP live, clean transaction data, some existing segmentation | 14–18 months |
| Moderate maturity | CDP exists but static; some modeling; limited merchandising link | 20–26 months |
| Low data maturity | Campaign tools only; no CLV scoring; no formal BI layer | 28–36 months |
Two qualifications apply. First, these timelines assume organizational change happens in parallel with the technology build — in practice it almost never does. When org change lags, add six to twelve months to each horizon. Second, and more fundamentally: the media channel model’s apparent payback of two to six weeks is an accounting illusion built on co-funding recovery, inflated attribution, and ignored margin damage. The real comparison is not eighteen to twenty-four months versus six weeks. It is eighteen to twenty-four months to genuine positive return, versus accelerating damage that never appears on a dashboard.
WHAT THE LEADERS ACTUALLY DO
A small number of retailers have run both models simultaneously — maintaining short-term campaign performance while building the intelligence system underneath it. Four cases are worth examining in detail: three that have completed meaningful stages of the transition, and one that is mid-journey and all the more instructive for it.
Kroger — the most complete case
Kroger created 84.51° as a wholly owned data science subsidiary in 2015, structurally separating the intelligence engine from campaign execution. The media channel — Kroger Precision Marketing — sells targeted audience access to CPG brands and generates over one billion dollars annually. The BI system feeds category managers, store operations, and pricing teams with customer intelligence that is entirely invisible to the media operation. The result is that Kroger’s customer data has become a product. CPG brands now treat access to it as a must-buy, not an optional media spend. What was once a cost centre is now a revenue line.
Tesco — the original blueprint
The Clubcard program launched in 1995 is the foundational case. In the decade that followed, Tesco grew UK grocery market share from 19% to 31% — the largest sustained share gain in modern British retail history. The mechanism was not the vouchers. It was that basket data restructured entire category strategies, store formats, and the tiered range architecture that competitors were copying blind. By the time Sainsbury’s or Asda understood why a particular value tier was winning, Tesco had already moved to the next insight cycle.
The cautionary note: when financial pressure hit post-2012, Tesco harvested the short-term media model too aggressively. The Clubcard became a discount mechanism rather than an intelligence engine. Market share followed. The lesson runs in both directions.
Amazon — the asymmetric case
Amazon never separated the two models. Every customer interaction is simultaneously a media moment and a data capture event feeding the intelligence system. Prime membership retains 93% of members after year one and 98% after year two. The advertising business generates over forty-five billion dollars annually — almost entirely margin, built on data superiority. Private label strategy, assortment decisions, and pricing are all downstream outputs of the same customer intelligence. Their competitive moat is not logistics. It is that the BI system improves with every transaction, compounding a data advantage that widens every year.
Carrefour — the most instructive live case
Carrefour is deliberately included here not as a success story, but as the most relevant case in progress — and the most honest illustration of what the transition looks like from inside a legacy hypermarket operator. Under Alexandre Bompard since 2018, the intent has been made explicit: Carrefour Links monetizes customer data externally with CPG partners, internal CRM is formally expected to feed category teams, and AI investment signals the BI architecture being built underneath. The Carrefour+ loyalty program has over ten million active members in France. Retail media revenue is targeting €200 million annually by 2026. The ambition is credible and the early gains are real. The honest assessment is that Carrefour is approximately five to seven years behind Kroger.
What makes it the most instructive case is precisely what it has not yet completed: formal CRM-to-merchandising integration at scale, the shift away from supplier co-funding dependency, and the organizational realignment that gives commercial teams shared accountability for customer outcomes. These are not technology gaps. They are the same structural barriers every legacy retailer faces — visible here because Carrefour is far enough into the transition to have encountered them.
For most traditional retailers reading this article, Carrefour is the closest mirror. The gap between ambition and execution it reflects is not a failure of leadership. It is an accurate picture of how hard the organizational change actually is — and how long it takes even when the strategic will is present.
THE THREE DECISIONS THAT SEPARATE THEM
Across all four cases — and most clearly in the contrast between Kroger’s completed transition and Carrefour’s live one — the pattern is consistent. Every retailer that has successfully made this shift shared three structural choices that their competitors did not make:
- They separated the BI engine from campaign execution. Data science and customer intelligence operate independently of the marketing calendar. The insights are not hostage to this month’s campaign plan.
- They gave merchandising teams direct, formal data access. CRM intelligence formally inputs into range planning, space allocation, and category strategy. This is where the majority of the margin opportunity lives, and it is the step most retailers skip entirely.
- They monetized the data externally. Once the capability is mature, the customer intelligence becomes a product. A retail media network priced on data quality and targeting precision transforms CRM from a cost centre into a revenue line, changing the entire investment calculus.
THE COMPOUNDING MOAT
The most important strategic implication of this framework is not the ROI multiple in year one. It is what happens in year five.
After three to five years of the BI operating model, the data asset is so rich and the organizational capability so embedded that a competitor starting from zero cannot close the gap with investment alone. They would need years of transaction history they simply do not have.
The retailers operating on annual range reviews and quarterly sales data are always reacting. The retailers with a functioning BI system are always anticipating. That gap compounds. The time to start is not when the competitive pressure becomes visible. By then, the window has closed.
THE LEADERSHIP QUESTION
This is not a problem any single function can solve. The CMO cannot fix organizational silos alone. The CTO cannot build a BI system whose outputs no commercial team is accountable for acting on. The CEO cannot mandate data-driven decisions without restructuring the incentives that currently reward short-term campaign volume.
The conversation that needs to happen at board and executive committee level is a simple one, with a difficult answer:
THE DIAGNOSTIC QUESTION
Are we using CRM to change what we know about our customers and our categories — or are we using it to send them messages? If the honest answer is the latter, the data asset is depreciating, not compounding. And every quarter that passes without addressing the organizational structure makes the transition harder and the competitive gap wider.
A PRACTICAL DECISION FRAMEWORK: THE QUICK WIN MODEL
For retailers who want to begin operating the BI model without dismantling their existing campaign infrastructure, a useful starting point is the Quick Win Campaign architecture. It illustrates concretely what “intelligence-led communication” looks like in practice.
The logic operates as a five-layer sequential pipeline. A campaign is only proposed when all five layers produce a valid output:
| Layer | Decision layer | What it does |
| 01 | Segment isolation | Identify Basket Builders — customers purchasing across 4+ categories who anchor store economics through visit regularity and category breadth. |
| 02 | Fatigue guardrail | Apply a dynamic suppression window calibrated to each customer’s own purchase rhythm, not a blunt fixed interval. |
| 03 | Category prioritisation | Focus effort on categories with high strategic value and declining sell-out — where intervention is commercially necessary. |
| 04 | Hero SKU identification | Find the specific item that is at early risk of lapse — not the globally best-selling SKU, but the right SKU for this customer. |
| 05 | NBO offer selection | Let a predictive model choose the offer mechanic — replenishment reminder, volume discount, or bundle — based on what the customer will actually respond to. |
The critical design principle is what this framework deliberately avoids. It does not promote items the customer would have bought anyway. It does not push the highest-selling SKU globally. It does not default to a discount. Every layer is designed to ensure the intervention is genuinely necessary, genuinely personal, and genuinely incremental.
That is the difference between a media channel and an intelligence system: one optimises for message delivery, the other optimises for commercial outcome.
THE CONCLUSION
The CRM paradox is real, well-documented, and commercially costly. It persists not because retail leaders lack awareness of it, but because the organizational structures, incentives, and funding models that sustain the media channel approach are deeply embedded.
Closing the gap requires three things: structural separation of intelligence from execution, formal data-sharing between CRM and merchandising, and a long enough time horizon to let the compounding work. None of these are technology problems. They are leadership decisions.
The retailers who make them will not see the payoff in the next quarter’s campaign report. They will see it in three years, when their competitors are still reacting to customer behaviour they are already anticipating.
ABOUT HYPERTRADE
Retail practitioners. Not platform vendors.
We wrote this article because we see the CRM paradox in almost every retail market we operate in. Hypertrade is a retail intelligence company founded by a former Carrefour executive, with over two decades of operating experience across Southeast Asia, the Middle East, and Africa — not observed from the outside, but built from within.
Ariane RDS — our decision engine — is the practical answer to the BI system model described in this article. It connects loyalty and transaction data to commercial decisions across assortment, promotions, pricing, and CRM campaign execution, with financial value attached to every recommended action. When a commercial decision is approved, the CRM campaign brief is generated automatically. The communication layer follows the intelligence. Not the other way around.
| Live demo on your data We run Ariane on your own transaction data and show you exactly what your commercial team should be deciding — and what each decision is worth. No generic walkthrough. Your categories, your SKUs, your numbers. 45 minutes · No commitment · hypertrade.ai | CRM diagnostic session A structured conversation using the five-dimension framework in this article to map where your CRM operation currently sits, where the largest commercial gaps are, and what a realistic transition roadmap looks like for your specific context. For CEOs, CMOs, and CTOs · hypertrade.ai |
About this article
This article draws on industry benchmarks from dunnhumby, McKinsey, BCG, and Bain retail practice research, and on documented case outcomes from Tesco, Kroger, Amazon, and Carrefour. It was developed as a strategic framework for retail executives navigating the transition from campaign-led to intelligence-led CRM.
