Deep Loyalty: The Quality of Future Profit

Rob Clements | FORCEFWD

Three customers renew. One has compared the alternatives and actively prefers the business. Another has not reconsidered the decision for years. The third was preparing to leave but accepted a substantial discount.

All three are retained. All three may produce valuable revenue. But their renewals give the business different reasons to believe they will return again, and potentially very different economics.

The first provides some evidence of preference. The second may remain for years, although recent behaviour tells us little about what would happen if comparison became easier. The third may represent a profitable save, but that depends on the contribution remaining after the discount and whether further subsidy will be required.

Retention matters. The difficulty is interpreting what it means.

Where continuing customer demand represents a material part of expected revenue or profit, the business needs to understand more than whether customers stayed. It needs to understand the causes of staying, their durability and their economic value.

A retention rate cannot establish those things by itself.

This distinction has a substantial intellectual history. Dick and Basu’s 1994 framework connected customer loyalty with both relative attitude and repeat patronage, recognising that behaviour and preference need to be considered together. The problem did not begin with AI or modern loyalty programmes.

What interests me is how to turn that distinction into a management discipline.

I call my mental model Deep Loyalty. Its governing idea is that loyalty is future-profit engineering: the discipline of converting uncertain future customer demand into demand that is more predictable, durable and economically valuable.

The goal is to give the business better reasons to expect future customer profit, and better evidence for deciding what to invest in to improve it.

That requires two related forms of work. The business must understand the demand it already has, including the uncertainty surrounding it. It must also change the conditions that produce future demand, where doing so creates incremental economic value.

Understanding a customer relationship more accurately does not automatically make it more profitable. A business can forecast declining profit very precisely. Deep Loyalty therefore combines causal diagnosis with profitable intervention. Greater confidence is useful only if it is justified; stronger economics still have to be earned.

Retention is the starting evidence.

Retention diagnosis, economic and durability tests, investment decision and feedback loop

Figure 1. Retention begins the investigation. Causal diagnosis, economic evaluation and durability testing inform investment decisions and expectations of future customer profit.

Why the reason for staying matters

Repeat behaviour is causally ambiguous.

Preference, realised value, familiarity, habit, inertia, contractual constraint, switching friction and subsidy can all contribute to the same observed outcome. Several may operate within one customer relationship.

These causes are not interchangeable.

A customer who stays because the business consistently solves an important problem presents a different exposure from one who stays because comparing alternatives is exhausting. Both may be profitable today. The difference becomes commercially significant when the conditions protecting the relationship change.

That is why I would avoid dividing customers into permanent categories of “loyal” and “disloyal”. Commitment and inertia can coexist. A customer may trust the business, value its service and also appreciate the convenience of doing nothing.

The management task is to understand that mixture.

Inertia can be economically valuable. There is no need to pretend otherwise. But treating untested inertia as evidence of commitment can lead the business to overestimate the resilience of its demand.

Byron Sharp’s How Brands Grow, and subsequent work from the Ehrenberg-Bass Institute, provide an important boundary. Mental availability is the likelihood that a brand is noticed or thought of in buying situations; physical availability, which Sharp also calls purchase availability, is how easy it is to find and buy. Both can support repeat choice without requiring strong commitment. They are commercial assets, not weaknesses for Deep Loyalty to dismiss.

A loyalty strategy should not be used to disguise an availability problem. Deep Loyalty does not assume that deepening existing relationships is the best route to growth. Where being easier to think of and buy offers the stronger return, that is where the investment should go. The task is to understand the causes, durability and economics of continuing demand, and to fund interventions only where the evidence supports a profitable effect.

Longevity also needs separate economic scrutiny. Reinartz and Kumar’s research in a noncontractual direct-marketing setting found a more complicated relationship between customer lifetime and profitability than the assumption that longer relationships necessarily produce higher profits. Their findings were specific to that setting, but they provide a clear reason to test the relationship rather than assume it.

Deep Loyalty asks leadership to keep three judgements distinct: what behaviour has been observed, what probably caused it and what economic value it creates.

The programme is one mechanism

Loyalty programmes can contribute to all three judgements. They can influence behaviour, structure recognition, collect data and make value more visible.

Their effectiveness still needs to be demonstrated.

Consider a programme whose members spend more and remain longer than non-members. That difference may reflect a programme effect. It may also reflect self-selection: customers who already purchase frequently have more reason to join and collect the benefits.

Leenheer and colleagues examined this problem using Dutch grocery data. They found a positive effect of programme membership on share of wallet, but the estimated effect was seven times smaller after accounting for the endogeneity of membership than in a naïve model. That is a finding from one study, not a universal discount factor for programme performance. Its relevance is methodological: who joins matters when estimating what membership causes.

The business should therefore distinguish member value from value caused by membership.

It should also distinguish behavioural effectiveness from economic effectiveness. An intervention may increase frequency or retention while producing an inadequate return after reward costs, service costs, cannibalisation and subsidy.

The programme can be useful. It belongs inside a wider system that governs the reasons customers stay and the economics of influencing them.

The Five Forces of Deep Loyalty

My proposed diagnostic model has five Forces: Organisational Belief, Value Realisation, Friction, Moats and Money.

These are interacting conditions. They are not stages in a journey, and I am not presenting them as a statistically validated index. Their purpose is to organise the questions leadership needs to investigate, assign responsibility for the answers and identify where intervention may improve future customer value.

The Five Forces as interacting conditions around the Deep Loyalty management system

Figure 2. The Five Forces are interacting conditions within one management system. Money supplies the economic discipline for evaluating the system and its interventions.

Organisational Belief

Organisational Belief is what the company’s systems prove it believes customers are.

The evidence sits in targets, incentives, investment decisions, operating policies and tolerated trade-offs. If a business says it values long-term relationships but rewards retention teams solely for gross saves, it may encourage spending that improves retention without improving contribution. If service teams are measured on closing cases, unresolved customer problems may disappear from the dashboard while remaining in the relationship.

The question is whether the operating system supports the customer value the business says it wants to create.

Value Realisation

Value Realisation concerns how promised value becomes experienced value.

A proposition can be persuasive enough to acquire a customer and still fail to deliver the outcome that would give them a reason to return. The business needs to understand when value appears, where it fails and whether it is realised reliably after the initial purchase.

For a software business, that may involve successful adoption and useful outcomes. For a retailer, it may involve availability, product performance, delivery or resolution when something goes wrong. The relevant evidence depends on what the customer came to achieve.

Friction

Friction concerns how effort is removed, shaped or ethically protected.

Some friction is accidental: duplicated work, unnecessary waiting or internal complexity transferred to the customer. Some supports informed decisions, safety or confidence. Other friction protects revenue by making comparison or departure unnecessarily difficult.

These require different decisions. Removing a useful safety check is not equivalent to simplifying cancellation. The business should understand what each form of friction does and whose interests it serves.

Moats

Moats are accumulated reasons to stay that customers would endorse.

Trust, integrated workflows, recognised preferences, community, earned status and a history of good outcomes can make a relationship more valuable over time. The customer has something worth preserving.

A trap protects the relationship through punitive exit conditions. Moats and traps may coexist, so the analysis needs to identify which mechanism is doing the work. A useful test is whether customers recognise the accumulated value and would choose to preserve it if leaving were reasonably straightforward.

Money

Money is the economic control system.

It tests whether funded activity causes profitable incremental behaviour. A material intervention needs a behavioural hypothesis, a credible counterfactual, an appropriate evaluation horizon and a full account of its economics.

For a programme, that includes reward costs, redemption, breakage and associated liabilities. For a service intervention, it may include delivery costs and the contribution from changed purchasing or renewal behaviour. The analysis must avoid counting the same cost twice or confusing accounting timing with economic value.

If the behaviour would have occurred anyway, the expenditure may be subsidy. If the behaviour changes but the incremental contribution does not justify the investment, the intervention may be effective without being economically attractive.

The Five Forces bring these conditions into one conversation. Their value will depend on the quality of the evidence and decisions they produce.

Easier comparison changes the exposure

Agentic commerce gives this model an additional stress test.

AI-assisted shopping already includes tools designed to compare products against customer preferences and constraints. Their limitations matter too: OpenAI’s shopping research documentation acknowledges possible errors in prices and availability, as well as restrictions on automated access to some retailers. Current comparison tools should not be confused with complete, impartial visibility across a market.

My argument is conditional. As agents become more capable and adopted, and where customers delegate decisions to them, they can reduce the effort involved in comparison and switching.

A capable agent with the necessary information and permissions could monitor a renewal, evaluate alternatives and explain the trade-offs. That would reduce some of the protection the existing relationship receives from inattention, confusing value calculations or mild switching effort.

The consequence is exposure. Demand that depended on those protections may become less secure.

This gives us the agent test:

If an AI agent put the alternatives side by side and made switching easy, would you still win on merit?

That is a diagnostic question, not a prediction that agents will always select the cheapest option. Merit means value on the customer’s terms. Trust, continuity, identity, familiarity, accumulated value and confidence may all belong in that judgement, alongside price and speed.

The second question is the human veto:

If an agent found a cheaper or faster alternative, how many customers would tell it, “I still want to stay with them”?

The veto can also be a standing instruction: “Keep choosing them unless there’s a very good reason not to.” Such an instruction would be a useful signal of preference, although it would still need to be compared with subsequent behaviour. It asks whether the customer wants technology to preserve the relationship, even while making alternatives easier to evaluate.

Deep Loyalty should remain useful regardless of how quickly agentic commerce develops. Easier comparison simply makes the distinction between commitment and untested inertia harder to ignore.

What leadership should do with the model

Where continuing customer demand is material, the causes of that demand deserve coordinated leadership attention.

Product, service, pricing, finance, operations, data, technology and brand can all influence whether customers return. The decisions required may exceed the authority of the loyalty team. CEO sponsorship should give the system ownership and resolve conflicts between functions; it does not require the CEO to run the programme.

I would begin with one commercially important customer cohort and one decision the business needs to make.

For example, before increasing renewal incentives, examine the existing saves. Estimate what would have happened without the offer, identify which customers changed their behaviour, calculate the contribution remaining after the intervention and assess whether the effect persists.

Then investigate the Five Forces. Perhaps the offer is compensating for a failure in Value Realisation. Perhaps an operating policy creates avoidable Friction. Perhaps a valuable Moat exists but is poorly understood. Perhaps the incentive is the most economical response. The model should help determine which explanation survives investigation.

Where feasible, use experiments or holdouts to estimate incremental effects. Where they are not feasible, use the strongest available alternative and make the limits of causal inference explicit. Long-term trust and brand effects will not always resolve neatly within a short measurement window; that calls for an appropriate horizon and judgement about uncertainty.

The evidence can then inform investment, operating changes and customer assumptions within the financial model.

The standard I want people to adopt

I define Deep Loyalty as a customer’s durable, economically valuable preference to continue choosing a business, even when alternatives are visible, comparison is honest and switching is reasonably easy.

Not every retained customer needs to meet that definition. Businesses can earn valuable revenue from habit, convenience, occasional purchases and relationships with limited emotional attachment.

The objective is to understand the basis of that value and manage it deliberately.

I want leaders to adopt a habit of asking:

What proportion of our expected future customer value rests on reasons to stay that are understood, economically proven and likely to survive easier comparison?

That question connects customer preference with causal evidence, operating decisions and capital allocation.

A better retention rate may be part of the answer. So may improved contribution, more consistent value realisation or a reduced dependence on subsidy. The relevant result is future customer profit supported by stronger evidence and more durable reasons to recur.

That is what I mean by future-profit engineering.

Frequently asked questions

What do you mean by the quality of future profit?

In this mental model, quality refers to how credible, durable and economically sound expected future customer profit is.

I would examine the contribution expected after relevant costs, the evidence that customers will keep choosing the business, and how vulnerable that contribution is to ongoing subsidy, easier comparison or switching.

For example, two customer cohorts could have the same projected contribution. One may return because the business consistently delivers value; the other may need repeated discounts. Those projections deserve different scrutiny, although discounts can still be a profitable intervention.

This is a management judgement, not a formal accounting measure or a guarantee. Stronger reasons to expect profit do not make future profit certain.

Is Deep Loyalty the same as retention?

No. Retention records who stayed. Deep Loyalty examines why they stayed, whether those reasons are likely to endure and what the relationship contributes economically. Retention can be valuable whether it comes from preference, habit or convenience. The result alone does not establish durable preference or prove that an intervention created incremental profit.

Does Deep Loyalty contradict Byron Sharp’s work on availability?

No. Deep Loyalty is a proposed management discipline, not a competing empirical law of brand growth. Sharp and the Ehrenberg-Bass tradition emphasise mental and physical availability and the importance of reaching more category buyers.

The Double Jeopardy pattern shows that smaller brands typically have fewer buyers who are also slightly less loyal. Brand size and penetration therefore need to be considered before differences in loyalty metrics are credited to a programme or relationship intervention.

Deep Loyalty does not claim that retention-led growth generally outperforms acquisition, or that every customer needs a deep relationship with the brand. It asks which causes of continuing demand can be understood and influenced profitably. Investment in retention still has to justify itself against alternative uses of capital, including availability and acquisition.

Does this mean businesses should abandon loyalty programmes?

No. Programmes can help customers realise value, recognise their relationship with the business and influence behaviour. The question is what the programme causes and whether that effect is profitable after the relevant costs. Member spending, enrolment and redemption are useful observations, but they do not establish incrementality by themselves.

Is Deep Loyalty simply a forecasting exercise?

Better forecasting is one use of the evidence, but the aim is to improve the conditions producing future profit. A business can forecast declining contribution very precisely. Deep Loyalty combines causal diagnosis with profitable intervention: improving the reasons customers return, testing what changes because of the investment and deciding whether the incremental contribution justifies its cost and risk.

How should a business measure Deep Loyalty?

I would begin with a commercially important cohort and a decision, rather than inventing a single loyalty score. Examine retention alongside contribution, the reasons customers give for choosing the business and evidence from their behaviour. Use experiments or credible counterfactuals to test interventions, with an appropriate evaluation horizon. The Five Forces organise the investigation; the conclusions depend on the evidence, its limitations and the assumptions that remain.

What changes when customers use AI agents?

If agents make comparison and switching easier, they can expose demand that relied on inattention, confusing value calculations or mild switching effort. The useful test is what would still win under honest comparison, and whether the customer would instruct their agent to preserve the relationship. That is a conditional stress test, not a claim that agents will always choose the cheapest alternative.

Continue the conversation

To discuss applying Deep Loyalty to your business, visit FORCEFWD.