Treating Acquisitions and Pricing with Separate P&Ls is Where Value Leaks

What to Read Next

1. Gieni ABX Brings Autonomous Organisational Health Checks to Salesforce Environments

2. Mod Op Unveils Connected Intelligence Platform ORION

3. Onton Debuts AI Model for Trustworthy Product Discovery

4. FreeWheel Launches Video Content Report

Customer acquisition cost (CAC) has been one of the defining metrics of marketing performance. Lower CAC, higher ROAS and cheaper conversions became the benchmarks by which growth teams judged success.

Digital businesses are beginning to realise that efficient acquisition alone does not create sustainable growth. Customers acquired at the lowest cost are not always the customers who generate the highest lifetime value (LTV), contribute the most gross profit or remain loyal over time.

At the same time, artificial intelligence is transforming how businesses understand customer behaviour. Predictive models are allowing organisations to move beyond historical averages and static cohort analysis, making LTV a real-time operational signal rather than a backward-looking reporting metric.

“Of course, CAC / CPA discipline matters, capital efficiency is real, and it can become truly expensive if it is not in place. But the question should never only be “how cheaply can we acquire?” Rather, it should have been “how cheaply can we acquire someone worth keeping?” says Álvaro Martínez, Sr. Director MENA – Growth Marketing and Pricing, talabat.

In this interview, Álvaro Martínez discusses why growth marketing and pricing should work as one system, why acquisition metrics need to evolve beyond CPA, how businesses should rethink LTV:CAC, why customer frequency matters more than basket size, and how AI will make lifetime value actionable at an individual level.

Excerpts from the interview;

How does dual ownership of growth marketing and pricing shape the way you think about CAC and LTV?

Most organisations treat acquisition and pricing as separate functions with separate P&Ls. That separation is actually where value leaks. When you own both, you stop optimising them independently and start treating them as a system.

Concretely, it changes the way I frame CAC. If I know that a cohort acquired on a promotional price point has an 18-point lower retention rate than one acquired at full price, then the headline CPA is almost meaningless; it’s a vanity number. 

The relevant question becomes: what is the fully-loaded cost of acquiring a customer who actually stays?

• CAC becomes a cohort-quality signal, not just a channel efficiency signal (i wouldn’t mind to pay EUR 40 CPA if the acquired user 3 month LTV was EUR 300, so it comes down to a matter of unit economics, not a siloed metric)
• Pricing incentive decisions get evaluated like-for-like against other peer incentives. The efficiency game should consider the best of all options for driving long-term sustainable growth
• Budget allocation shifts toward channels and offers that produce durable GMV, not first-order volume (here, it is also key to understand what your adoption/retention strategy looks like)

Dual ownership forces the team to hold two truths simultaneously; we need to grow the base, and we need to grow it with the right base. That discipline is hard to replicate when the two functions are siloed.

Have marketers over-indexed on acquisition efficiency at the expense of long-term value?

Yes. And I’d go further: much of the industry built its measurement infrastructure around this bias, making it structural, not just behavioural. The question is always: what channel drives the cheapest CPA versus which channel drives the best user cohort at an acceptable cost. Small semantics, big difference.

Last-click attribution, platform-reported ROAS, and weekly CAC targets all optimise for the moment of acquisition. They don’t capture what happens in months two through twelve. 

The result is that growth teams became very good at producing first orders cheaply and very bad at knowing whether those orders were worth producing. It is certainly an uncomfortable shift.

• At the category scale, the damage compounds: aggressive CAC targets pull spend toward high-volume, low-quality segments (retaining low quality cohorts becomes a long-term open wound)
• Incentive structures reinforce this; many growth teams are bonused on new users, not on retained cohort GMV
• It is normally misunderstood by finance teams, hence growth and marketing leaders need to take ownership and accountability to knowledge share the best way to measuring success of this

Of course, CAC / CPA discipline matters, capital efficiency is real, and it can become truly expensive if it is not in place. But the question should never only be “how cheaply can we acquire?” Rather, it should have been “how cheaply can we acquire someone worth keeping?” Those are different problems.

Ultimately, the fix isn’t abandoning CAC. It’s pairing it with an x-day cohort GMV or repeat rate threshold, so acquisition efficiency is always evaluated against retention quality.

Is the LTV:CAC ratio still a reliable benchmark, or does it oversimplify modern digital businesses?

It can be a great benchmark for true performance, but it all depends on which LTV period benchmark is considered and how it is calculated. The core problem is that LTV is usually a forecast, and most teams compute this forecast based on historical averages rather than the most recent cohort-specific curves. 

A 3:1 LTV:CAC ratio tells you very little if the LTV is calculated on a three-year average that includes customers acquired in a completely different competitive environment or at a different price point. Hence, two variables which are key to getting it right:

• Var1: GP / GMV / inc. GMV-based LTV
• Var 2: LTV baseline should consider comparable periods to ensure comparability

Additionally:

• In markets with high promotional density (which describes most of MENA food delivery), churn curves are non-linear. Averaging them flattens the signal and can neglect structural issues.
• The ratio conflates margin and revenue. A 3:1 on gross revenue in a low-margin category is structurally different from a 3:1 on contribution margin (see Var 1 mentioned above)
• It ignores the time-value dimension entirely; a 3:1 over 9 months is a much worse business than a 3:1 over 3

It is good to use LTV:CAC as a direction-setting ratio, not a performance target. The operational target should be cohort-level contribution margin at 90 and 180 days, preferably on GP and within cohorts at comparable periods.

Beyond acquisition, what are the biggest drivers of LTV?

In our context, food delivery, a high-frequency category in MENA, the answer is order frequency, not basket size and not retention in isolation. We are playing the frequency game.

Basket size has a ceiling driven by meal occasion economics. Retention is partly a function of frequency. But frequency is the variable that compounds. 

For example, a customer who orders 2.5x per month vs. 1.5x per month is not 67% more valuable over one year; they’re structurally more valuable because they’re also more likely to retain, more likely to try new verticals, and more likely to respond to loyalty mechanics.

The drivers that actually move frequency:

• Habit formation in the first 30 days: the cohort data consistently show that ordering behaviour in weeks one to four is the strongest predictor of 6-month LTV. Activation, not acquisition, is the leverage point. Investment here is critical.
• Occasion breadth: customers who use the platform across multiple day-parts (lunch and dinner, not just dinner) show materially higher frequency
• Category expansion: a customer who adds grocery or convenience to their food ordering has a retention curve that looks fundamentally different from a food-only user

Basket size matters, but it’s largely driven by occasion type and restaurant selection, which are harder to move directly. Frequency is more actionable. If you’re trying to increase LTV, the highest-leverage investment is in the 30-day activation journey, not in the acquisition funnel.

What role will AI and predictive modelling play in improving how we understand and optimise LTV?

It will shift LTV from a backward-looking cohort average to a forward-looking individual signal, which changes what you can do with it operationally.

Currently, most LTV models produce segment-level outputs: “Customers acquired via channel x in market y have an average 6-month LTV of z.” That’s useful for budget allocation but not for real-time intervention. Predictive modelling at the individual level changes the decision surface.

• Churn prediction becomes preventive rather than reactive; you intervene at day x when the signal appears, not at day y when the customer has disengaged
• NBA (Next Best Action) engines can use predicted LTV as a weighting variable, so high-LTV-potential customers receive disproportionate investment in their early lifecycle
• Pricing personalisation can be calibrated against individual LTV curves, you don’t need to offer a blanket discount to re-engage; you can size the incentive to the expected return

The honest constraint is data quality and model refresh cadence. A predictive LTV model that runs monthly is already partially stale in a category with daily/weekly ordering cycles. The real unlock is when predictions update in near-real-time and feed directly into CRM and paid media decisioning.

AI doesn’t make LTV a better metric; it makes LTV actionable at a granularity that was previously impossible. The teams that will win are the ones that close the loop between the prediction and the intervention, not just the ones that build the model.

ALSO READ: How Full-Funnel Retail Media is Starting to Take Shape in Practice