Customer acquisition cost across telecom channels

Pull an operator's acquisition budget apart and it sits in four ledgers that rarely meet: a media plan owned by marketing, a commission schedule owned by sales ops, a property line owned by retail, and a credit line that finance books as a revenue deduction. Only the first carries the word marketing. Divide that one by activations and you get a number that can be a third of what the subscriber actually cost, which is how an operator ends up defending a channel that loses money on every line it opens.

Media-only cost against loaded cost

Media-only CACCACCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → is ad spend divided by activations. Loaded CAC adds everything you pay only because that subscriber arrived:

• dealer and agent commissions, plus volume spiffs and any residual on the line
• allocated retail cost: lease, rates, staff hours, fit-out amortisation
• device subsidy, the gap between wholesale handset cost and the upfront price paid
• acquisition credits: bill credits, free months, port-in bonuses, referral payouts
• onboarding: SIM or eSIM logistics, credit check fees, porting charges, the first support contact

Loaded CAC = (media + commissions + allocated channel overhead
              + subsidies + acquisition credits + onboarding)
             / (activations that survive the qualifying window)

Two choices inside that formula move the answer more than any efficiency work: what you allocate to acquisition, and what you count in the denominator. Publicly cited fully-loaded estimates for US postpaid owned retail sit in the low hundreds of dollars per subscriber (estimates; carriers do not disclose the line). Before you compare your figure to anyone's published one, the definitions have to match, which is a separate exercise the benchmarking lesson handles.

Four channels, four cost shapes

Dealer and agent commission. Purely variable. No activation, no cost, which makes it the simplest channel to model and the hardest to police. The commission is rarely one figure: an upfront per-activation payment, a tier bonus for hitting monthly volume, sometimes a trail on revenue for the first year. Against it sits a clawback window, commonly 90 to 180 days in dealer agreements: if the line dies inside it, the dealer repays. Dealer CAC for a cohort is therefore provisional until the window closes, and any dashboard reading it at month end is reading a draft. The second-order effect matters more. Pay on gross adds and you buy gross adds, including lines a dealer activates against its own float and keeps breathing until the clawback date passes.

Retail footprint. Almost entirely fixed. The store costs the same in a slow month, so CAC per activation is a throughput function: halve the activations and you double the CAC without anything changing in the marketing. It is the only channel whose unit cost gets worse when demand falls, which makes closure decisions self-reinforcing, since footfall does not transfer cleanly to the surviving stores. Mint Mobile built its proposition on deleting this line entirely: an MVNO riding T-Mobile's network, selling three, six and twelve-month prepaid blocks paid upfront, no owned estate. T-Mobile bought its parent, Ka'ena Corporation, with the deal closing in 2024.

Referral. giffgaff, the UK MVNO on O2, pays existing members to recruit new ones (a credit of a few pounds once the new SIM is topped up) and pays members "payback" for answering support questions in its community forum. Two lines of loaded CAC fall at once: the payout is a success fee, so waste is near zero, and community support cuts the onboarding contact cost that most CAC models never itemise. The ceiling is arithmetic, member base times participation rate, so referral cannot carry a growth quarter alone. The failure mode is cannibalisation: some share of referred joiners were coming anyway, and each one converts a free acquisition into a paid one. The only honest read is a holdout, withholding the referral prompt from a random slice of the base and measuring the gap.

Celebrity-led direct. Ryan Reynolds took an ownership stake in Mint Mobile in 2019 and fronted the advertising himself, with his agency Maximum Effort producing spots at a fraction of conventional broadcast budgets. Two accounting consequences. His compensation sat in equity, so it never touched the marketing ledger at all; media-only CAC looked excellent and the real cost showed up as dilution at exit. And brand-led direct spend is fixed within a flight: it does not scale with volume, so CAC falls as the work lands. That inverts the arbitration. With dealers, more volume costs proportionally more. With brand-led direct, more volume is how you make the number good, and the exposure is attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition →, because you cannot A/B testA/B testA/B testing is a controlled experiment that compares two versions of something (A and B) by splitting traffic randomly to learn which performs better on a chosen metric.View full definition → a national campaign and the organic baseline you subtract decides the answer.

Where the loaded number diverges

A worked allocation

Take one store costing 38,000 a month in lease, rates, staff and fit-out amortisation, handling 420 counter transactions of which 180 are new activations. The rest are bill payments, SIM swaps, repairs and retention saves.

• Charge the whole store to acquisition: 38,000 / 180 = 211 per activation.
• Charge the 55% of floor time that a staff-hour sample attributes to acquisition: 20,900 / 180 = 116.

Add 60 of average handset subsidy across that mix, 15 of porting and onboarding cost, and 25 of digital media that drove the visit. Loaded CAC is 216 against a media-only 25. Same store, same month, and the number the board sees depends on the allocation rule alone. Write the rule down once, and do not change it mid-year without restating the history, or every year-on-year CAC comparison becomes an argument about accounting.

What prepaid distribution does to the denominator

Safaricom sells through a Kenyan network of hundreds of thousands of outlets, mostly M-Pesa agents who move airtime, SIMs and cash in the same transaction. Commission per SIM is small in absolute terms, but so is prepaid ARPU: a few dollars a month, not fifty. There is no room for a handset subsidy in the same equation, which is why device affordability arrives as a separate financed product, Safaricom's Lipa Mdogo Mdogo daily-payment plan, rather than as an acquisition cost buried in CAC.

The harder problem is the denominator. A SIM sold is not a subscriber acquired. Multi-SIMming is normal across East Africa, so a "new line" is often an existing person's second or third SIM, bought for a promotion and dormant within weeks. Kenya's registration rules add ID verification at the point of sale, which raises the per-activation compliance cost and cuts fake registrations, but it does not stop the same human buying four SIMs legitimately. If commission pays on registration with no activity threshold, the scheme funds dormancy directly. The workable definition counts SIMs still revenue-generating at 30 or 90 days, and gross SIM sales can overstate that by a wide margin. Whether the resulting cost clears is a question of the payback horizon the LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → lesson models, not of CAC on its own. For sector-level structural context on operator economics, see the GSMA's Mobile Economy reports.

Blended CAC and the mix decision

Blended CAC is a weighted average, so it moves when mix moves even if no channel's cost has changed. Shift ten points of volume from dealers to referral and blended CAC drops while nothing has improved. Report by channel, loaded, under one allocation rule, then argue about mix.

The arbitration a marketing leader actually makes is between cost shapes, not costs. Variable channels (dealer, agent, referral) buy volume you can switch off next month. Fixed channels (retail estate, brand-led direct) buy capacity you own for the length of a lease or a campaign. Cutting the variable ones is fast and visible; cutting the fixed ones takes a year and pushes CAC up in whatever remains while it happens.

The diagnostic to run on any channel: if its volume halved next quarter, would CAC per activation hold or double? Anything in the second group needs a demand floor underneath it before it deserves more budget.

🎬 [VIDEO: "How Telecom Companies Make Money" - youtube.com - search this title on YouTube for an accessible walkthrough of telecom unit economics including acquisition and retention cost drivers]

Key takeaways

• Loaded CAC adds commissions, allocated retail cost, subsidies, acquisition credits and onboarding to the media line; media-only can understate the real figure several times over.
• Dealer commission is variable and provisional until the clawback window (typically 90 to 180 days) closes, and paying on gross adds reliably buys gross adds.
• Retail CAC is a throughput number: fewer activations mechanically raise it, and the allocation split between acquisition and service work can move it by half.
• Referral (giffgaff) is a success fee with near-zero waste but a hard volume ceiling and a cannibalisation problem only a holdout can measure; celebrity-led direct (Mint Mobile) can hide its cost in equity and gets cheaper per line as volume grows.
• In prepaid distribution such as Safaricom's, define the denominator as lines still generating revenue at 30 or 90 days; multi-SIMming and dormancy make gross SIM sales a poor proxy for acquisitions.