Your ARPU is an average. Your margin lives in the differences it hides.
In telecommunications the unit that decides profit is not the subscriber, the plan or the tower. It is the combination: this customer, on this plan, through this channel, consuming this slice of network at this hour. A blended ARPU and an allocated network cost flatten all of that into one number. Time-driven costing puts the differences back.
Cost and Profitability Consulting · 150+ models since 2010 · TDABC
Telecom economics are dominated by shared network capacity and per-customer cost-to-serve, neither of which a blended ARPU reveals. The same patterns that hold across every sector apply here: industry research shows cost-to-serve varies two to three times between customers who look identical, and studies consistently find roughly 30 percent of any fully costed business is unprofitable. TDABC assigns network capacity and service cost to the customers, channels and plans that actually consume them, so each shows its real margin instead of an average. We do not publish a telecom-specific benchmark, because rigorous figures for this sector are not in our base; the method, not an invented number, is what makes the difference visible.
A shared network and a per-customer cost the average erases.
A blended ARPU hides who pays
Average revenue per user says nothing about average cost per user. Two subscribers on the same plan can cost very different amounts to serve. Industry research shows cost-to-serve varies two to three times between apparently identical customers.
Network is a shared, peak-driven capacity cost
The expensive thing is capacity at peak, not bytes on average. Allocating network cost by simple volume over- or under-charges customers depending on when they consume. Studies consistently find most organisations do not measure the cost of unused capacity.
Churn and cost-to-serve are rarely costed together
A high-touch, complaint-heavy subscriber can be unprofitable even before churning. Retention spend aimed at the wrong segment destroys value.
Channel cost is invisible in margin reports
A retail store, a call centre, a reseller and self-serve digital have very different costs. Counted at blended cost, the cheap channels subsidise the expensive ones silently.
PEAK CAPACITY, NOT AVERAGE VOLUME
Illustrative analogue, drawn from a large IT organisation, not a telecom benchmark. The network is sized for the peak, so the peak is what the capacity actually costs.
Two parameters, no surveys.
A capacity cost rate per resource group (network element, call centre, field crew, store) and time equations that describe how each customer, plan or channel consumes those resources. The cost drivers that matter here are peak network consumption, support-contact frequency, channel of acquisition and service, plan complexity, field and truck-roll events, and provisioning churn.
Subscriber cost = network capacity share (peak-weighted) + support contacts x minutes per contact x call-centre rate + field / install events x crew time + channel-specific acquisition and servicing cost + billing and provisioning overhead
Illustrative structure, not a measured benchmark. The peak-weighted network share is where two subscribers on the same plan diverge.
The whale curve is the honest map.
Across sectors, industry research shows the top 20 percent of customers generate 150 to 300 percent of profit while the bottom 10 to 20 percent destroy 50 to 200 percent of it. In telecom that shape typically shows up as a band of high-touch or low-usage subscribers that the average quietly hides. Two analogous patterns from the IT and digital world show the levers, because telecom shares their economics. A large IT organisation priced shared IT capacity using peak-capacity pricing across several time zones, charging consumers for the capacity they drove at peak rather than a flat average, the same logic that applies to peak-driven network capacity. A global software vendor, moving from licence to subscription, lifted average selling price and margin by aligning price with the cost-to-serve of an ongoing relationship; telecom is already a subscription business, and the unfinished work is aligning each plan's price with its real cost-to-serve. These are illustrative analogues from adjacent sectors, anonymised and with adjusted figures, not telecom benchmarks.
SUBSCRIBERS, RANKED BY TRUE MARGIN
Transversal whale curve, the agnostic shape applied to subscribers. Not a telecom benchmark; the lens, not a sector number.
Rich data, weak allocation.
Telecom usually has rich raw data (usage, network, CRM), so Data & Technology tends to be a relative strength. The weak dimensions are almost always Cost Allocation, network and support cost spread by simple volume rather than by peak capacity consumed, and Profitability Visibility, margin seen at company or plan level but never per subscriber, channel and service. The gap is method, not data, and the seven dimensions are read qualitatively here, with no invented sector score.
The network runs itself, the bot answers. Where does margin go?
AI changes telecom cost on two fronts at once. On the network side, AI-driven operations and predictive maintenance reduce the field events and capacity buffers that drive cost. On the customer side, AI absorbs tier-1 support and reshapes the channel mix, which moves cost-to-serve faster than pricing can follow. The operators who win are the ones who already know their per-customer and per-channel cost, so they can re-price as the cost base shifts. This is a question of decision quality, not a regulatory countdown.
Two ways into the sector's cost.
Frequently asked questions
- How do you measure true customer profitability in telecom?
- Load network capacity (peak-weighted), support contacts, channel cost and provisioning onto each subscriber using time equations, then compare to revenue. ARPU alone cannot do this.
- Why is a blended ARPU misleading?
- Because cost-to-serve varies two to three times between customers on the same plan, industry research shows. Average revenue says nothing about who is actually profitable.
- How should network cost be allocated?
- By the capacity each customer drives at peak, not by average volume. Peak is what sizes the network and therefore the cost.
- Do you have a telecom-specific benchmark?
- No. We do not publish sector figures we cannot stand behind. We apply transversal evidence (whale curve, cost-to-serve variance, capacity costing) and the TDABC method to your own data.
See which subscribers, channels and plans actually pay.
ProfitAudit 360 builds the per-subscriber, per-channel picture on a TDABC base. Or take the Profit Check first: five minutes, no data upload, and it points to where your blended ARPU is most likely wrong.