Mobile operators have been migrating from microwave to fibre backhaul for over a decade, and the migration logic has barely changed: more capacity, better latency, lower OPEX per Gbps, future-proofed for 5G densification. All true. None of it tells you whether the case actually closes at your network.
The case closes — or does not — on four assumptions that almost every internal business plan gets wrong.
1. Traffic growth that compounds, not extrapolates
Most backhaul models assume linear traffic growth: take last year, add 30%, repeat. That is a reasonable shorthand when you are sizing a circuit. It is a terrible basis for a 10-year capex decision.
Real traffic growth is non-linear and bursty. It accelerates around use-case inflexion points (video offload, live streaming, fixed-wireless replacement) and decelerates when the network gets congested. A defensible migration case models the inflexion points, not the trend line.
2. The civils cost you are actually going to pay
Per-site fibre build cost is the single biggest sensitivity in any migration model. And it is the assumption least likely to be defensible, because the cost is often pulled from a desktop GIS estimate based on aerial distance, not actual route length.
The aerial-to-actual multiplier varies enormously by terrain:
- Dense urban — actual route is typically 1.3 to 1.5x aerial distance.
- Suburban — 1.5 to 1.8x.
- Rural with terrain — 2x or more, before you account for permitting delays.
If your model uses 1.2x for everything (a surprisingly common default), you are understating civils cost by 20-50% on most of the network. That 22% IRR you are showing the board is probably 14% in reality.
"On a recent 4,000-site review, applying terrain-corrected civils factors moved the programme NPV by $94M — and changed the recommendation on three regional clusters."
3. Microwave residual value (or lack of it)
Migration cases routinely assume that the existing microwave links can either be redeployed elsewhere in the network or sold for residual value. Sometimes that is true. Often it is not.
Older microwave gear that does not support modern modulation schemes has limited internal redeployment value, and the second-hand market for it is thin. The "savings" from removing it are real (less licensing, less power, less maintenance), but the residual asset value is usually closer to zero than to book.
Be honest about this in the model. The migration case does not need it to close — it just needs you to stop pretending it is a benefit.
4. Time-to-revenue, not time-to-handover
The dirty secret of backhaul migration is that fibre activation does not equal traffic migration. Once the fibre is at the tower, the operator still has to commission it, run it in parallel with the microwave for a soak period, validate performance, and then cut over. That is typically 2 to 4 months between first light and revenue substitution.
If your model assumes activation = revenue, you are overstating year-one IRR by 8-12%. Fix the timing assumption and the case still works — but the cluster prioritisation might change, because the clusters with the longest soak windows are the ones to defer.
The next time someone hands you a backhaul migration deck, ask to see the four assumptions above before you ask about the conclusion. If the answers are not specific to your network and your terrain, the conclusion is not yet investment-grade.