The most expensive mistake in fibre rollout is not technical — it is sequencing. The wrong cluster gets built first, and a programme that should have hit IRR targets in year three drags into year five, or never gets there at all.
Most rollout sequences we review have been chosen on a mix of three things: where the operator has existing infrastructure, where senior leadership has visibility, and where competitors are already building. None of these are bad inputs. None of them are sufficient inputs either.
What yield-weighted prioritisation actually means
Yield-weighted prioritisation answers a simple question: if I can only build five clusters this year, which five maximise programme NPV?
Doing this properly means scoring every candidate cluster against four dimensions:
- Demand density — addressable revenue per km of network, segmented by willingness-to-pay tier.
- Build cost intensity — per-home-passed civils cost based on actual ground conditions, not desktop GIS.
- Activation lead time — months from first dig to first revenue, including PoP availability and back-office readiness.
- Strategic option value — does this cluster unlock adjacent expansions, or is it terminal?
Each cluster gets an NPV. Each NPV is then risk-adjusted. The clusters get ranked. The top of the list is what you build first — even when the political and geographic temptation is to do something else.
"In one engagement, switching the build order on the same set of 18 clusters lifted three-year cumulative free cash flow by 31% — without changing a single capex line."
Why operators get this wrong
The honest answer is that the modelling required to do it well sits between three teams that rarely talk to each other: commercial (who own the demand), engineering (who own the cost), and finance (who own the discount rate). When the model lives in three spreadsheets, you get three different rollout orders depending on whose meeting you are in.
The fix is not technical. It is operational. The cluster prioritisation has to live in one model, owned by one team, with version control and assumption transparency. That single source of truth is what lets the executive committee actually decide.
The numbers that change the conversation
When we do this properly, three things almost always emerge from the analysis:
- Between 15-25% of the planned year-one clusters should be deferred.
- At least one cluster nobody had on the radar belongs in the top 10.
- The IRR uplift from re-sequencing alone is usually 3-6 percentage points.
That last point is the one that wakes up the investment committee. A 4-point IRR improvement on a $500M programme is real money — and it costs almost nothing to capture, because you were going to spend the capex anyway.
If you would like a confidential review of how your current rollout sequence stacks up against a yield-weighted alternative, our investment diagnostic is the fastest way to get a defensible second opinion.