Capacity planning without the guesswork

Capacity planning tends to fail in one of two directions. Either sites are added ahead of demand that never materialises, or thresholds are crossed quietly and the first signal is a complaint. Both are expensive; the second is also reputationally costly.

Busy is not the same as constrained

The most common analytical error is treating high utilisation as evidence of a capacity problem. A cell can run hot and still deliver acceptable experience. Another can look moderate while a specific resource (signalling, scheduler, backhaul, licence) is the actual binding constraint.

Separating those cases is what makes a forecast worth budgeting against. Otherwise you are ranking cells by symptom and buying equipment for the ones that shout loudest.

Forecast the threshold, not the trend

• Model when each cell reaches its capacity threshold, not just where traffic is growing
• Rank interventions by impact per unit of spend, so cheap fixes are exhausted first
• Use the operator's own assumptions, because a generic model will not survive a budget review
• Re-run automatically, so the plan is current rather than annual

Optimise before you build

The cheapest capacity is usually the capacity you already own. Parameter, tilt and neighbour work, signalling reduction and load balancing frequently recover meaningful headroom, and they are reversible in a way that a new site is not.

That is also the honest sequencing argument to make internally: exhaust configuration before capital. A plan that shows optimisation was tried first is considerably easier to defend than one that opens with a purchase order.

Network managers should be able to say exactly when, what and where to invest, and what it will cost if they do not.