Capacity planning is the practice of knowing when you will run out of bandwidth or infrastructure capacity before your subscribers tell you. An ISP that upgrades infrastructure reactively, in response to subscriber complaints or NOC alerts about 95th-percentile utilisation exceeding 90%, is always one growth surge behind. An ISP that plans capacity proactively has upgrade orders in process before the current capacity is saturated.
The practical challenge is that capacity upgrades have lead times. A transit bandwidth upgrade from your provider requires a contract amendment and provisioning time. A new OLT card or chassis for a saturated GPON site requires procurement, shipping, and installation. If the decision to upgrade is not made until utilisation is already critical, the lead time means weeks of degraded service while the upgrade is in process.
The Metrics That Drive Capacity Decisions
Peak utilisation is the primary metric, specifically the 95th percentile of peak hourly traffic on each critical link over a rolling 30-day period. The 95th percentile eliminates the top 5% of traffic (brief spikes that are not representative of sustained load) while capturing genuine peak demand. Monitor this for every link where utilisation determines subscriber experience: transit uplinks, core inter-site links, aggregation uplinks from access to core.
The upgrade trigger threshold should be set such that an upgrade is initiated when the 95th percentile reaches a level that gives enough lead time for the upgrade to complete before utilisation reaches a degraded state. For a transit link with a 30-day upgrade lead time, setting the trigger at 70% 95th percentile gives approximately 3-4 months of headroom at typical growth rates before the link is at risk. For a link with a 1-week upgrade lead time, the trigger can be set higher.
Contention ratio per PON port tracks how many subscribers are sharing each GPON or EPON port and what their aggregate utilisation looks like during peak. A PON port at 1:64 split with 60 active subscribers consuming an average of 10 Mbps peak each is approaching the 600 Mbps theoretical limit of a GPON port's available downstream bandwidth, even though the 2.488 Gbps raw capacity suggests plenty of headroom. The contention calculation must account for actual subscriber usage, not theoretical subscriber count.
Connection tracking table utilisation on CGNAT routers is a capacity metric that is easy to overlook until it becomes critical. As subscriber count and activity grows, connection tracking tables fill and new connections are dropped when the table is full. Monitor this metric specifically and set upgrade triggers well before the table approaches capacity.
Traffic Forecasting
Traffic growth at most Pakistani ISPs follows two patterns: organic subscriber growth that increases aggregate demand roughly linearly with subscriber count, and step-change growth from specific events (a corporate customer adding a high-capacity CIR circuit, a residential building connecting to FTTH, a WISP acquiring subscribers from a competitor).
For organic growth, extrapolating from recent 90-day trends gives a reasonable 3-6 month forecast for capacity planning purposes. The extrapolation does not need to be sophisticated: if traffic on a link has grown 15% over the past 90 days, a further 15-20% growth over the next 90 days is a reasonable planning assumption absent specific intelligence about the subscriber pipeline.
For step-change growth, the subscriber management and sales pipeline is the input: if your sales team has an active pipeline of corporate clients that would collectively add 500 Mbps of CIR capacity, that demand needs to be reflected in capacity planning regardless of where current utilisation sits.
Defining the Upgrade Decision Process
Capacity planning is only useful if it leads to decisions. Define explicitly: who reviews capacity metrics and how often (monthly is the right cadence for most ISPs), what the upgrade triggers are for each link category, who has the authority to initiate a capacity upgrade, and what the standard lead time is for each upgrade type.
Putting this process in writing, even as a simple document listing the trigger thresholds and the approval process, converts capacity planning from an ad-hoc reaction to a systematic discipline. It also makes the capacity planning defensible to an ISSC or board that asks why infrastructure spending is increasing: you can show that upgrades were triggered by specific metrics reaching defined thresholds.
For operators who want NOC-level visibility into utilisation trends with automated alerting when thresholds are approached, NOC Intelligence provides the triage and alert classification layer that surfaces capacity signals before they become subscriber complaints. For the NMS infrastructure that generates the utilisation data capacity planning depends on, NOC Enablement & Monitoring covers the monitoring stack design. For capacity planning as part of a broader network architecture and growth strategy engagement, Network Design & Optimization and ISP Consulting & Advisory cover both the technical and strategic dimensions.