A fibre network gets built twice. The first build passes premises: duct, cable, joints and cabinets laid along streets on a construction schedule set by permits, wayleaves and civil crews. The second connects customers: a drop, a termination and a router installed when somebody orders service, which can be weeks or years later. The two flows run on different clocks, consume different materials and compete for the same crews and the same yards. Planning them as one programme hides the connection backlog until it is the only thing customers notice.
The weekly build review opens on one number: premises passed against target. The chart is close to plan, the board likes it, and the contractor payments are tied to it. Further down the pack, connection orders in areas that went live last quarter are waiting three weeks for an installer, because installers were moved onto splicing to keep the premises passed line on target. The depot nearest the backlog is short of optical network terminals while another depot holds a pallet of them. The yard behind it has a growing pile of cable offcuts that are too short for any route still to be built.
Each of these comes from treating a fibre rollout as one programme with one measure, when it is two flows of work and material that start at different times, run at different speeds and are judged by different people. Fibre rollout planning is mostly the work of keeping those flows separate where they differ and making the trade-offs explicit where they share resources.
Two demand streams with different clocks
The build stream passes premises. Its demand is the construction programme: which design areas are built in which weeks, by which crews, using which method. Its materials are duct and sub-duct, feeder and distribution cable in several fibre counts, joint closures, splitters, street cabinets and, for aerial routes, poles and fixings. Its measures are premises passed, cost per premise passed and schedule adherence, and its horizon is months.
The connect stream turns a passed premise into a paying customer. Its demand is orders, which arrive only after an area is declared ready for service. Its materials are drop cables, customer terminations, optical network terminals and routers. Its measures are connections, order to install lead time and first visit success, and its horizon is days to a couple of weeks between an order and the appointment.
The streams share almost nothing in their demand signal and almost everything in their resources. The same contractors, depots, vans and splicing teams serve both. That is why a plan that reports one combined programme looks healthy while the connect stream quietly loses customers.
The design area is the unit of demand for the build
Build demand comes from network design. Each design area has a bill of materials derived from the route: metres of each cable type, the number of closures and splitters, chambers to install or reuse, poles to erect or replace. Multiply the design areas scheduled for a month by their bills of materials and you have the material demand for that month.
The catch is that a design passes through stages, and its quantities move at each one. A desktop design drawn from maps and records becomes a surveyed design once crews have walked the route, then a detailed design ready for construction, then an as-built record once the work is done. Surveys find blocked duct that has to be dug out or bypassed, poles that need replacing before they can carry cable, and access problems that reroute a section of the network. A route planned as duct reuse can turn into new trench, which changes both the materials and the crew.
Plan with an uncertainty that depends on design maturity. Areas still in desktop design should carry a wider margin on materials than areas at detailed design, and expensive or long-lead items such as cabinets are best committed once detailed design is approved. The margin itself should be measured rather than assumed: compare designed with as-built quantities for areas already completed, by build type, and use the observed ratio as the planning allowance for the next ones.
Cable arrives on drums, and drums do not divide evenly
Cable is bought and stored on drums of fixed lengths, and routes are cut from them. A drum cut for a long route leaves a remainder that may be too short for any route still to be built, and those remainders accumulate in yards as stock that counts on paper and cannot be installed. Waste in fibre builds is largely decided at the moment drums are assigned to routes.
That makes drum allocation a planning decision in its own right. Matching route lengths to drum lengths before issue, keeping a register of partial drums with their remaining lengths, and assigning short routes to partial drums first all reduce the offcut pile. Counting cable stock in total metres overstates what is usable, so the inventory view that matters is drums by fibre count and remaining length.
Fibre count adds a mix problem. High count feeder cable, distribution cable and drop cable are different items with different lead times, and a shortage of one is not covered by surplus of another. The same applies to closures and splitters, where a design change from one network architecture to another can strand stock bought for the old one.
Ready for service is the date both streams hang on
An area becomes ready for service when its cabinet is powered, its splices are tested and the network records show every premise as connectable. That date ends the build stream's work on the area and starts the connect stream's demand, so a slip in it moves two plans at once. It also tends to slip in a particular way. Most of an area is finished on time, and a few streets wait on a blocked duct, a missing wayleave or a power connection to the cabinet.
Operators handle the tail differently. Some declare an area ready only when every premise can be served, which protects the customer experience and delays revenue on the streets that were finished weeks earlier. Others release an area in parts, which brings orders forward and makes the connection forecast harder, since part of the area is selling while the rest is still a construction site. Either choice is reasonable. What matters is that the planning system knows which one is in force and holds ready for service as a date per premise or per street rather than per area.
The forecast of that date deserves its own measurement. Compare the ready for service date planned at each stage of an area's design with the date achieved, by build type and contractor. The connect stream's installer and equipment plan should be built on the measured date, because a plan built on the hoped-for one puts installers in a region a month before there is anyone to connect.
Take-up decides the connection plan once an area is ready for service
When an area goes live, connection orders follow a recognisable pattern. A first burst comes from households that were waiting for fibre, then a slower climb as contracts with existing providers expire and marketing reaches the rest. The shape varies with the area. Another fibre network in the same streets slows take-up. On open-access networks orders arrive through retail providers, so their campaigns drive the curve. Apartment blocks need building access and internal wiring before anyone inside can connect, which delays their share of the curve.
Forecast connections per area from the curves of comparable areas that are already live, indexed by weeks since ready for service and adjusted for competition and channel. That forecast is what sizes installer capacity by region, terminal and router stock by depot, and drop material by area. It also gives the build plan a commercial input it usually lacks: areas expected to take up quickly are worth building first, because the revenue arrives sooner.
Recovered equipment is a second source of supply. Routers and terminals collected from cancelled and moved customers can return to stock after testing, so the purchase plan for customer equipment should net off the expected recovery rather than assuming every installation needs a new unit.
Crews are the shared constraint
Civil crews, cable crews, splicing teams and installers are separate skills, often supplied by different contractors, and contractors are commonly paid per unit: per premise passed, per splice, per connection. Payment by unit shapes what gets done first, and it means the plan of record may differ from what crews choose to work on in a given week.
The costly trade-off sits between build and connect. Moving installers or splicers onto build protects the premises passed target, and the connection backlog grows in areas already live. Long lead times lose orders, because customers cancel and stay with their existing provider, and those lost connections are the revenue the build was meant to produce. The trade-off is often made by a construction manager looking at one measure.
Plan crew capacity by skill and by region, with an explicit allocation between build and connect, and put the cost of the connection backlog next to the cost of a missed build milestone in the same review. When someone moves a crew, they should see what it does to both lines.
Stock sits in contractor yards and engineers' vans
Most fibre material is issued free to contractors, who hold it in their own yards and vans until it is installed. That stock is easy to lose sight of. It is issued against a design area, used partly on that area, partly on the next, and the remainder may or may not come back when the area closes.
Reconcile issued with installed for each design area at close, using the as-built record. The gap shows loss, theft and design drift, and it differs by contractor in a way that is useful to know before the next contract. For customer equipment, count van stock as part of the network. A terminal sitting in an installer's van in the wrong region is as unavailable to a waiting customer as one that was never bought.
Where this stops
The pace of a fibre build is set by permits, traffic management approvals, wayleaves and access to poles and ducts owned by others. No planning system can predict a highway authority's decision, and none can see a competitor's build plan until crews appear in the street. What planning can do is measure how far dates and quantities move, and set commitments to match.
The data limit is the as-built record. It usually lags construction by weeks, and in some contractor relationships it is never fully reconciled, which leaves the design quantities as the only figures the planning system ever sees. Every margin calculated from design alone inherits that blind spot.
This week, take the last twenty design areas that closed, compare designed with as-built quantities for each material by build type, and calculate the ratio. That distribution is the allowance the next material plan should carry, and it will usually show which build type the current plan understates.
More on this in industry on the DemandLabs platform.
