Wednesday Reality Check : The Reservoir Was Full
At six in the morning, the tap in the house at the top of the hill runs well enough to fill a kettle while its owner finds a cup. An hour later, the same tap takes long enough that she leaves the kettle in the sink and goes to open the shutters. Down in the valley, the water company’s level recorder shows the service reservoir close to full throughout. It has dropped a little since dawn, as it does most mornings. If the operator had only that trace and the customer had only the kettle, they would each have good reason to wonder what the other was talking about.
The morning draw
In a gravity-fed system, what matters for pressure at the house begins with the difference in height between the reservoir’s water surface and the point where the water is delivered. With the water at rest, a ten-metre difference in elevation gives roughly one bar of pressure. A house low in the valley starts with more of that difference in its favour. The one at the top of the hill starts with less. A broad reservoir can hold a great deal of water across a relatively small change in surface level. Its volume tells the operator how much water is in reserve. The elevation of its surface helps explain what that water can do.
Once water starts moving, it loses some of its available head to friction along the way. The loss increases with flow and depends on the length, diameter and condition of the pipe, among other things. For part of its journey, the water going to the hill shares a main with water going to other streets. The owner can open exactly the same tap to exactly the same position on two successive mornings and get a different result because of what other people draw through that main. A network with several routes may redistribute some of that flow. A street served through a single branch has fewer options. The complaint has an address. Its cause may sit several junctions away.
Some of that head disappears over a very short distance. A tight elbow typically costs more head than a sweeping bend at the same flow. A tee also has a different loss depending on whether the water passes straight through or turns into the branch. A fitting can occupy a few centimetres on the drawing and still account for a meaningful share of the loss along a short route. On a long, narrow main, friction along the pipe may matter much more.
Now add a housing development further down the hill, supplied through part of the same main. Its total daily consumption may fit comfortably within the town’s water budget. The reservoir may refill every night. The development can still change what happens at breakfast if its busiest hour coincides with the busiest hour on the existing network. A connection tested while most of the new houses stand empty tells the engineer something about the pipework on the day of the test. The first winter with every house occupied tells her something else. In between, the town may have issued every certificate required to open the development.
What the network can spare
There are several ways to get more pressure to the top of the hill, and their consequences differ. Raising the supply head across the entire network can put lower streets under more pressure than they need. A booster serving a separate upper zone can change the conditions on the hill without imposing the same increase on the valley. A larger main or a second feed can reduce the losses along the route. An elevated tank within the upper zone can cover part of a short peak if it has sufficient head and the network can replenish it in quieter hours. Enlarging the existing reservoir at the same water-surface elevation would buy the town time during a supply shortfall. It would do little for a morning pressure drop caused by the shared main while the reservoir level stays high.
The water company can distinguish some of these possibilities by recording pressure on the hill over the course of the day and comparing it with flow through the main. The timing matters. A technician who arrives after the morning draw can take a perfectly sound reading and leave with the impression that the trouble has cleared. Even a permanent gauge at the pumping station gives a limited account of what happens at a distant junction. Where the instruments sit determines which explanation the operator can readily test. This gives the householder’s report an engineering value of its own. She can say when the water stopped coming out fast enough to fill a kettle.
Four projects reach the same junction
In a software portfolio, four delivery teams can have enough people to build four changes. One changes an order flow, one moves a warehouse onto a new interface, one adjusts the rules for refunds and one replaces a reconciliation job. The portfolio review shows room in all four plans. Each project also needs the same small group to explain what the integration layer does with an order it has already seen, establish which system owns the final status, and trace discrepancies once several systems start exchanging real data. A first estimate might allow for a few days of specialist work per project. The actual demand comes in bursts, often during the same fortnight because the projects share a release calendar.
At first the teams work around the wait. The warehouse project builds against its best reading of the interface. The refunds team uses test data that represent a clean transaction. Neither decision looks particularly expensive while the specialist is busy elsewhere. When testing finally exposes the gaps, some of the work returns. A question that would have taken half an hour during design now needs an engineer to reconstruct a transaction, a tester to reproduce the discrepancy and someone from the business to decide what the outcome should have been. If the resulting fix misses the next environment slot, the delay can survive long after the question has been answered. The portfolio counts the specialist hours spent investigating the defect. The extra days often turn up under the project that had to wait.
The integration ticket can look complete on the sending team’s board while the receiving engineer still has to recover the decision behind it. Take a warehouse retry. If the two teams have agreed what happens when the warehouse resends an order after a timeout, the receiving engineer can test that behaviour directly. If the ticket simply says ‘support retries’, somebody must reconstruct the failure before deciding whether a second message should create a second order. Both tickets cross the same boundary. Only one carries enough of the original decision to survive the crossing.
Water and work part company here. Opening more taps cannot give the water company a second version of the same litre to deliver. An organisation can create work by making work wait. It can also create a queue in advance of any real need. Once project managers learn that a request made at the point of need will sit unanswered, they begin reserving the people they expect to need later. They hold a slot while a design is still changing or escalate a request before the team can explain what it requires. A booking starts to express the cost of being caught without one. The calendar grows fuller and becomes less useful as an account of what anyone is ready to do.
The economics of a full calendar
A motorway offers a closer comparison for the queue. A ramp meter controls how quickly vehicles join the main carriageway. It can help preserve flow through a busy merge, but it also asks drivers to wait somewhere else. If the ramp fills, that queue can spill onto local streets. The point where waiting occurs changes who bears the cost and who gets to see it. In the portfolio, the same fortnight of delay looks different before a project starts and after a team has built against an unresolved assumption. Before the start, it appears as an awkward choice about which commitment should move. After the start, it spreads across rework, idle test slots and engineers pulled away from other projects. Much of that cost gets recorded as ordinary delivery effort.
The apparently idle hour of a specialist has a peculiar value in this setting. If the hour has already been promised to a fifth project, the business can point to the work it bought. If it remains available, its value may lie in a design question answered before three teams implement different interpretations. Sometimes the hour passes quietly. Its benefit across a quarter appears partly in events that never enter a report, such as a failed test that never needed to be written up or a release that did not require a week of diagnosis. A finance review can readily attach a cost to the unbooked hour. It has a harder time attaching a value to the weeks that remain uneventful because somebody had time to look.
The choice between adding engineers to a product team and widening access to integration expertise has a similar accounting problem. The product hire gives one director a visible increase in capacity. Pairing on a legacy interface, documenting how it handles retries, or building representative test cases initially takes time from the people everyone already needs. For a period, the queue can get worse. The return comes later and across several teams, each of which can now answer a class of questions without waiting. Some questions will still need the specialists. What changes is how often four independent plans become one queue.
The reading at seven
A few years later, the town considers another development. The engineer brings two traces to the planning meeting. One shows the reservoir level over a week. The other shows pressure on the upper branch over the same period. The first moves gently within its normal range. The second falls sharply most mornings and recovers by mid-morning. On the reservoir chart, the hour when people first noticed trouble is almost impossible to pick out. On the other chart, it is the most recognisable feature of the week.
At the next portfolio review, a director puts four project plans on the screen. Each has a line for integration and each has a different date for the first request. Three months later, all four lines will come to mean the same two people in the same week. For now, the plans are reviewed one by one. There is room in every plan.
Technical references
Member discussion