What Can Load Curves Teach Us About Project Development?
A systems-based view of how development effort shifts over time.
Over the course of developing energy projects, I commonly think about work in linear terms: schedules, milestones, and cash flows. Effort is expected to build, peak, and taper, forming a continuous S-shaped curve that reflects defined project phases such as design and engineering, permitting and approvals, commercial contracting, and project financing. Variability in workload is often treated less as signal and more as noise, something to be managed, flattened, or explained to investors after the fact. But when project activity is observed over a sufficiently long period, and at a sufficiently granular level, a different pattern can begin to emerge.
This data story draws on four years of weekly project data from a single energy development, including recorded hours, workstream classifications, and activity types. Rather than focusing on individual tasks or milestones, the data is examined at the level of aggregate weekly effort, allowing the full distribution of work over time to be observed. Viewed in this way, the project does not operate as a continuous, evenly paced process. Instead, its workload organizes into a distinct structure.[1]
At the most basic level, weekly effort separates into three recognizable ranges: higher-intensity periods, moderate operating periods, and lower-intensity periods. These are described here as “peak,” “mid-peak,” and “off-peak” conditions, and the separation between these ranges is quite distinct. Each band occupies a different portion of the overall timeline, with higher-intensity weeks forming a relatively small share concentrated in the early to middle periods, while moderate and lower-intensity weeks make up the majority of observed activity across the middle and later periods.[2]
This data pattern was initially explored in a previous article and data story in which I suggested that development is not best understood as operating at a steady baseline with occasional deviations. Instead, it appears to move through distinct operating states, each with its own characteristic level of effort.
However, the structure of the development data becomes more interesting, and perhaps less intuitive, when those same weeks are reorganized not chronologically, but by intensity.
When weekly effort is ranked from highest to lowest and plotted as a continuous distribution over time, the resulting curve takes on a familiar shape. The highest levels of effort occur over a very small percentage of weeks. Effort then transitions through a middle range and extends into a long tail of lower-intensity weeks. This type of curve is well known in another domain: energy systems, where it is referred to as a load duration curve. And when compared to a normalized profile of Ontario electricity demand over the same period, the two curves align closely in overall shape.[3] Both exhibit a concentrated set of high-load conditions, a broad middle range, and an extended lower-load tail.[4]
It is important to be precise about what this does and does not imply. The comparison does not mean to suggest a causal relationship between electricity demand and development effort. Rather, the comparison is strictly descriptive. It simply shows that two independently generated datasets share a similar structural distribution when normalized and ranked.[5]
In energy systems, the shape of a load duration curve reflects the aggregation of many underlying drivers: human activity patterns, operational constraints, infrastructure limits, and temporal coordination across a network.[6] The curve is used not only to describe demand, but also to inform system design, determining how different types of resources are deployed to meet varying levels of load.[7]
Seen in this light, the similarity raises an interesting research question: if development effort naturally forms a distribution with characteristics comparable to an energy load, is it helpful to think of project development as a system with its own “load shape,” driven by demand and supplied by resources? This exercise does not require the analogy to be exact; it only requires that the comparison provide a structured way to interpret what is observed.[8]
To explore this systems view of development further, we look beyond the ranked distribution of development effort to reveal a second layer of data structure. Project work is not only shifting in intensity over time; it is also organized internally in a non-uniform way. Different types of work, such as strategy and communications, knowledge development, and analytical validation, are not distributed evenly across the project. Instead, they are concentrated in specific functional areas.
Strategy and knowledge-related work are carried primarily within the commercial-finance function, forming a centre of gravity. Analytical validation, by contrast, is more distributed, with a greater share falling within engineering-design and administration-management. This indicates that the project’s workload is not a homogeneous pool of effort, but rather a composition of distinct functional nodes, each responsible for different types of work.
When this internal structure is examined alongside the band-based pattern of peak, mid-peak, and off-peak effort, a more complete picture of a system begins to emerge. As the project moves between these conditions, the distribution of work shifts across functions. Engineering-design is more prominent during higher-intensity periods, while administration-management increases its share during lower-intensity periods. At the same time, the mix of activities evolves. Strategy and communication-oriented work become more prominent as overall intensity declines, while knowledge- and validation-heavy work become less dominant.
This shifting of work across bands indicates complexity, suggesting the system is non-linear, adaptive, and self-organizing. Lower-intensity periods are not simply mini versions of peak periods; they are qualitatively different. The project remains active, but the nature of that activity shifts from technical execution toward coordination, alignment, and preparation. Effort is not simply expanding and contracting in response to workload. It is reorganizing itself as conditions change, both in terms of who is doing the work and the type of work being done. This is where the energy system analogy can become useful.
In energy systems, different types of resources serve different portions of the load curve. Some operate continuously, such as nuclear and large hydro, providing a stable base level of supply. Others, like gas plants, are brought online to meet intermediate demand, while additional gas capacity and battery storage are deployed only during peak conditions. At the same time, intermittent resources act as “negative demand,” reducing the load that dispatchable generation must supply. Demand-side strategies also play a role, shifting flexible loads away from peak periods and redistributing activity across the curve.
While a development project is not an energy system, the parallel can be instructive. Some types of project work appear to function as a steady core, persisting across all bands. Others expand and contract with intensity, while some activities shift in timing depending on when they are most needed, most effective, or most available.
This does not mean that project work can or should be managed in the same way a utility or system operator manages energy supply and demand. However, it suggests that system-based concepts, such as operating states, load distribution, and functional specialization, may offer a useful lens to view development. Which resources function as “baseload,” providing a consistent foundation of strategy and knowledge? Which are “dispatchable,” responding to specific needs as they arise? And which are “self-scheduling” or “intermittent,” with output shaped by external conditions?
The data story that follows explores these patterns across five visualizations. It describes not only how much work occurred over time, but how that work was distributed, where it was carried within the organization, and how its composition changed as the project moved between different operating conditions.
My underlying thesis is straightforward: development behaves as a system. The visualizations seem to support this, showing that as the level of effort shifts, the structure of the work reorganizes with it.
The implication is more open-ended. If development naturally takes shape as a system, the question is not whether the load curve can or should be changed, but whether it can be interpreted and used to achieve better balance.
Please click on the button to view the full interactive data story for dynamic charts and layered narrative. Below is a static version of the data story. Curious readers are encouraged to scroll down and read the notes.
[1] This analysis is based on a single project over a four-year period. The findings are descriptive of this dataset and should not be generalized without comparison to additional projects.
[2] A k-means clustering approach was used to classify weekly effort into three distinct operating ranges. Weekly effort values were normalized and grouped into three clusters, with each week assigned to the nearest cluster centroid through an iterative distance-based procedure. The resulting centroids defined characteristic levels of project activity and were used to identify peak, mid-peak, and off-peak conditions.
[3] Data source: IESO Hourly Demand Reports (2022–2025), corresponding to the years of project development. The hourly dataset (Date, Hour, Ontario Demand) was aggregated to a weekly level to match the granularity of the project data, which is recorded and managed in weekly intervals. Using hours or days would introduce artificial precision not supported by the underlying dataset, while monthly aggregation would smooth out meaningful variation in intensity.
[4] The correlation between the ranked development effort curve and the ranked Ontario electricity demand curve is 0.995.
[5] While the literature does not suggest there is a quantitative relationship between energy load profiles and organizational workloads, several studies support the analogy, particularly in areas such as peak-shaving, workload shifting, and resource allocation.
[6] To underscore the challenges of balancing an energy system, one of my former partners liked to say, “Electricity moves at the speed of light, while natural gas moves at 18 miles per hour.”
[7] Modern energy systems are planned using advanced analytics and modelling that determine what to build, where to build it, and when it must come online. For the most part, project development has lagged in adopting this type of data-driven approach, with many projects still emerging as solutions in search of clearly defined demand.
[8] There is a substantial body of work using systems and complexity theory to model and optimize energy systems, along with some research applying these ideas to energy projects and governance. However, little attention has been given to translating the complexity of supply–demand network systems into the design of internal organizational structures, workflows, and resource allocation within project development teams. Linking detailed energy system models with organizational and system-dynamics models of project development remains a relatively underexplored and promising area for further research.