Workforce planning and productivity

A schedule says what will be built and when. A workforce plan says who will build it: how many workers of each trade, when they are needed, where they will come from, and how productive they are likely to be. On Elmwood, the difference between a schedule that works and one that does not often came down to labour: whether there were enough carpenters for the forming cycle, whether the electricians could staff two floors at once, and whether overtime was helping or hurting. This lesson covers how to turn a schedule into a workforce plan, how to manage the peak, and what affects labour productivity.

The labour market behind the plan

Construction labour is scarce in much of Canada. The federal government said in 2026 that the country will need more than 1.4 million new skilled trades workers by 2033, and BuildForce Canada's annual forecasts warn that retirements will remove a large share of the experienced workforce over the next decade, even in provinces where demand is levelling off. Major public investment in housing, transit, hospitals and energy is adding demand at the same time. For a contractor, this means that labour cannot be assumed to be available on short notice: the workforce plan must be made early, shared with unions and subcontractors, and revisited often.

From schedule to labour curve

Every activity in the schedule has a quantity and a production rate, which together give the labour hours it needs. Spreading those hours over the activity's duration and adding up all activities gives a labour histogram: the number of workers of each trade needed in each week of the project. Plotted cumulatively, the hours form the familiar S-curve.

# Weekly carpenter demand from the Elmwood forming schedule (simplified)
activities = [
    # (name, start week, duration weeks, labour hours)
    ("podium walls and columns", 1, 8, 28_000),
    ("podium decks", 3, 10, 34_000),
    ("tower floors 2-4", 13, 13, 42_000),
    ("tower floors 5-7", 24, 12, 38_000),
    ("stairs, shafts and penthouse", 34, 8, 16_000),
]
hours_per_worker_week = 40
weeks = range(1, 42)
demand = []
for w in weeks:
    h = sum(hrs / dur for _, s, dur, hrs in activities if s <= w < s + dur)
    demand.append(h / hours_per_worker_week)
peak = max(demand)
print(f"peak demand {peak:.0f} carpenters in week {demand.index(peak) + 1}")
for w in range(1, 42, 3):
    print(f"week {w:2d}: {demand[w - 1]:5.0f} " + "#" * int(demand[w - 1] / 4))

peak demand 172 carpenters in week 3
week  1:    88 #####################
week  4:   172 ###########################################
week  7:   172 ###########################################
week 10:    85 #####################
week 13:    81 ####################
week 16:    81 ####################
week 19:    81 ####################
week 22:    81 ####################
week 25:   160 #######################################
week 28:    79 ###################
week 31:    79 ###################
week 34:   129 ################################
week 37:    50 ############
week 40:    50 ############

The histogram shows sharp peaks: about 172 carpenters in weeks 3 to 8, when the podium walls and decks overlap at full crew sizes, about 160 in weeks 24 and 25, when the upper tower floors start while the lower floors finish, and a smaller rise when the stairs and penthouse begin. Between the peaks, demand falls to about 80. A curve like this is expensive to staff: workers hired for the first peak must either be laid off in the dip, and may not come back, or kept on with too little to do. Luis's next step was to smooth it.

Levelling the curve

Resource levelling means adjusting the timing of activities within their float, or changing crew sizes and sequences, so that the workforce rises and falls more smoothly. On Elmwood, Luis stretched the podium walls and decks over longer durations with smaller crews working in parallel, started the lower tower floors two weeks later and the upper floors two weeks earlier, again with smaller crews working longer, and kept the stairs and penthouse finishing in week 41 as before. The total hours did not change and the finish date did not move, but the peak dropped from about 172 carpenters to about 111, roughly a third lower. A smaller, steadier crew is easier to recruit, easier to supervise and more productive, because workers learn the building and the repeated floor cycle.

Levelling has limits. Some activities have no float, some sequences are fixed by structure or by the owner's phasing, and some peaks cannot be avoided. In those cases, the plan should show when extra workers are needed early enough to request them from the hall or recruit them, and when they will be released, so that workers know what to expect.

Planning each trade and each subcontractor

The general contractor's workforce plan must cover the subcontractors as well, because on a building like Elmwood they supply most of the workers. At the start of the project, Calder Ridge asks each major subcontractor for its own labour curve, based on the master schedule, and combines them into a site-wide plan. The combined plan shows the peak number of workers on site, which drives decisions about site facilities, parking, hoisting, washrooms, lunchrooms and orientation capacity, as well as the risk that one trade's shortage will hold up others.

What drives labour productivity

Labour productivity, the amount of work produced per worker-hour, varies enormously in construction, and much of the variation is under management's control. Studies of construction work consistently find that a large share of a typical worker's day is spent not on direct installation but on waiting for materials, information, equipment or other trades, travelling around the site, and reworking mistakes. Contractors that reduce these losses gain more than any wage difference.

The real cost of overtime

When a project falls behind, overtime is the quickest response, and for a week or two it works. Over longer periods, studies by industry groups and researchers have found that productivity per hour falls as long weeks continue: workers tire, mistakes increase, absenteeism rises, and the extra hours produce less than expected. Estimates vary, but a crew working 50 to 60 hours a week for more than a few weeks commonly loses 10 to 20 percent or more of its hourly productivity, and the loss grows the longer the schedule continues. Combined with premium pay, the cost per unit of work can rise sharply.

# Cost per unit of work for a carpenter crew under different weekly schedules over 8 weeks (illustrative)
base_rate = 82.54          # loaded cost of a straight-time hour ($)
ot_rate = 82.54 + 0.5 * 48.50 * 1.24   # loaded cost of a time-and-a-half hour ($)
schedules = {
    # name: (straight hours, overtime hours per week, productivity factor over 8 weeks)
    "40 h/week": (40, 0, 1.00),
    "50 h/week": (40, 10, 0.90),
    "60 h/week": (40, 20, 0.82),
}
for name, (st, ot, prod) in schedules.items():
    cost = st * base_rate + ot * ot_rate
    output = (st + ot) * prod          # units of work, where 1 unit = 1 productive straight-time hour
    print(f"{name:10s} cost/week ${cost:8,.0f}  output {output:5.1f} units  cost per unit ${cost / output:6.2f}  output vs 40 h: {output / 40:4.2f}x")

40 h/week  cost/week $   3,302  output  40.0 units  cost per unit $ 82.54  output vs 40 h: 1.00x
50 h/week  cost/week $   4,428  output  45.0 units  cost per unit $ 98.39  output vs 40 h: 1.12x
60 h/week  cost/week $   5,554  output  49.2 units  cost per unit $112.88  output vs 40 h: 1.23x

Working 60 hours a week for eight weeks produces only about 23 percent more output than working 40, at roughly 35 to 37 percent more cost per unit of work. If Luis needs 23 percent more output, a second, smaller crew or a second shift will often deliver it more cheaply, provided the workers and supervision are available. Overtime remains the right tool for short bursts: a pour that cannot stop, a week of recovery after a weather delay, or work that must be done when the hospital's existing building is quiet.

Measuring productivity on site

Calder Ridge measures productivity by comparing earned hours with actual hours. Each cost code has a budget of hours per unit of work. Each week the foreman reports quantities installed, which are converted to earned hours at the budget rate, and the payroll gives actual hours. A productivity factor below 1.0 means the crew is using more hours than planned.

# Weekly productivity by cost code: earned hours vs actual hours (Elmwood, week 22, illustrative)
codes = {
    # code: (budget hours per unit, units installed this week, actual hours)
    "column forms (each)":         (14.0, 18, 280),
    "deck forming (m2)":           (0.85, 980, 900),
    "wall forms (m2)":             (1.20, 240, 330),
    "stair forms (flight)":        (32.0, 3, 112),
}
te = ta = 0
for code, (rate, qty, actual) in codes.items():
    earned = rate * qty
    te += earned; ta += actual
    pf = earned / actual
    flag = "  <- review" if pf < 0.9 else ""
    print(f"{code:28s} earned {earned:6.0f} h  actual {actual:4d} h  PF {pf:4.2f}{flag}")
print(f"overall PF {te / ta:.2f}")

column forms (each)          earned    252 h  actual  280 h  PF 0.90
deck forming (m2)            earned    833 h  actual  900 h  PF 0.93
wall forms (m2)              earned    288 h  actual  330 h  PF 0.87  <- review
stair forms (flight)         earned     96 h  actual  112 h  PF 0.86  <- review
overall PF 0.91

Wall forms and stair forms are running behind. When Luis asked the foremen why, the answers were specific: the wall forms were waiting for embedded conduit from the electrical subcontractor, and the stair crew had been short a journeyperson for two days. Neither was a question of effort. The productivity numbers are useful only as a way of finding and removing obstacles, and Calder Ridge's supervisors are taught to ask what got in the crew's way before asking whether the crew worked hard enough.

Planning the end of the job

Workforce plans often stop at the peak, but the run-down matters too. As each trade finishes, workers must be released in an orderly way that respects the collective agreement's layoff provisions and the employment standards notice rules, keeps the most versatile workers for the punch list, and gives good workers a path to the next project. Calder Ridge holds a monthly meeting during the last six months of each project to match workers finishing on one job with needs on the next, which has done more for retention than any bonus.