Dynamic Capacity Sizing (Timing & Expansion)
How to increase capacity?
Improve Processes
Outsource (some) Production
Extend Existing Facility (Brownfield)
Add shift to Existing Facility
New Facility (Greenfield)
5 Capacity Timing Strategies
Leading or Lagging
Hybrid timing
Demand chasing (ideal but rare)
Follow your competitor - less dependent on own forecasts but needs confidence in competitive position
Ideally, w/o frictions, timing is a non-issue: track/chase demand
Temporary workers
Changing shifts to move capacity from one time slot to another in weekly or monthly shedules
e.g., Walmart uses 10 min segments to schedule shifts
Real life frictions
Lead time complicates fast, lumpiness precise, economies of scale (EoS) cheap, and ireversibility reversible adjustments
The result is that capacity dynamics often involves fairly large increments & fall between 2 extremes
leading
lagging
Leading & Lagging Timing Strategies
Decisions to make:
Time (when)
lead or lag
Size (by how much)
many small or one big
Comparison
Advantages of Leading
Aids market development
Competitive barrier to entry
No shortages
Allows signalling & credibility
High service levels
Not sensitive to start-up problems with new capacity
abundant capacity = good response
more likely to observe true demand, can exploit "up-side" of forecast
Advantages of Lagging
Delays capital expense
Less dependent on accurate forecasting, but difficult to measure excess demand if there is no waiting list
Option value of waiting
New and/or cheaper technology may be available in future
Low excess capacity risk
Increased control of channels & customer relaitonships
High cost efficiency
no under-utilisation
Hybrid Timing Strategy
between leading & lagging: Smoothing
2 types of smoothing
Inventory
Backorders
build inventory with excess capacity to protect against better than expected demand increase during capacity building lead time and/or delay expansion a bit
sell capacity before you have built it, customers join a "waiting list"
Economies of scale in Capacity Investment
2 Capacity Investment Cost Models
1) Linear CapEx function
C(K) = 
2) Power CapEx function
C(K) = with 0 < α < 1
Scale Economies in Capacity Expansion
Constant Marginal Cost
Decreasing Marginal Cost
Scale economies result from fixed cost being spread over more units
Scale economies result from spreading of fixed cost & volume discounts, installation, learning, etc.
Capacity Timing Drivers
A classic timing model - no residual uncertainty (level 1)
Basic Timing Model
Linear demand growth
CapEx cost function
Continuous-time discount rate r
D = gt
C(K)
Decision Variable: timing t (size K = gt)
PV(t) = =
Optimal timing condition for power CapEx C(K) =
Capacity Timing Optimisation Drivers
insights from the basic timing model
Driver change:
Greater scale economies (α ↓)
Driver change:
Smaller discount rate (r ↓)
Driver change:
Greater demand growth (g ↑)
Capacity size K ↑ & Timing t ↑
Cost of one expansion C(K*) ↑
Capacity size K ↑ & Timing t ↑
Cost of one expansion C(K*) ↑
Capacity size K & Timing t
depends on investment cost
Impact on cost of one expansion: depends
Some benchmarks from the model
(linear growth rate)
For strong economies of scale
(power CapEx, alpha = 0.57, c0 = 0)
r = 10%
expand every 10 years
r = 20%
expand every 5 years
Shape of the cost-curve
(as a function of capacity size)
small deviations from optimality do not affect costs much
it is better to err on the side of more capacity
Waiting for better information
Waiting is valuable if
Waiting is problematic if
the value of waiting lies in more accurate forecasts
during the waiting period companies may perform several tests or observe what happens elsewhere
to achieve better forecasts
investment opportunity does not disappear
future capacity costs less
uncertainty can be reduced substantially
e.g., success in one market informs success in others
e.g., there is a long enough product life cycle
first mover advantage is important
production learning (cost reduction through experience) is important
Strategic vs. Tactical
Strategic Timing
Tactical Timing
often suppresses tactics such as inventory carry-over
Tools
captures key capacity investment drivers (economies of scale & uncertainty) over longer time frames
Maximise NPV in models without uncertainty
Real options: option value of waiting
often suppresses economies of scale or uncertainty
Example Tool
captures key adjustment tactics (capacity-inventory-waiting triangle) over shorter time frames
with hard-to-flex capacity given
helps decide when to build inventory, maintain a backlog, or make easy-to-flex capacity adjustments (overtime, sub-contracting, temporary employees)
to maximise profits (sales, inventory, backlogs, hiring, overtime, subcontracting)
most models assume error-free demand forecasts but safely inventories and safety capacities can be added to meet service level & responsiveness requirements
Leadtimes, Irreversibility, and Service Levels
Irreversibility
Practical way to deal with uncertainty & growth
Capacity Leadtimes
with uncertainty, a leading strategy is made more difficult
note: "safety" or "buffer" (capacity, inventory, waiting) used to avoid lost sales is a combination of building early & building larger (capacity), and preparing to use inventory or backlogs (waiting)
because it requires forecasting further into the future
The more irreversible adjustments are, the larger the "hysteresis" or region of inaction is for capacity adjustments
i.e., they cannot be undone without losing a substantial part of investment
Each asset type may have its own hysteresis
estimate an optimal expansion time t* using the classical timing model with discount rate & economies of scale
in a model, create leading, lagging, and hybrid expansion options using the optimal expansion time* for the most likely demand-growth scenario
use a decision-tree calculation to decide. whether you should wait for more information
use simple trial and error improvement strategies to check if you can improve expected NPV further in order to find the amount of capacity to invest in now
in a model, model uncertain demand with multiple demand scenarios
find strategy with highest Expected NPV
use newsvendor analysis as a check
when calculating capacity investment costs, consider responsiveness requirements, in make-to-order, service, or fast innovation environments by using a maximum utilisation
if marginal cost of capacity/operating profit for marginal capacity unit is much less than 0.5, the leading strategy should be more attractive
e.g., 85%
Guidelines for dynamic capacity strategies
1) Align with overall strategy & product life cycle
Example of modular capacity & construction
Pharmadule
about the company
think about capabilities related to service levels, responsiveness, quality, (→ leading) and cost (→ lagging) together with the overall pace of change (high → smaller t )
2) Update forecasts to reduce uncertainty
3) Streamline project management (reduce capacity lead-times)
work in parallel + do activities independent of timing & sizing in advance
eliminate activities and reduce activity times & rework
use standard design/ re-use previous designs
4) Add flexibility to allow responsiveness to demand
a) modularise capacity & construction ( to reduce lumpiness)
b) increase adjustment flexibility (low fixed costs, reversible)
c) adaptable & flexible capacity (can be repurposed)
keep initial scope generic/ flexible to allow for multiple products, technologies
producer of modular production lines to multiple industries, including pharmaceuticals
areas
Swedish company, founded 1986
pre-fabrication shortens the time-line for building plants
design & engineering
modular fabrication
process & utility installations
shipping & assembly
commissioning & qualification
handover & start-up