Presented at the OECD expert meeting "Construction Risk Management in Infrastructure Procurement: The Loss of Appetite for Fixed-Price Contracts", held on 17 May 2023 at the OECD, Paris and online.
2. 2
More than 75% of projects are late with 50% late by at least 5 months
3. 3
Why do so many projects fail to finish on time? Put simply, complexity and bias.
Incredibly
complex
project
schedules
Overheads and time-
bound costs spiral
Contractor’s profit
margin squeezed and
quality affected
Trust between
contractor and owner-
operator breaks down
Business case for project
is weakened or collapses
Owner-operator’s
business is devalued
Legal issues ensue with
liquidated damages
awarded
Reputational damage
hurts everyone involved
Contracting and
baselining
exercises
undermined
Unexpected
delays
materialise
without warning
Key milestones
missed again
and again
Failure to
mitigate risky
activities
Inaccurate
forecasting
Salience
bias
Availability
bias
Optimism
bias
Failure to
mitigate risks
the project
team hasn’t
factored
Failure to
mitigate
boring risks
4. 4
nPlan’s Unified AI Platform is a data-derived schedule and risk forecasting solution
delivered
via
Machine learning
Past project
schedules
How long might my
project take?
What might go
wrong with my
project?
What can I do to reduce
risk or capture
opportunity?
nPlan
Platform
nPlan
Portfolio
nPlan
Insights
The platform uses past project data and machine learning to generate unbiased forecasts of project outcomes;
results are shared via our Portfolio, Insights, and Bid products.
nPlan
Bid
5. 5
Two key areas of focus are identifying unknown unknowns and providing independent assurance
nPlan seeks to answer the question “when will my project most probably finish?”
● The things that keep you up at night –
what might the team not be thinking of
that you should be? What might derail
all your planning?
Identifying Unknown Unknowns
● Making sure the team is focused on
the most valuable risk areas with a
comprehensive, 3rd party analysis
Providing Independent Assurance
7. 7
Gather
information
about the
project,
context,
objectives
Run the
simulation
Reach
consensus on
probability of
risks occurring
Identify
activities in
schedule to
which to apply
discrete risks
Define, apply
uncertainties to
activities
Run the
simulation
QSRA
nPlan
Obtain schedule
file
Time and effort
Human processing
(Compounding) error
Computer processing
nPlan’s approach solves for QSRA in a nimble, data-driven way that allows for rapid iteration and
scenario comparison
9. 9
Win more work with credible, realistic plans that give clients total peace of mind
Quantify key risks that could consume contingency and profit to protect margins
Establish the optimum contracting model for achieving project objectives, every time
Assurance of sub-contractor & master schedules to correctly set expectations and deliver
Implement mitigation strategies for different risks to control schedule and increase margins
Deliver more active risk management without straining finite resources
Benefits depending on the project dynamic
10. Has amassed over
75 Construction schedules
creating the biggest
dataset of its kind in
the world
600,000
$22M
In funding from backers
including Google Ventures, the
founder of Deepmind, and the
former CEO of McKinsey
Secured Customers include:
Founded in the
UK
in 2017
Team of
- and growing
Hypothesis: we can use AI to identify
risks and forecast projects
Taking optimism bias (and a bunch of other biases)
quite literally out of the equation and going a long way
towards solving the problem of delayed projects
11. C A S E S T U D Y
How Rail Projects Victoria used AI-powered
forecasting and risk management to choose
the best tender for its South Geelong to
Waurn Ponds Duplication Project
12. About the Project
Project Name:
Project Type:
Project Value:
Works to start:
South Geelong to Waurn Ponds Duplication
Railway construction
~$900M AUD
Late 2022
13. 13
nPlan’s AI learnt
how rail projects
are delivered
RPV shared bid
schedules with
nPlan
The bid schedules
were processed in
nPlan Project
RPV used nPlan’s
findings to evolve
the bid schedules
nPlan processed
the iterated bid
schedules
1 2 3 4 5
How nPlan worked with RPV
RPV selected
a tender
6
17. 17
100x
ROI of client
investment in nPlan so
far:
Value of unmitigated
total risks identified by
nPlan @ P80
Value of risks identified
and actioned so far:
$19.4M
RESULTS AT TENDER AWARD
$8.25M
18. 18
Benefits for Rail Projects Victoria
Easily assess
whether schedules
are realistic
Benefits for chosen contractor
Quantify project
risk contingency
Compare bid
schedules with
quantitative data
Develop risk mindset
from the tender phase
Formalise key
milestones which work
for both contract
parties
Rewarded for
developing a
rigorous, realistic
schedule
Start project from
position of trust
with owner-
operator
Start negotiation
phase with a de-
risked schedule
Receive ongoing
insights from nPlan to
help with identification
of risks
Build trust with
contractor partner
before a contract is
awarded
20. We have defined a new category;
Algorithm-led Assurance
Our Algorithm-led Assurance product is being used by
contractors, owner-operators, and government agencies.
We’ve proven we can acquire data at an incredible scale,
and provide valuable services today.
This has created a big and new
opportunity for us in insurance
Through the largest capacity providers in the world, we can
underwrite the outcomes of construction projects. We will
do this through a world-first product; project over-run
insurance, starting with a $500m capacity.
Our first product is scaling up
20
21. 0
2
4
6
8
10
12
14
16
18
20
22
We are developing the world’s only
project over-run insurance
21
● The Owner or Contractor enters an agreement
with us (the Insurer), against the impact of
severe delays.
● nPlan’s assurance product is mandated as a risk
control measure, and provides the risk
parameters for the insurance to be priced.
● The product payout is based on a time trigger,
contingent on sufficient conditions remaining
true throughout the project.
Jul
2023
Oct
2023
Jan
2024
Mar
2024
Jun
2024
Sep
2024
Dec
2024
Feb
2025
May
2025
Aug
2025
Nov
2025
Jan
2026
Apr
2026
Jul
2026
Oct
2026
Dec
2026
Mar
2027
Jun
2027
Sep
2027
Dec
2027
Feb
2028
May
2028
Likelihood
(%)
Most likely Insured Range
End date
22. This is a big opportunity, and we ran a backtest with our reinsurance
partners on our data to quantify it
22
Project Size
Between $30m and $1b
Dataset
10 Years of data, 16985 projects
(~1% of the world’s total projects
in that period)
Underwriting
15% of project cost
0.8% premium
Capacity
$815b of total underwriting
$10b capacity requirement
$6.5b of premiums
$1.6b total payouts
PnL
$2b management fees for the
MGA
$3b syndicate profit
30% return for syndicate
23. The next steps for Insurance is to build the initial basket of projects that require
underwriting
- The basket must be sufficiently diverse, demonstrating sufficient non-correlated
risks
- The exact policy terms and conditions to be agreed and finalised
- Fill the capacity requirements based on the basket of projects and quantum of risk
to be traded.
If you would like to discuss partnering on a
project or portfolio of projects - please let
me know