Nailing the numbers

I recently ran into a customer who thanked us for our work on a project. Though he said our labour was wrong, so he had the team double it. He then proceeded to tell me that they lost the $3m project by $300k. Afterwards I wondered, maybe our labour wasn’t so bad?

I get that quite a bit; builder’s presenting their labour as fact. Whereas the reality is, it is just as much an estimate as ours. So, who is right? The builder’s labour based on experience or our detailed labour build up refined on nearly 4,000 projects.

Labour is people doing tasks. Estimating labour comes down to three steps:

1. Determining the tasks involved.
2. Establish what factors drive the work on the task.
3. Measuring those factors.

These three steps underpin all common labour estimating methods. The main difference is how much you roll up the tasks. The more you combine tasks, the simpler the job appears but the more detail is lost and the broader the assumptions become.

Common labour estimating methods

Based on a project schedule - Estimate the number of weeks for each aspect, multiply by the number of staff on site.

Good

• Quick to do
• Easy to understand
• Scalable, big and small jobs

Bad

• Requires experience on similar projects.
• Can be very inaccurate.
• Highly subjective. Varies widely between estimators.

Based on unit rates - allow a certain amount of time per building element.

Good

• Reasonably quick once set up.
• Easy to understand.
• Scalable, big and small jobs.

Bad

• More time-consuming to measure.
• Requires manual input to reflect job specifics.
• Can ignore project complexity.

Based on materials - allow a constant amount of labour against each item in a materials schedule.

Good

• Easy to do.
• Reasonably detailed.
• Materials can be a good indicator of work.

Bad

• Measure is time consuming.
• If materials are wrong, so is the labour.
• Missing non-material activities.
• Does not consider complexity.

Bottom-up task analysis - we call this activity-based costing.

Good

• Very detailed.
• More closely reflects real-world.
• Highly repeatable.

Bad

• Very time consuming.
• Requires a lot of set up.
• Requires information for each construction system.

Our experience with ceiling batten labour

To start with we used hours per m2 sourced from a book, NZ Construction Labour Constants.  It was easy. Measure the room area and times by the factor and you have your hours.

Then builders complained that we weren’t allowing enough time on small rooms.

What we did to improve

We could have used hours per metre of ceiling batten, but that would still understate small rooms.

We could have looked at the house and thought, “mmm that will take 4 people 9 days”. But how could we train all our team to do this?

Instead, we used activity-based costing.  We analyse the activities and build up the labour times based on these:

1. Get the materials into the room and set up dumpy level
        - x minutes per room
2. Install the perimeter channel/batten
        - x minutes per metre of perimeter
3. Install battens
        - x minutes per m2

Of course, reality is more nuanced than this.  There are other factors like materials and build complexity to consider.

Use the power of computers

This approach is very time consuming to do with traditional estimating methods. By harnessing the power of computers and our 3D model, it takes us no time and gives very accurate labour hours that respond to the design. 

A last thought - Track what happens on site and give feedback to your estimator. You can only improve estimates by closing the loop.

Nick Clements, MNZIQS RegQS ICECA

Nick won the inaugural NZIOB 2024 Digital Technology award for his work on 3D estimating in residential construction.  Nick is a Registered QS and Chair of NZIQS Auckland Branch.

His business, YourQS, specialises in providing cost estimating services to residential builders, architects, and homeowners for both new build and renovation projects. They have completed over 3,200 projects since launching in 2019 working for around 300 builders nationwide.