StruXure - Industry Trends and Insights
Over the past 25 years, the construction industry has evolved dramatically. Over the next 25 years, it may be unrecognizable. From handwritten notes and graph paper takeoffs to automated estimates and Building Information Models (BIM) the industry has come a long way, but there is still room to innovate.
Serious builders must own their data to control their destiny. In this brief, I outline three critical reasons why. A basic understanding of statistics and normal distributions is assumed. For a refresher, reference the links in the sources. Let’s dive in…
This article also appeared on LinkedIn? Read it there: [link]
Each project provides an opportunity for recursive improvement. The critical path of the schedule should dictate what facets of the job must be completed sequentially Vs. the tasks that can occur in parallel. Contracts that were written poorly should be corrected. Terms that were overlooked or omitted should be added. Vendors who performed poorly should be noted and systems should automatically flag them when they bid work for a future job.
"The vendor claimed the crane was onsite by a certain date, but aerial photos of the jobsite, taken daily, proved the crane arrived days after the vendor claimed. The dispute was put to rest, the owner saved money ($35,000 per week), and the job continued as scheduled."
Learn how struXure.co handles daily reports and phots.
On the procurement side, items that have a long-lead time must be ordered early (only after plans are finalized). Any potential changes should be flagged if they change designs on previously ordered items.
Similarly, profitable jobs and team leads should be studied. Superintendents who solve problems on the fly are invaluable. Project engineers who properly document daily reports can pay for themselves in a week. A StruXure.co client was paying $35,000 a week to rent an extension crane for use on a jobsite. The vendor claimed the crane was onsite by a certain date, but aerial photos of the jobsite, taken daily, proved the crane arrived days after the vendor claimed The dispute was put to rest, the owner saved money, and the job continued.
These are simple concepts in theory but difficult in practice. A midsize general contractor may have anywhere from 10 - 50 jobs running simultaneously. The complexity begins to compound when you factor in subcontractors, architects, and owners.
Knowing that a job is 10% into the defined schedule of values, but has an excessive # of RFIs created by the building team, is a red flag. How would a builder know this?
There are several ways, but the simplest would be by comparing their historical data to the current project. If the number of RFIs is two or three standard deviations outside the norm (for a similar size project with similar duration), the project warrants closer inspection.
In a normal distribution (a standard bell curve), data splits into a predictable pattern:
A project two standard deviations outside the norm, means it is relatively rare (5%) of all projects completed (make sure you have a statistically significant dataset (at least 30 viable observations).
Data can be sliced and diced a variety of ways, but projects could be grouped like-for-like, for example by:
Businesses either grow or die. Stasis is not a viable option.
We are still in the early innings of the AI-revolution. Frontier-labs and established software vendors recognize the only way to differentiate their models is to access data that others cannot (i.e. not publicly available). Like other raw materials (oil / gold), data has massive potential value, but it must be refined and processed after it has been collected, to fully recognize the value. Proprietary data is scarce, so the ability to train your internal model on your unique data-set will result in better outcomes.
Builders, architects, and owners should be aware of this and weary of any companies accessing, monitoring, and monetizing their data.
Seemingly every week the major AI players release a new model touting increased capabilities. In addition, open weight models (like Kimi) allow anyone to download, run locally, and customize their own LLM. Open-weight models are artificial intelligence systems where the trained numerical parameters, or weights, are made publicly available.
Having control of your data - the ability to download, disconnect, and re-direct an AI language model to your dataset will allow the maximum flexibility to adapt to a rapidly changing environment.
Sources: