I was five years old. My father had just put his life savings on the line to start a wire manufacturing business and build his first factory. On most days, he’d have left for his construction site before I woke up for school and would be back only around dinner. His colleagues would often dine with us. Our dining table resembled a war room, energized with discussions to complete the construction on time.
Every Sunday, I would accompany him to the site. I loved learning about the giant lego project and perhaps, more importantly, getting to spend time with my Dad. A couple of months in, I noticed that his fiery optimism was replaced with a tense, grim demeanor. I learned from my Mom that Dad was worried about delays in the project.
The project was debt financed, which meant that we’d lose our life savings and our home if production didn’t start on time. As children, we think of our parents as invincible; capable of protecting us from the evil world. The day my Mom switched our cereal brand for a cheap knockoff — my safety bubble burst. That feeling stayed with me.
I started Doxel because my Dad’s experience was not unique. Even in 2018 — years after my Dad’s factory was finally delivered — 98% of large projects continue to be delivered, on average, 80% over budget and 20 months behind schedule.
Doxel is artificial intelligence that stops that from happening.
Within two short years, owners like Kaiser Permanente, and general contractors like DPR Construction are already using our solution.
Why do projects experience cost overruns and schedule delays?
Low labor productivity.
While productivity has more than doubled in manufacturing since ’95, productivity in construction has stayed flat. In the U.S. productivity hasn’t significantly changed since the Roosevelt administration.
The process of manufacturing gains from its inherent repeatability. Construction, on the other hand, is quintessentially unique — every project is different, every design, every material, every team.
But here’s the kicker — along with the benefits of repeatability, most manufacturing processes have real-time reporting to employ measures to counter falling production rates. Thousands of embedded sensors relay production rates to a control room, alerting operators if failures are detected and enabling them to react to inefficiencies in minutes.
Unlike their manufacturing counterparts, construction managers don’t have real-time feedback on progress and quality.
While decision-makers often know how many person-hours are being spent, they don’t have an accurate measure of how much work is accomplished, or if that work is even being done correctly. The reason for this is apparent — most projects involve several teams that are installing millions of components to high accuracies. With so much to inspect, it’s nearly impossible to accurately track progress. Before they know it, the managers realize that months worth of person-hours and millions of dollars have been overspent, leading to nasty overruns and delays.
I don’t blame them. Just take a look at how complicated a construction project can be — they have to build this for the first time, build it correctly, deliver on schedule and do it all without any feedback.
Doxel uses AI-based computer vision software to give managers real-time feedback on schedule, budget and quality.
It’s the sophisticated control room of manufacturing, built for construction.
How do we do it? Doxel uses autonomous devices to scan every inch of a site on a daily basis with LIDAR and HD cameras.
Our proprietary AI algorithm processes the visual data, inspects installation quality, and quantifies how much material has been installed correctly.
Finally, our cloud-based dashboard ingests data on person-hours spent on the job, compares it to progress measured by AI and predicts a project’s cost-at-completion and completion date based on current productivity (2 minute video on how Doxel works below).
Doxel is the canary in the coal mine for construction overruns.
How valuable is that? In one of our projects, our real-time progress tracking led to a 38% increase in productivity and helped the project be delivered 11% below budget (read the full case study here)
What’s transformative about this?
Comprehensive data capture: Capturing data indoors — where 80% of a construction budget is spent — has been a major challenge, until now, owing to the complex nature of construction sites. Doxel uses drones for outdoor, and ground rovers for indoor tracking.
AI-based automation: Majority software vendors out there simply use visualization software to overlay captured data onto a 3D design, leaving progress tracking to humans. Managers would rather walk around a jobsite than sift through terabytes of photos and laser scans.
Our breakthrough AI software can automatically analyze visual data to measure installed quantities and inspect quality. This game-changing end-to-end solution produces accurate progress and quality reports in real-time.
That’s how Doxel keeps molehills from becoming mountains.
Cross-referencing with budget and schedule: Our software is smart enough to mine through thousands of line items in the project budget and schedule, and update every line item with measured progress. This hyper-granular, real-time reporting delivers an X-ray vision on where a project stands with respect to budget and schedule.
A Controller at a global general contracting firm told me they’re spending about $11 million every day on construction projects.
She told me that if they’re 90% accurate on progress assessments, they’re having a very good day. Meaning that $1.1 million is getting spent every day on unaccounted work.
That changes today.
With Doxel’s solution, they’ll never have to guess again. Multi-billion dollar construction giants will now know exactly how much they owe for a day’s work.
Behind this software is a breakthrough 3D semantic understanding algorithm that was built by my professional soulmate and co-founder, Robin Singh, and the wonderful engineers who’ve chosen to join us on this journey.
Indoor construction environments are challenging for computer visions algorithms because of the density of components, clutter and ruthless lighting. Some of the challenges are occlusion, disambiguating junk from installed materials and visibility.
Robin and our team created a deep learning algorithm that classifies objects in real-world construction environments at a remarkable level of reliability.
We realized that the amount of data we’d need to do this with conventional 2D computer vision techniques would be gargantuan. We turned to 3D computer vision, because theoretically, a 3D neural network would require fewer datasets for training than 2D. This was uncharted territory since, at the time, there was no publicly available work which demonstrated reliable classification with real-world 3D data.
Today, I’m happy to say that the network is alive and performing phenomenally well.
I’m also thrilled to announce, that Andreessen Horowitz has led a $4.5 million seed investment into our company and Lars Dalgaard, GP at the firm, is joining us in our journey.
As the Founder & CEO of SuccessFactors, which he grew from nothing to a $3.7 billion acquisition by SAP, Lars is an enterprise software legend who’s often credited with orchestrating the first major SaaS acquisition. But that’s not the only reason why we chose to partner with him. What’s more important is that he relishes feisty conversations, puts heart into his deals and is just as audacious as we are.
I’m also grateful to our early backers, Alchemist Accelerator, Pear.VC, Steelhead Ventures and SV Angel who’ve been an integral part of our mission.
My story had a great ending — my father successfully paid all the loans and went on to acquire multiple factories.
Not all stories do.
A project that runs over budget can cause job losses, increase in healthcare, transportation and energy costs, and higher taxes. And sometimes, just sometimes, it can cause a five year old to grow up faster than he should.
Here’s hoping we can change that.
With lots of love and dreams for the future,
Chief Executive Officer, Doxel