Saun: “ . . . at this point we [are] still struggling to understand the info needed. I think it is time to call it a day. I don’t want to waste your time or ours.”
Fletcher: “Saun, I was going to suggest the same. It’s very simple stuff.”
And, that is the closing e-mail exchange with DSLD’s Saun Sullivan, as he and I mutually agreed not to further waste each other’s time trying to analyze the numbers behind the company’s operating performance.
By all industry accounts, DSLD Homes is an emerging force in homebuilding. Only in existence since 2008 – a mere six years – the company is lighting it up from a revenue/closings growth perspective. DSLD is already an NHQ silver award winner, has already been named Builder of the Year by Professional Builder, and is listed at No. 25 in the 2014 Builder 100 (No. 39 in the 2014 PB Housing Giants).
The history of success is actually longer than DSLD’s short existence: prior to starting DSLD, managing partner Saun Sullivan had started another homebuilding company; by 2006, he and his partners had sold six year-old PCC Home Builders/Provident Homes to the biggest homebuilding enterprise in the nation, DR Horton.
This was to be, at most, an informal engagement, the start of which consisted of a casual suggestion and an alternative to consider.
In March, on his own initiative, Ryan Nash, DSLD’s Product Development Director, attended the inaugural Pipeline workshop. Two months later, in May, Saun sat in on the breakout session I led at the Housing Leadership Summit. At the end of the HLS session, Saun and I had a brief discussion, during which he suggested that I make arrangements to come to Denham Springs, and see whether we could work together.
When I later emailed Saun the PowerPoint presentation and some links to other SAI resources, I proposed a different approach: “I am certainly interested in what you have in mind. But, let’s start with something that won’t cost you anything. Send me your monthly sales, starts, completions, and EOM work-in-process for some meaningful period of time; I will tell you what I see. Give me the data for 24 months.”
Saun turned the matter over to Ryan. Over the course of two months, involving numerous e-mails, telephone calls, and data files, I worked with Ryan to get the data; I explained what I needed and why I needed it. The data quest was adjusted to fit the context of what Ryan felt their management technology system could provide, in deference to the Rayco-inspired way that DSLD regards its production process and says it mandates even-flow.
Throughput – the output from the process – did not seem to be a problem, but the historic work-in-process proved impossible to obtain. I checked my explanation; it was clear, exactly what we teach at every Pipeline workshop. Yet, every data point DSLD submitted turned out to be a throughput measure. Not only was the work-in-process data apparently unavailable, but the data DSLD did submit indicated that they did not understand what periodic work-in-process is, that it is consistently-spaced point-in-time measures that have to be averaged – in some fashion – to produce a periodic measure, a measure over time.
Without those numbers, there was no way to look at the performance of DSLD’s production system. To be sure, there were DSLD’s assertions of even-flow production, assertions of fast, completely reliable cycle times, but no way to measure any of it.
Without access to those numbers, none of the long-accepted methods for gauging the operating performance of any production process could be used.
Even-flow? There was evidence that suggested DSLD’s production was not quite as even as was being asserted, but there was no way to determine whether it was or wasn’t. Every production system in existence will buffer itself from the effects of variation, in one of three ways. Even if DSLD was attacking variation in its process, there was no way that all of the variation had been eliminated; and, there was evidence of variation in the data that was provided.
How does DSLD’s production system handle that variation? Is it being buffered with higher inventory? Longer durations? Excess capacity? Some combination of all three? I have my theories, but there is no way of knowing the answer, without the data.
When we look at a production system, these are the kinds of questions we want to answer: How many completions can we produce with a planned, finite, and controlled number of houses under construction? How many houses do we need to have under construction to achieve our desired rate of completions? Do we have too much WIP? Not enough WIP? What cycle time do our job schedules need to meet in order to produce the desired rate of completions with a planned, finite, and controlled number of houses under construction?
Consider the matter of DSLD’s cycle time. One of the principles of production physics is Little’s Law. In manufacturing, it is commonly used to measure the rate of production, expressed as x-number of units per period of time (for example, 1,000 units per hour); less frequently, it is used to measure cycle time, expressed as x-number of periods for each unit (in this case, 3.6 seconds per unit).
When I first learned about Little’s Law, it was in the context of how it was being used at Motorola, and I wondered if the same principle could be used in homebuilding. As it turned out, the only adjustment was to replace the completion rate with a throughput number, and specify the time period.
Little’s Law results in what we call calculated cycle time, as opposed to measured cycle time, which is the mean (average) duration of a specified range of production units. Measured cycle time has its forensic value in solving problems, eliminating errors, and removing waste from a production process, but calculated cycle time provides the only true, useful picture of homebuilding production as a system.
The variables in the equation are inventory and throughput, and if any two of the three measures involved are known, Little’s Law can be modified to solve for the third.
I gave DSLD this example of annual (360 day) cycle time:
CT = 120 days WIP = 80 houses T = 240 homes
CT = (WIP ÷ T) x Days
(80 ÷ 240) x 360 = 120 days
WIP = (CT x T) ÷ Days
(120 x 240) ÷ 360 = 80 houses under construction
T = (WIP ÷ CT) x Days
(80 ÷ 120) x 360 = 240 closings
The data I was receiving from DSLD was producing shorter cycle times than even DSLD’s job schedules stipulated; Ryan conceded that those cycle times were not possible, based on DSLD’s “two separate straight-line even-flow schedule models” used for their frame start-to-completion sub-process (narrower scope than a start-to-completion process); and, he gave me the cycle times of both scheduling models, in terms of work days and calendar days.
I then told Ryan what DSLD’s WIP would have had to be in order to produce that many completions in the period of time we were examining; DSLD couldn’t say what their work-in-process had been during that period.
I told Ryan that I wanted to see what DSLD’s cycle time was for the frame start-to-completion sub-process, based on the actual data on completions and on the average number of homes in that phase, for the 24 monthly periods we selected. I told him I was not interested in assertions; I wanted something I could verify.
I told him that, going forward, we could look at production in any or all of DSLD’s sub-processes, but we should get the numbers on frame start-to-completion done first, because that sub-process was reported to be the most stable, with the least variation. I said, let’s see what we have, and we can move forward from there.
At that point, I also told Ryan that we had been at this task for almost two months, and I had been clear from the beginning on what I needed. I reminded him that I had told Saun that my insight would cost DSLD nothing. I was contributing my time, because I was intrigued with what DSLD said it was able to do. I told Ryan this was not rocket science, and to please tell me that DSLD’s management technology system knows how many units are at what stage of completion at any point in time. I told him to stop wasting my time, and get me what I need.
This story is about DSLD Homes, only from the standpoint of the question that it raises: If this is the situation at a homebuilding enterprise with phenomenal growth and industry recognition, with a management team that benchmarks relentlessly, that has made the effort to enforce even-flow production, what does this mean for anyone else?
No doubt, there would have been valuable insights gained from looking at these numbers; that has been the case with almost every previous opportunity we have been given to analyze operating performance. These insights could have validated DSLD’s assertions, or perhaps refuted them; they could have bolstered DSLD’s convictions, or perhaps convinced them to look further, to push deeper into an understanding of the production principles that drive operating performance and business outcomes.
It didn’t happen.
We should learn from every outcome, whether good or bad. We don’t learn when we are ambivalent about outcomes. Given their success, given their achievements, given their reputation, given what they consider to be important, I am betting their inability to provide this data, and the resulting lost opportunity to analyze homebuilding production as a system is making DSLD Homes mad as hell.
The next Pipeline workshop will be held at the Ponte Vedra Inn and Club, Ponte Vedra Beach, Florida, on October 15-16, 2014. Cost is $795.00. Early registration, open through August 15, 2014, is $645.00.
Delivered by SAI Consulting. Sponsored by BuilderMT and Big Builder (Hanley Wood).