With the flood of new technology solutions to improve space usage and management, “optimize” seems to be the word of the day. As one example, the business press highlighted a recent acquisition target as providing “software that optimizes offices.” In fact, the acquired company provides meeting room scheduling, admittedly very good scheduling, but not what most data scientists would define as true optimization and not what most of us would call “offices.” Similar examples abound in real estate and workplace software descriptions.
Certainly new workplace software and tools are valuable. Sensors, IoT systems and software, and mobile apps that show workers’ locations provide better data on space usage. Office and meeting room schedulers improve efficiency in finding workspaces and coworkers. These tools can show opportunities and provide “actionable insights.” In simple situations, this information may be enough for optimization. In complex situations, where there are many different ways to take advantage of these opportunities, much more is needed for “optimization.” Overlooking other better solutions can cost millions.
For example, when the data suggests that an activity-based workplace might be more appropriate, are workplace strategists finding the best combinations of desks and conferences rooms of various types and sizes, or simply a better combination? Or, when planners learn that some business groups need less space due to new workplace strategies and others need more space due to increasing demand, are they finding the facility and occupancy strategies that save the most money while also maintaining, or potentially improving, business productivity, or are they just saving some money, leaving millions of dollars on the table?
As real estate users face increasing competition, rising real estate costs and constant pressure to be more responsive, most organizations don’t have the luxury to be satisfied with just improvements. They don’t have the time for slow, inefficient manual processes to make decisions and they want to be able to find solutions themselves rather than call a consultant.
Most new real estate technology improves data collection and process efficiency. There are few advances that address decision making. Real estate and workplace planners are making more “data-driven” decisions, but the tools they use are essentially still “manual,” relying on basic summary statistics, dashboards and spreadsheets.
Decision Making and Optimization in Other Disciplines
In many disciples, better decision-making tools were adopted decades ago. We’ve seen much less adoption in corporate real estate, which includes office, R&D and other space types for knowledge workers. These space types include corporate locations, as well as educational, medical, government and non-profit workplaces.
In flight scheduling, inventory management and even NFL game scheduling, planners use mathematical optimization to find the “best” solution. With this approach, computer algorithms search through the many different alternatives to find the ones that best meet the business goals and constraints. Imagine how much worse our airline experiences would be if airlines scheduled flights using the same tools that most organizations use for real estate and workplace planning — post-it notes, spreadsheets and graphic drag-and-drop.
Mathematical optimization falls within the broad Merriam-Webster’s definition of artificial intelligence (AI) tools, that is, software and tools with the ability “to imitate intelligent human behavior,” but uses different algorithms than most current AI applications. Technically, these algorithms are called “mixed-integer linear programming” or “constraint programming,” but we can leave these details to the experts.
Military agencies have a long history of using these techniques, beginning after World War II, when planners wanted to find the lowest cost diet to provide soldiers with adequate nutrition. More recently, the Army used mathematical optimization in their 2005 Base Realignment and Closure (BRAC) analysis. This approach provided defensible data on alternatives to use as background for political discussions. The GIS team at the National Aeronautics and Space Administration (NASA) Langley Research Center (LaRC) in Virginia has used a combined GIS-Optimization model to reallocate space usage across its 3.7 million square foot campus with more than 300 buildings.
These techniques have also been adopted in other real estate classes, where outcomes can be measured more easily. Logistics and supply-chain management applications optimize manufacturing, warehouse and distribution capacity and locations. Analytics for retail sites, service centers, and fire and police station locations can identify how to reach customers and residents most cost-effectively.
Optimization for Corporate Real Estate
For more than twenty years, I’ve been using optimization models for real estate and workplace planning on consulting projects for companies including Google, Cisco, US Bank, and PacifiCare Health Systems (now part of United Health Care Systems). With my Core Planning software, we’ve found millions in overlooked cost savings and productivity improvements, all in minutes rather than the hours and days needed for traditional approaches. Projects can range from tactical blocking and stacking to multi-facility strategic analyses for campus, regional and national plans that support mergers, acquisitions, growth, and consolidations. Recommendations include which locations to acquire, dispose, renew, cancel, reconfigure and renovate and which business groups to relocate.
Our algorithms evaluate data on business growth and requirements, space utilization, facility costs, conditions and configuration, lease expirations, disposition alternatives, capital improvement requirements, and real estate market conditions. When appropriate, we include salaries and benefits, hiring, training and severance costs. A typical large project could evaluate these factors for twenty business groups across fifteen buildings to find the best solution that meet business requirements in just minutes. With so much computer processing speed and efficiency, computers are much faster than people at these jigsaw-puzzle-like problems.
Other CRE decision-making applications use optimization and simulation algorithms to test different workplace layouts and determine the best number of different types of offices and meeting rooms and the best number of exam rooms in hospitals. These tools can also balance the additional costs of short-term leases, cancellation options and co-working spaces with the chance that space in a long-term lease won’t be needed in the future. Predictive analytics helps building managers identify when to replace or provide maintenance for building equipment and how to reduce energy usage.
Credibility, Speed and Flexibility
As these tools provide better information for decision making, CRE teams can build more credibility with senior executives. On a project to consolidate the headquarters operations of two merging insurance companies, typically an emotionally-charged decision, the company Treasurer said he had never seen a major decision proceed so quickly and smoothly. Our search algorithms showed that the planners’ intuitive recommendation provided the best way to meet business goals. In the executive meeting, senior executives asked about other alternatives. Because optimization techniques automated the most time-consuming aspects of answering what-if questions, we had already prepared answers, providing senior executives with greater confidence in our recommendation and requiring just two meetings for a final decision.
The speed and flexibility of these tools can support a more agile workplace and the increasing recognition that worker adjacency and collaboration bring increased productivity. As one planner noted, with traditional trial-fit planning approaches, it might take a few days to find a good seating arrangement. By then, the business needs may have changed and a new plan would be needed. Optimization software can provide real-time, comprehensive plans that support new business requirements in minutes. In co-working situations, when workers come to the office, planners can develop plans to meet the specific needs for that day in real time.
Overcoming Slow Adoption in Corporate Real Estate
With so many benefits, why haven’t these tools received greater adoption? Until now, a primary challenge has been gaining widespread recognition that these tools can capture all of the relevant data needed for CRE decisions. Planners would often ask in disbelief how computer models could possibly consider the many different factors needed to make recommendations? How can a computer possibly replace planners’ intuition and experience? With the widespread demonstration of artificial intelligence techniques, such as machine learning and neural networks that are applied to speech and object recognition and chat and game bots, more people are open to considering how other mathematical techniques, such as optimization, can be applied in corporate real estate.
One challenge in applying optimization for CRE is understanding what computers can do well, when human input is needed, and how to build tools that integrate computer efficiency and users’ expertise. In its $2 billion “AI Next” initiative, DARPA (Defense Advanced Research Projects Agency) is also emphasizing these challenges. While there have been major accomplishments with AI, there continue to be many limitations in what computers can do. They do some things very well, and for the rest, the computers must become “partners in problem-solving.”
For real estate and workplace planning, computers are very good at evaluating quantitative trade-offs. For example, should we renovate space with a long-term lease or sublease it and move into space that better meets the business group needs? Or, how can we relocate growing business groups to maintain adjacency requirements with the least disruption for high-value workers? And how do we evaluate these decision when there are twenty different business groups in twenty different facilities, with each decision dependent on the other decisions.
The difference between real estate and workplace planning for knowledge workers and for other real estate uses is that we need to evaluate subjective productivity-cost trade-offs. We may know that some business groups would be more productive if they were located in their preferred locations with key adjacencies to the right teams, but will the increased productivity offset the additional costs. Since computer models can’t make these subjective evaluations, we develop a set of alternative “optimal” solutions that meet different sets of business requirements. Then, we rely on decision makers to evaluate the cost-productivity trade-offs.
These approaches are not replacing planners and decision makers. They are automating the tedious, time-consuming activities that keep planners struggling to find a good solution into the early hours of the morning. Automating these tasks leaves more time for more valuable activities including understanding business needs and evaluating the results.
Applying Technical Solutions to Practical Challenges
Another major challenge is combining the modeling expertise of technical specialists with the experience, insights and intuition of the content experts. First, project teams must identify which pain points can be addressed with technical solutions. Then, they must highlight the priorities and nuances so that technical specialists can choose the right techniques and identify what’s most important so that the approach is appropriate and manageable. Being able to bridge this gap can be key to developing software that takes minutes to solve rather than hours.
This challenge is even greater when developing tools that planners can use on their own, without the assistance of optimization experts. In our fast-paced business environment, planners don’t want to have to call a consultant each time they need to evaluate a new situation. In addition to having the right techniques and priorities, these tools must have simple and meaningful user interfaces that capture planners’ expertise and intuition and present the appropriate data in an orderly and understandable way.
Partnering with SpaceIQ, a leading SaaS workplace management software company, we are now working on developing these tools.
With mathematical optimization, planners will be able to move past traditional manual decision-making techniques. They will be able to evaluate large amounts of data, forecast potential outcomes, automatically search through alternatives, test what-if scenarios, and find the best outcomes, all in minutes thanks to advanced search algorithms and data processing efficiency.
We’ve talked about data-driven decisions for years, but the decision-making tools planners use are often still spreadsheets and simple summary statistics. We won’t be able to truly optimize until we start adopting better decision-making tools as well.