Executives around the world strive for AI to create new sources of business value. However, a fact-based separation of myth from reality remains rare: Where do we focus? How do we scale AI? How do we achieve competitive advantage?
In this webinar, Philipp Gerbert, senior partner & managing director at the Boston Consulting Group and coauthor of the recent MIT SMR AI report, “Artificial Intelligence in Business Gets Real,” discusses the findings from the report and dispels several myths around scaling and reaping the benefits of AI.
In this webinar, you’ll learn:
How AI leaders differ from laggards in their AI strategies
What drives competitive advantage — and how this affects AI focus
What is different about Chinese companies’ approach to adopting AI
How business and operating models (need to) evolve when scaling AI
There are 335 million users on Twitter, 1 billion users on Instagram and a whopping 2.23 billion users on Facebook. A lot of people who could become your customers are hanging out on social media.
You may already have a social media account, but are you getting tons of new followers each month who are interacting on your platform? If not, you need to step up your social media game.
Whether it's on Facebook, Instagram, Twitter or LinkedIn, growing your social media presence is important to get your business in front of more consumers and make more sales. With some simple tips, you can turn your lonely social media accounts into popular communities bustling with activity.
Get ready for your popularity to explode next year. Here are six tips for improving your social media presence in 2019.
1. Use the right tools.
If you want to improve your social media presence, you need to post on social media often and at the optimal times. In fact, according to studies, businesses should be tweeting 15 times per day. But as a business owner, you don't have enough time in your day to constantly be logging in to all your social media accounts. That's why you need to use the right tools.
Editor's note: Looking for help managing social media for your business? Fill out the below questionnaire to have our vendor partners contact you with free information.
Using tools to schedule posts will help you stay active on social media and save you time. With Buffer, you can manage all of your social media accounts in one place, schedule posts weeks in advance, post at preferred times when the most users will see them, and analyze how well your posts are doing. Using tools to help with your social media accounts will ensure you're posting often enough to improve your social media presence.
2. Focus on customer service.
Social media is the new platform for customer service. Instead of giving a business a call or waiting hours (sometimes days) for a company to respond to an email, many customers who are having issues now turn to social media to get help from brands. In fact, 54 percent of customers prefer social messaging channels for customer care over phone and email.
So, instead of focusing so much on selling, you need to focus more on helping people. Using your social media platforms as a customer service platform will send more users to your social media and make it easier for them to solve their problems. You can even create a Facebook chatbot that can answer questions when you're not around. Customers will also be more willing to buy from you when they trust that you'll help them with their issues quickly.
3. Promote your social media accounts.
Don't expect your customers to know about all your social media platforms and go searching for them themselves – you've got to tell them they exist. If you make it easy for customers to follow or like your social media profiles, they'll be more willing to take action.
Start promoting your social media accounts to your customers and ask them to interact with you on your profiles. You can add social media icons to your site, add your social media feeds to your sidebar, tell people to follow you on social media at the end of your blog posts, and promote your social media accounts in your email marketing. The more people are aware of your presence on social media, the bigger it will be.
4. Pay attention to what's trending.
One of the best ways to create a bigger social media presence for your business is to piggyback on popular social media trends. This will get your profile in front of a ton of users who never knew about your business before and could even help your posts go viral.
You can stay on top of trends just by taking a look at what's going on. For instance, on Twitter you can see the top trending hashtags on the left side on the homepage in the Trends for You section. Look at what hashtags you can use to get more views. While it helps to use hashtags that relate to your business, they don't have to be strictly relevant. If your accounting company can create a fun post for #NationalDonutDay, go for it.
5. Focus on eye-catching visuals.
It can be difficult to get noticed in crowded social media feeds, so you need to make your business stand out. One of the best ways to stand out on social media and increase your presence is to focus on using eye-catching visuals.
Try to avoid using too many boring stock photos. Instead, focus on sharing images that will connect with users. Share high-quality behind-the-scenes photos and people using your products. You can even share your customers' photos of them interacting with your product with their permission, which also creates social proof.
6. Actively engage with your audience.
Would you keep talking to someone who never returned your messages? Probably not. Users won't want to follow you on social media or interact with you if they get nothing in return. That's why it's so important to actively engage with your audience on social media.
If you want to create a social media presence, then your social media accounts need to become a community. That means when someone comments on your posts, you should reply to them or at least like their comment. You can also pose and answer questions in posts, and show appreciation to your followers and customers by thanking them and sharing user-generated content. Engaging with your followers will not only increase your social media presence, but help you develop real relationships with customers.
Get ready to have your best year on social media yet. Following these tips, not only will you get more likes and follows than ever before, but you'll gain more loyal customers too.
In 2008, I was designing advertising products at Google. For the first time in my young career, I was going to lots of meetings, and my job had become as much about convincing, cajoling, and coordinating as it was about designing. My manager told me about a team that was working on the Google Help Forum. They needed a designer, and he thought the project was a good way for me to try my hand at a consumer-facing product. The project itself was unremarkable apart from one feature: my work had to be approved by Marissa Mayer.
Marissa was a VP personally responsible for reviewing and authorizing every change made to Google.com. She was a smart, passionate, forceful, and somewhat feared decision-maker who critiqued, gave direction, and (hopefully) approved proposals from various teams. During our meetings, I often envisioned an upside-down pyramid, where the time of many rested on the decisions of one. We were product managers, designers, and leaders. Each one of us would take Marissa’s decisions back to our team, make a plan, and get to work. Our team members were another layer in the pyramid. Their time also depended on her choices.
We all make choices, every day, about how to spend our time. Most of our decisions are small, but over time they add up, and eventually they become decisions about how we’re spending our lives — and our work lives. When managers are careless with their decisions, it creates big problems for their teams. But when they are deliberate and thoughtful, it can create opportunities and give their teams the time they need to do valuable work.
Below I’ve put together a list of tips to help leaders of all kinds be deliberate with their choices, based largely on my years advising startup founders on product, marketing, and management at Google Ventures — and my subsequent work studying and experimenting with personal time-management techniques for my book Make Time. They fall under three categories: the environment you create, the expectations that you have, and the example your choices and actions set.
1. Treat new tools as debt. Before you add a new product, process, or platform to your company, ask yourself if it’s worth it. There will always be new technologies and processes you can adopt — an app promising better communication, a service promising smarter collaboration. But these products don’t always deliver. And when you’re overeager about trying shiny new things, it can hurt your team more than it helps them. People may become bogged down incorporating a new tool into their workflow, or scattered while attempting to learn a new process. Of course these things can be useful if the timing is right and the strategy is solid, but they also come at a cost.
2. Block as a team. Blocking your calendar is a simple and defensible way to make time for the work that matters. You can supercharge this tactic by agreeing to block your calendar as a team. When everyone in a group or department has the same “do not schedule” blocks on their calendar, it’s much easier to spend that time focused on work.
3. Make your workplace a place for work. Ironically, most offices are not great for getting work done, and open floor plans deserve most of the blame. Moving walls may not be realistic, but you can change the default behavior of your team by instituting Library Rules. Jason Fried, co-founder and president of 37signals and co-author of Rework, has a brilliant suggestion: Swap one default (you can talk to anyone anytime) with a different default, one that everybody already knows (act like you’re in a library).
4. Keep it small. Large teams have more overhead than small ones. Complicated projects have more unknowns than simple ones. Long timelines encourage people to take on unnecessary work. This probably seems obvious, but my experience is that most leaders make things bigger than they need to be. Keep teams, projects, and timelines as small as possible.
5. Reward the right behaviors. The 21st Century workplace is full of rewards for long hours and fast responses: compliments, promotions, and cultural badges of honor. If you want to get better, more valuable work from your team, think about which behaviors you reward — even if those rewards are small and unconscious. Instead of thanking someone who promptly replies to an after-hours email, encourage them to write a thoughtful response while at the office. Rewarding people who spend their time productively will encourage team members to practice that behavior, and discourage the notion that overwork is better work.
6. Have a contact contract. We have so many ways to keep in touch at work — writing emails, sending chats, scheduling meetings, hopping on calls. Which form of communication is the most appropriate, and when? You can help your team decide by having an open discussion about everyone’s preferences and then making guidelines that work for the majority. Think about timeliness, thoughtfulness, interruption, and synchronicity. The decisions you come to don’t have to be a literal contract, but they should create an understanding about when and how to communicate.
7. Don’t ask for updates. Nothing triggers anxiety like an email from the boss late in the day: “Hey, can you send me a quick update on Project Alpha?” This kind of message appears urgent — even if it’s not — and it will likely take time for your employee to respond. They may have to run numbers or ask collaborators for updates. A better way to keep tabs on projects is to ask your team for summaries. Explain to them that summaries come at the end of a project, or mark a milestone, and include: the results, the lessons learned, and what needs to happen next. It’s a semantic difference, but it’s significant. If you set clear deadlines, your team can anticipate when a summary is due, and plan updates around the data you want to see.
8. Be mindful of what you say, because everyone’s listening. When leaders make careless comments or suggestions, they can unintentionally change the workflow of their teams. As an employee, I’ve seen this happen many times, but my favorite example comes from Fried and his co-author, David Heinemeier Hansson, in their book about productivity, It Doesn’t Have to be Crazy at Work: “It takes great restraint as the leader not to keep lobbing ideas at everyone else. Every such idea is a pebble that’s going to cause ripples when it hits the surface. Throw enough pebbles in the pond and the overall picture becomes as clear as mud.” Leaders need to recognize the weight their words carry, and practice speaking with thoughtful intention.
9. Don’t expect consensus. Getting everyone to agree before moving forward with a decision can waste time if consensus is not realistic. In fact, a little conflict often inspires learning and innovation, especially on diverse, thoughtful teams. The key, then, is to collect input from everyone, consider your options, and then make a decision based on what you think is best given the information you have. Be transparent with your team about how you made the decision — what you considered, and why — and set time aside to answer questions. People should walk away with a clear understanding of your choice and how it affects their work. This will save you time later on.
10. Turn off the green dot. Your decisions about how you spend your time sets the example for your employees. As a leader, you might want them to know you’re available when they need you — but if being logged in and responsive at all times becomes your default, it might become theirs too. Projecting this kind of presence (whether in person or in the form of a logged-in “green dot”), sends the message that it’s okay for people to interrupt you whenever you’re needed, or worse, that the company values the appearance of availability over the time and focus needed to do great work. The solution is to create boundaries. Be straightforward about your time, when you need to focus, and when you are free. A good option is to create “office hours” — periods when anyone can drop in or schedule time with you — and regular check-ins with direct reports. These meetings will allow you to give people your undivided attention when you’re available to do so.
11. Be thoughtful, not reactive. When leading new initiatives, take the time to thoughtfully write your ideas down and consider them. Try not to “think out loud” in meetings. Even if you are brainstorming with others, avoid making a decision on the spot. Give yourself the mental space you need to feel confident that the decision you make is the best path forward. This will save time down the road, and help your team avoid unnecessary road blocks or last minute changes. Ask: How can I make this — product, service, or company — better right now? What are the first steps?
12. Take real breaks. Leave the office early. Take a weekend getaway. Go on a long vacation. And when you do, tell your team you’ll be out of the office and offline. Delegate people to make decisions while you’re out, or defer those decisions until you come back. Real breaks can make you a better leader, a happier person, and set the standard that people need, and deserve, time off.
If you’ve ever wished for better work, greater job satisfaction, or less stress for your team, you have the power to make those changes by rethinking the decisions you make about time. New behaviors have a funny way of becoming habits. What sounds crazy and new right now will seem normal and inevitable in a couple of years. Take these ideas as experiments you can run with, and start testing them tomorrow.
Antitrust deserves the attention it’s getting, and the tech platforms raise important questions. But the rise of big companies — and the resulting concentration of industries, profits, and wages — goes well beyond tech firms and is about far more than antitrust policy.
In fact, research suggests that big firms are dominating through their use of software. In 2011, venture capitalist Marc Andreessen declared that “software is eating the world.” Its appetizer seems to have been smaller companies.
What’s Driving Industry Concentration
Most industries in the U.S. have grown more concentrated in the past 20 years, meaning that the biggest firms in the industry are capturing a greater share of the market than they used to. But why?
Research by one of us (James) links this trend to software. Even outside of the tech sector, the employment of more software developers is associated with a greater increase in industry concentration, and this relationship appears to be causal. Similarly, researchers at the OECD have found that markups — a measure of companies’ profits and market power — have increased more in digitally-intensive industries. And academic research has found that rising industry concentration correlates with the patent-intensity of an industry, suggesting “that the industries becoming more concentrated are those with faster technological progress.” For example, productivity has grown dramatically in the retail sector since 1990; inflation-adjusted sales per employee have grown by roughly 50%. Economic analysis finds that most of this productivity growth is accounted for by a few companies such as Walmart who used information technology to become much more productive. Greater productivity meant lower prices and faster growth, leading to increased industry dominance. Walmart went from a 3% share of the general merchandise retail market in 1982 to over 50% today.
All of this suggests that technology, and specifically software, is behind the growing dominance of big companies.
IT Does Matter
In 2003, then-HBR-editor Nick Carr wrote an article (and later a book) titled “IT Doesn’t Matter.” Carr took issue with the common assumption “that as IT’s potency and ubiquity have increased, so too has its strategic value.” That view was mistaken, he argued:
“What makes a resource truly strategic—what gives it the capacity to be the basis for a sustained competitive advantage—is not ubiquity but scarcity. You only gain an edge over rivals by having or doing something that they can’t have or do. By now, the core functions of IT—data storage, data processing, and data transport—have become available and affordable to all. Their very power and presence have begun to transform them from potentially strategic resources into commodity factors of production. They are becoming costs of doing business that must be paid by all but provide distinction to none.”
Carr distinguished between proprietary technologies and “infrastructural” ones. The former created competitive advantage, but the latter were more valuable when broadly shared and so eventually became ubiquitous and were not unique to any company. IT would temporarily create proprietary advantages, he predicted, citing Walmart as an example. Walmart is the country’s largest employer and largest company by revenue and it reached that position through an operating model made possible by proprietary logistics software. But Carr believed that by his writing in 2003 “the opportunities for gaining IT-based advantages are already dwindling” and that “Best practices are now quickly built into software or otherwise replicated.”
It didn’t turn out that way. Although rivals have tried to build their own comparable logistics software and vendors have tried to commoditize it, Walmart’s software acumen remains part of its competitive advantage — fueled now by a rich trove of data. While Walmart faces new challenges competing online, it has maintained its logistics advantage against many competitors such as Sears.
The “Full-Stack” Startup
This model, where proprietary software pairs with other strengths to form competitive advantage, is only becoming more common. Years ago, one of us (James) started a company that sold publishing software. The business model was to write the software and then sell licenses to publishers. That model still exists, including in online publishing where companies like Automattic, maker of the open source content management system WordPress, sell hosting and related services to publishers. One-off licenses have given way to monthly software-as-a-service subscriptions, but this model still fits with Carr’s original thesis: software companies make technology that other companies pay for, but from which they seldom derive unique advantage.
That’s not how Vox Media does it. Vox is a digital publishing company known, in part, for its proprietary content management system. Vox does license its software to some other companies (so far, mostly non-competitors), but it is itself a publisher. Its primary business model is to create content and sell ads. It pairs proprietary publishing software with quality editorial to create competitive advantage.
Venture capitalist Chris Dixon has called this approach the “full-stack startup.” “The old approach startups took was to sell or license their new technology to incumbents,” says Dixon. “The new, ‘full stack’ approach is to build a complete, end-to-end product or service that bypasses incumbents and other competitors.” Vox is one example of the full-stack model.
The switch from the software vendor model to the full-stack model is seen in government statistics. Since 1998, the share of firm spending on software that goes to pre-packaged software (the vendor model) has been declining. Over 70% of the firms’ software budgets goes to code developed in-house or under custom contracts. And the amount they spend on proprietary software is huge — $250 billion in 2016, nearly as much as they invested in physical capital net of depreciation.
How Big Companies Benefit
Clearly, proprietary software is providing some companies advantage and the full-stack model is dominating the software-vendor model. The result is that large firms are gaining market share. But to explain that, one needs to explain why some companies are so much better at developing software than others and why their innovations don’t seem to be diffusing to their smaller competitors the way Carr thought they inevitably would.
Economies of scale are certainly part of the answer. Software is expensive to build but relatively cheap to distribute; larger companies are better able to afford the up-front expense. But these “supply-side economies of scale” can’t be the only answer or else vendors, who can achieve large economies of scale by selling to the majority of players in the market, would dominate. Network effects, or “demand-side economies of scale,” are another likely culprit. But the fact that the link between software and industry concentration is pervasive outside of the tech industry — where companies are less likely to be harnessing billions of users — suggests network effects are only part of the story.
Part of the explanation for rising industry concentration, then, seems to hinge on the fact that software is more valuable for firms in combination with other industry-specific capabilities. These are often referred to as “intangible assets,” but it’s worth getting more specific than that.
Research suggests that the benefits of information technology depend in part on management. Well-managed firms get more from their IT investments, and big firms tend to be better managed. There are other “intangible” assets that differentiate leading firms, and which can be difficult or costly to replicate. A senior executive who has worked at a series of leading enterprise software firms recently told one of us (Walter) that a company’s ability to get more from an average developer depended on successfully setting up “the software to make software” — the tools, workflows, and defaults that allow a programmer to plug in to the company’s production system without having to learn an endless number of new skills.
Patents and copyright also make it harder for software innovations to spread to other companies, as do noncompete agreements that keep employees from easily switching jobs. But one of the biggest barriers to diffusion — and therefore one of the biggest sources of competitive advantage for the firms that excel at software — comes down to how companies are organized.
In 1990, Rebecca Henderson, now a professor at Harvard Business School, published a paper that provides a theoretical basis for the success of full-stack startups. At the time, multiple thinkers were grappling with the question of why big, successful, cash-rich companies were sometimes unseated by new technologies. Incumbent companies aren’t necessarily bad at using new technologies, Henderson argued, based on her study of the photolithography industry. In fact, incumbents were great at using new technologies to improve individual components of their products. But when a new technology fundamentally changed the architecture of that product — the way everything fit together — the incumbent struggled.
Her point was that a company’s way of doing things is often deeply interconnected with the architecture of the products or services it creates. When the architecture changes, all the knowledge that was embedded in the organization becomes less useful, and the company’s way of doing things goes from advantage to disadvantage.
For example, Walmart’s competitive edge depended on having an organization and business model that took advantage of its logistical prowess by emphasizing Everyday Low Prices, large assortment, and rapid response to changes in tastes. Even though its larger competitors such as Sears spent heavily on IT, they could not compete effectively without making fundamental architectural changes. If all Walmart had done was apply IT to one component of the retail system — say, digitizing catalogs or bringing them online — Sears might have been in a better position to compete. But Walmart changed not only how supply chains, product decisions, and pricing worked, but how they related to each other. Sears’ entire existing way of doing things was suddenly a disadvantage.
As Dixon, the VC, clearly recognized, these architectural innovations can create openings for startups. “Before [Lyft and Uber] were started, there were multiple startups that tried to build software that would make the taxi and limo industry more efficient,” Dixon has noted. If Uber had merely created software for dispatching taxis, incumbents would have been well positioned to adopt it, according to Henderson’s theory. One “component” of the service would have been changed by technology (dispatching) but not the entire architecture of the service. But ridesharing startups like Uber and Lyft didn’t didn’t just make taxis more efficient; they fundamentally changed the way the different pieces of the system fit together.
Architectural innovation doesn’t necessarily result in startups displacing incumbents. It can also determine who prevails in a competition between larger, older firms. In November 2007, Forbes put the CEO of Nokia on its cover and asked, “Can Anyone Catch the Cell Phone King?” Apple had launched the iPhone just months before.
Why was Apple, a company with no prior experience in phones, able to overtake the cell phone king? Earlier this year, Harvard Business School professor Karim Lakhani asked this question to a group of conference attendees. One of us (Walter) listened as the technology experts in the audience listed all the ways the iPhone was superior: touch screen, app store, web browser, etc. Lakhani then provided the dates at which Nokia had offered those features: an app store in 2001, a touch screen in 2002, a web browser in 2006. Why, then, did Apple prevail?
Lakhani’s answer is that Apple had the right architecture to bring phones into the internet age. Apple and Nokia both had plenty of the intangible assets necessary to excel in the smartphone business, including software developers, hardware engineers, designers. But Apple’s structure and culture were already based around the combination of hardware and a software ecosystem to which third parties contributed. It already had experience building hardware, operating systems, and software development kits from its PC business. It had built a software platform to deliver content to mobile devices in the form of iTunes. Steve Jobs initially resisted letting developers build apps for the iPhone. But when he eventually gave in, the app store became the iPhone’s key advantage. And Apple was able to manage it because of its existing “architecture.”
Like any theory, architectural innovation can’t explain everything. If experience building operating systems and SDKs were so key, why didn’t Microsoft invent the winning smartphone? Apple’s particular acumen in product design clearly mattered, too. But architectural innovation helps explain why certain capabilities are so tough to replicate.
Spreading the Benefits of Software
The challenge for policymakers worried about industry concentration, markups, and the power of giant companies is to spread the benefits of the digital economy – of software – more broadly. Antitrust may be able to help in extreme cases, including in reining in the tech platforms and their ability to buy up competitors. But policymakers should also consider ways to help software and software capabilities diffuse throughout the economy. To some degree, economies of scale will simply increase the average size of firms, and that’s ok. But banning non-competes would help employees spread their knowledge by moving jobs. Reforming patents, which aren’t always necessary to protect software innovation and are abused by patent trolls to the detriment of nearly everyone, would help, too. Anything governments can do to encourage the use of open source software could help as well. For example, the French government mandates that public administrative bodies thoroughly review open source alternatives when revising or building new information technology and to use the savings realized to fund further open source development.
Encouraging startups is another promising avenue, as these firms are able to organize around software capabilities to take on incumbents. Doing so through public policy isn’t always easy, but government funding can help when done well, and at the state and city level policymakers can encourage the formation of technology clusters. These policies would pair well with more aggressive merger review, to ensure that promising startups are not all swallowed up by the incumbents they’re challenging.
For companies, the takeaway is more obvious. Even if you’re not in the software industry, there’s a good chance your success hinges on your ability not just to use but also to build software. Using vendors often still makes financial sense, of course. But consider what makes your company unique, and how software might further that advantage. Investing in proprietary solutions that complement your strengths might be a good idea, especially for medium and large companies and for growth startups.
A Cloud on the Horizon
There is some good news: researchsuggests that cloud computing is helping smaller, newer firms to compete. Also, some firms are unbundling their advanced capabilities. For example, Amazon now offers complete fulfillment services including two-day delivery to sellers, large and small, on its Marketplace. It may be that Carr was right in principle but just had the timing wrong. But we wouldn’t bet on it. Some aspects of software will be democratized, including perhaps some areas where companies now derive competitive advantage. But other opportunities will arise for companies to use software to their advantage. One in particular stands out: even when machine learning software is freely available, the datasets to make it valuable often remain proprietary, as do the models companies create based on them. Policy may be able to help level that playing field. But companies that don’t invest in software and data capabilities risk being left behind.
“To be wise you must arrange your experiences on a lattice of models.”
— Charlie Munger
Organizations are awash in data — from geocoded transactional data to real-time website traffic to semantic quantifications of corporate annual reports. All these data and data sources only add value if put to use. And that typically means that the data is incorporated into a model. By a model, I mean a formal mathematical representation that can be applied to or calibrated to fit data.
Some organizations use models without knowing it. For example, a yield curve, which compares bonds with the same risk profile but different maturity dates, can be considered a model. A hiring rubric is also a kind of model. When you write down the features that make a job candidate worth hiring, you’re creating a model that takes data about the candidate and turns it into a recommendation about whether or not to hire that person. Other organizations develop sophisticated models. Some of those models are structural and meant to capture reality. Other models mine data using tools from machine learning and artificial intelligence.
The most sophisticated organizations — from Alphabet to Berkshire Hathaway to the CIA — all use models. In fact, they do something even better: they use many models in combination.
Without models, making sense of data is hard. Data helps describe reality, albeit imperfectly. On its own, though, data can’t recommend one decision over another. If you notice that your best-performing teams are also your most diverse, that may be interesting. But to turn that data point into insight, you need to plug it into some model of the world — for instance, you may hypothesize that having a greater variety of perspectives on a team leads to better decision-making. Your hypothesis represents a model of the world.
Though single models can perform well, ensembles of models work even better. That is why the best thinkers, the most accurate predictors, and the most effective design teams use ensembles of models. They are what I call, many-model thinkers.
In this article, I explain why many models are better than one and also describe three rules for how to construct your own powerful ensemble of models: spread attention broadly, boost predictions, and seek conflict.
The case for models
First, some background on models. A model formally represents some domain or process, often using variables and mathematical formula. (In practice, many people construct more informal models in their head, or in writing, but formalizing your models is often a helpful way of clarifying them and making them more useful.) For example, Point Nine Capital uses a linear model to sort potential startup opportunities based on variables representing the quality of the team and the technology. Leading universities, such as Princeton and Michigan, apply probabilistic models that represent applicants by grade point average, test scores, and other variables to determine their likelihood of graduating. Universities also use models to help students adopt successful behaviors. Those models use variables like changes in test scores over a semester. Disney used an agent-based model to design parks and attractions. That model created a computer rendition of the park complete with visitors and simulated their activity so that Disney could see how different decisions might affect how the park functioned. The Congressional Budget office uses an economic model that includes income, unemployment, and health statistics to estimate the costs of changes to health care laws.
In these cases, the models organize the firehose of data. These models all help leaders explain phenomena and communicate information. They also impose logical coherence, and in doing so, aid in strategic decision making and forecasting. It should come as no surprise that models are more accurate as predictors than most people. In head-to-head competitions between people who use models and people who don’t, the former win, and typically do so by large margins.
Models win because they possess capabilities that humans lack. Models can embed and leverage more data. Models can be tested, calibrated, and compared. And models do not commit logical errors. Models do not suffer from cognitive biases. (They can, however, introduce or replicate human biases; that is one of the reasons for combining multiple models.)
Combining multiple models
While applying one model is good, using many models — an ensemble — is even better, particularly in complex problem domains. Here’s why: models simplify. So, no matter how much data a model embeds, it will always miss some relevant variable or leave out some interaction. Therefore, any model will be wrong.
With an ensemble of models, you can make up for the gaps in any one of the models. Constructing the best ensemble of models requires thought and effort. As it turns out, the most accurate ensembles of models do not consist of the highest performing individual models. You should not, therefore, run a horse race among candidate models and choose the four top finishers. Instead, you want to combine diverse models.
For decades, Wall Street firms have used models to evaluate investment risk. Risk takes many forms. In addition to risk from financial market fluctuations, there exist risks from geopolitics, climactic events, and social movements, such as occupy Wall Street, not to mention, risks from cyber threat and other forms of terrorism. A standard risk model based on stock price correlations will not embed all of these dimensions. Hence, leading investment banks use ensembles of models to assess risks.
But, what should that ensemble look like? Which models does one include, and which does one leave out?
The first guideline for building an ensemble is to look for models that focus attention on different parts of a problem or on different processes. By that I mean, your second model should include different variables. As mentioned above, models leave stuff out. Standard financial market models leave out fine-grained institutional details of how trades are executed. They abstract away from the ecology of beliefs and trading rules that generate price sequences. Therefore, a good second model would include those features.
The mathematician Doyne Farmer advocates agent-based models as a good second model. An agent-based model consists of rule based “agents” that represent people and organizations. The model is then run on a computer. In the case of financial risk, agent-based models can be designed to include much of that micro-level detail. An agent-based model of a housing market can represent each household, assigning it an income and a mortgage or rental payment. It can also include behavioral rules that describe conditions when the home’s owners will refinance and when they will declare bankruptcy. Those behavioral rules may be difficult to get right, and as a result, the agent-based model may not be that accurate — at least at first. But, Farmer and others would argue that over time, the models could become very accurate.
We care less about whether agent-based models would outperform other standard models than whether agent-based models will read signals missed by standard models. And they will. Standard models work on aggregates, such as Case-Shiller indices, which measure changes in prices of houses. If the Case-Shiller index rises faster than income, a housing bubble may be likely. As useful as the index is, it is blind to distributional changes that hold means constant. If income increases go only to the top 1% while housing prices rise across the board, the index would be no different than if income increases were broad based. Agent based models would not be blind to the distributional changes. They would notice that people earning $40,000 must hold $600,000 mortgages. The agent based model is not necessarily better. It’s value comes from focusing attention where the standard model does not.
The second guideline borrows the concept of boosting, a technique from machine learning. Ensemble classification algorithms, such as random forest models consist of a collection of simple decision trees. A decision tree classifying potential venture capital investments might say “if the market is large, invest.” Random forests are a technique to combine multiple decision trees. And boosting improves the power of these algorithms by using data to search for new trees in a novel way. Rather than look for trees that predict with high accuracy in isolation, boosting looks for trees that perform well when the forest of current trees does not. In other words, look for a model that attacks the weaknesses of your current model.
Here’s one example. As mentioned, many venture capitalists use weighted attribute models to sift through the thousands of pitches that land at their doors. Common attributes include the team, the size of the market, the technological application, and timing. A VC firm might score each of these dimensions on a scale from 1 to 5 and then assign an aggregate score as follows:
This might be the best model the VC can construct. The second best model might use similar variables and similar weights. If so, it will suffer from the same flaws as the first model. That means that combining it with the first model will probably not lead to substantially better decisions.
A boosting approach would take data from all past decisions and see where the first model failed. For instance, it may be that be that investment opportunities with scores of 5 out of 5 on team, market size, and technology, do not pan out as expected. This could be because those markets are crowded. Each of the three attributes —team, market size, and workable technology — predicts well in isolation, but if someone has all three, it may be likely that others do as well and that a herd of horses tramples the hoped for unicorn. The first model therefore would predict poorly in these cases. The idea of boosting is to go searching for models that do best specifically when your other models fail.
To give a second example, several firms I have visited have hired computer scientists to apply techniques from artificial intelligence to identify past hiring mistakes. This is boosting in its purest form. Rather than try to use AI to simply beat their current hiring model, they use AI to build a second model that complements their current hiring model. They look for where their current model fails and build new models to complement it.
In that way, boosting and attention share something in common: they both look to combine complementary models. But attention looks at what goes into the model — the types of variables it considers — whereas boosting focuses on what comes out — the cases where the first model struggles.
Boosting works best if you have lots of historical data on how your primary model performs. Sometimes, we don’t. In those cases, seek conflict. That is, look for models that disagree. When a team of people confronts a complex decision, it expects — in fact it wants — some disagreement. Unanimity would be a sign of group think. That’s true of models as well.
The only way the ensemble can improve on a single model is if the models differ. To borrow a quote from Richard Levins, the “truth lies at the intersection of independent lies.” It does not lie at the intersection of correlated lies. Put differently, just as you would not surround yourself with “yes men” do not surround yourself with “yes models.”
Suppose that you run a pharmaceutical company and that you use a linear model to projects sales of recently patented drugs. To build an ensemble, you might also construct a systems dynamics model as well as a contagion model. Say that the contagion model results in similar long-terms sales but a slower initial uptake, but that the systems dynamics model leads to a much different forecast. If so, it creates an opportunity for strategic thinking. Why do the models differ? What can we learn from that and how do we intervene.
In sum, models, like humans, make mistakes because they fail to pay attention to relevant variables or interactions. Many-model thinking overcomes the failures of attention of any one model. It will make you wise.
It takes time and care to develop trusting relationships with the women we work with, particularly women who are different from us in some way. But the effort of understanding each other’s experiences is worth it, personally and professionally: We’ll feel less alone in our individual struggles and better able to push for equity.
We talk with professors Tina Opie and Verónica Rabelo about the power of workplace sisterhood. We discuss steps, as well as common snags, to forming deep and lasting connections with our female colleagues.
Tina R. Opie is an associate professor of management at Babson College.
Verónica Caridad Rabelo is an assistant professor of management in the College of Business at San Francisco State University.