#1: What Machine Learning Can Do for Your Business and How to Figure It Out
This is part 1 of the 6-part tutorial, The Step-By-Step PM Guide to Building Machine Learning Based Products. Follow the link for an overview of the entire series.
Investing in ML Is like Investing in Mobile 10 Years Ago — It Can Transform Your Business
Querying existing data for insights is a well known, widely adopted discipline. ML, however, is the next frontier in data analysis. It is a discipline where computer programs make predictions or draw insights based on patterns they identify in data and are able to improve those insights with experience — without humans explicitly telling them how to do so. As organizations have access to more data, machine learning enables them to draw insights from the data at scale, at a level of granularity that ranges from a single user interaction to worldwide trends and their impact on the planet. The use of those insights can also range from customizing an individual user’s experience at the pixel level to creating new products and business opportunities that don’t currently exist. Note that with ML you can go well beyond using internal data — the power of ML can often be enhanced by marrying internal with external data to drive new insights that were not previously possible.
Frank Chen from A16Z has an excellent primer on the potential applications of artificial intelligence, many of which require or will require machine learning. Some of these applications are future looking and not yet achievable with existing technology, but give a great sense of the possibilities.
Just like consumer companies started to think about investing in mobile 8–10 years ago, now is the time for companies to start exploring ML as a technology that can help drive business results. For companies focusing on leveraging existing ML technologies, there are several key themes for what ML allows you to do. These are not exhaustive or mutually exclusive, but rather represent different angles of thinking about potential impact on your business:
- Mass customization of a user’s environment, experience and system responses. Imagine that everything a person does or sees could be customized especially for them and even anticipate their needs and behaviors. That includes recommendations for products or services, ranked by level of relevance to them; tailored user experience or flows based on knowledge you have of the user, their behavior, other people like them or external data, including predicting what they would want to do next etc. At a smaller scale this could translate into customization of experience to segments of users rather than individuals.
- The ability to visually identify objects and automate or tailor experiences accordingly. Technology today can identify objects in photos and videos, including on live cam. Pinterest uses this to suggest similar / complementary objects to those in a photo the user is looking at; Facebook uses face recognition technology to suggest friends to tag in a photo, Amazon is building automatic store checkout based on visual identification of objects etc.
- Automatic retrieval, generation or processing of content. ML enables expedient processing of the massive amounts of content in the world. Common uses are document retrieval — e.g. finding all the documents that are relevant to a legal case (note that this goes beyond just keywords into contextual search), classification of documents by topic and keywords, automatic summary of the content, extraction of pertinent information from large amounts of content — e.g. finding specific terms in vendor contracts etc. “Content” here applies to all types of media, not just text.
- Predictions, estimates and trends at scale. ML enables predictions that are very expensive or difficult to make otherwise. ML is particularly useful for making predictions that otherwise require a high level of expertise such as the price of a home, or are even impossible for a human to make such as which content will do well on social media. Machines can also identify trends in data well before they become obvious to humans.
- Detection of unusual activity or system failures. Every system has failures and issues, but ML allows you to not just detect whether issues arise, but also whether those issues are unusual and alarming. This is particularly useful in various monitoring and security systems.
From a strategic perspective, ML can drive several types of business outcomes:
- Enhanced experience and functionality for your customers. The most common use case is mass customization — finding the products that are most likely to be relevant for your customers more quickly and efficiently, e.g. their best matches on dating sites, song they might like on music sites, products they may be interested in purchasing etc. The other use case is using predictions to get them intelligence about entities or situations that they would not have otherwise. This could be general — e.g. Zillow’s Zestimate values a house the same regardless of who’s looking at it, or customized to the individual customer — e.g. the rating a user is likely to give a movie they have not seen given their specific tastes.
- Internal functions, processes and business logic. Machine learning can save you time and make your resource investment more effective when it comes to business processes and decisions. For example: A lending company would like to prioritize its outreach to potential loan applicants. It needs to determine who wants a loan enough to actually take it if offered, but is still likely to be able to repay it. Prioritizing the most creditworthy customers is not necessarily the answer, since those customers usually have many options and are less likely to convert, so a more complex model is required.
- Expansion to new verticals and new products. Data can help you open completely new business opportunities — create brand new products for your existing customers, or serve segments or customers you haven’t served before. For example: Netflix can serve studios, which weren’t the core target audience, by selling them insights from its data on what themes and plot lines work for which audiences; Zillow can help real estate developers understand which building features will get them the highest return on investment, etc.
The decision which area to address first should depend on the potential business impact, as well as the complexity of the problem and the cost of achieving that impact.
“We Need to Do Something with Our Data” Is a Strategy, Not a Data Science, Problem
Many companies are looking to hire data scientists, the people who build ML models, because “we should do something with our data”. I’ve heard many executives in prominent companies say “we see our competitors buy data so we need to do this to stay competitive”, and then go hire a couple of data scientists hoping they’ll come up with some magic. This brings me to a big misconception about ML.
ML is not a magic wand for your business. The first challenge in ML is figuring out the business impact the technology aims to drive. ML is a solution — you need to first define the problem: What are the business results you’re hoping to achieve with ML? What benefit can ML provide for your customers? ML is a hammer — but if you don’t have a nail, a hammer is not particularly useful. To stretch the cliché even further, ML is a hugely varied set of hammers, and the kind of nail you have will determine which hammer you’ll pick and how you’ll use it. The precise problem you’re trying to solve will dictate everything — how the result will be used, what your model should predict and how it should be calibrated, what data you collect and process, what algorithms you test and many other questions.
At its core, “what problem are we solving?” is a business question, which means defining it is ultimately the responsibility of product managers and business leaders, not data scientists. Data scientists and other stakeholders should absolutely be involved in getting to the definition — just don’t throw the question at them and expect them to come back with answers. If you have data that you don’t know what to do with, conduct customer interviews and ideate with other customer-facing people across the organization. Data scientists can help you explore your data, ideate and iterate, but unless they have a lot of problem space expertise it would be difficult for them to come up with the business case on their own. In order to maximize the value of ML to the business you need an ongoing collaboration between product managers and data scientists, where it is the responsibility of the product managers to ensure that the problems being solved are the most impactful ones for the business.
Unpacking How ML Can Move Your Business Forward
While the possibilities with ML are endless, there are certain questions you could ask to figure out how the technology could apply to your organization. Here are some examples:
- Where do people in my company today apply knowledge to make decisions that could be automated, so their skills could be better leveraged elsewhere?
- What is the data that people in my company normally search for, collect or extract manually from certain repositories of information and how can this be automated?
- What is the set of decisions people at my company make? Can those decisions conceivably be made by a machine if it magically ingested all the data my people have?
Products and experience for existing customers
- What parts of my customer interactions are customized by people and could potentially be customized by machines?
- Do I have a clear segmentation of my customers based on their preferences, behaviors and needs? Is my product / experience customized for each segment?
- Can I customize the experience for each individual customer based on what I know about them or their interaction with my site / app / product? How could I create a better, faster or otherwise more delightful experience for them?
- Specifically, what are the decisions and choices I’m asking my customers to make today? Can those decisions be automated based on some knowledge I already have or could have?
- How can I better identify good vs. bad customer experiences? Can I detect issues that will negatively impact customer experience or satisfaction before they happen or spread?
New verticals or customers
- Do I have any data that could be useful to other stakeholders in the industry or in adjacent industries? What sort of decisions can it help these stakeholders make?
All the above
- What are the metrics or trends that if I could correctly predict would have a meaningful impact on my ability to serve my customers or otherwise compete in the industry, e.g. forecast demand for certain categories of products, cost fluctuations etc.?
- What are the key entities about which I gather data (people, companies, products etc.)? Can I marry that data with any outside data (from public sources, partners etc.) in a way that tells me something new or useful about those entities? Useful to whom and how? For example: Identify potential customers when they are on the verge of looking for your product, understand how external factors affect demand in your industry and react accordingly, etc.
Brainstorm some of these questions (and others) with your team and key stakeholders in the organization. If you’re not sure where to start — start somewhere. Just experimenting with some data can help you and your team figure out where you can go from there.
In part 2, we’ll discuss all the ML technical terms PMs need to understand, how technology choice is affected by your problem definition, and some of the modeling pitfalls to watch out for that have an impact on your business.
If you found this post interesting, would you please click on the green heart below to let me know, or share with someone else who may find it useful? That would totally make my day!