Einstein Discovery is a curious enigma in the Salesforce ecosystem. Most people in the Salesforce world have heard of Einstein, as it has been widely publicised, but the majority have little understanding of what it does and how they can use it. In fact, even a good number of thought leaders and consulting partners are in the dark around Einstein.
I think that three factors have contributed to this scenario:
- Lack of understanding around the general subject of artificial intelligence and machine learning, especially in the context of business deliverables.
- Confusion about the rather complex family of Salesforce Einstein products – including Einstein Analytics, Sales Cloud Einstein, Einstein insights, Einstein Prediction Builder, and Einstein Discovery. Phew!
- The challenge that Salesforce Account Executives have to effectively communicate a product family that is becoming increasingly diverse and complex.
This lack of understanding is unfortunate, because Einstein Discovery (ED) is an extremely powerful tool that can offer tremendous business value if correctly implemented and effectively employed.
The purpose of this post is not to publish a detailed technical treatise about the ED platform; I’ll leave that to others. Rather, my goal is to distill the jargon, bypass the hype, and introduce a very capable piece of kit to those who might benefit from it.
1. What is Einstein Discovery?
Supervised machine learning. Those thee words succinctly define the ED platform. However, what does this mean? How does a machine learn? And why does it need supervision, as if it were some delinquent child? Let me explain.
Machine learning (ML) is analytical model building that can be automated, where the systems “learn” from data and identify patterns, resulting in meaningful conclusions and predictions.
I like this definition of ML from the SAS institute web site – “Machine learning is a method of data analysis that automates analytical model building. It is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention.”
There are two types of machine learning: supervised, and unsupervised. They are fundamentally different, so let’s briefly set them apart.
In supervised machine learning you have a set of features, or variables, similar to the column names in a spreadsheet. You also have a response variable that you use the features to make predictions for. A simple example would be looking at the brand, age, size and condition of motor vehicles in order to make a prediction about their value. Supervised learning is what Einstein Discovery does, and it is very good at it.
In unsupervised machine learning, all you have are the features – there is no response variable. Unsupervised ML is a growing field as it is being used for applications where the amounts of data are enormous, but there is no one specific answer that is being looked for in the data. Unsupervised ML looks for patterns in the data without having any pre-defined notion of how the features might related to one another.
Now that we know the difference between these two, let us return to our understanding of Einstein Discovery. Here is an important statement for you to note:
Einstein Discovery is a tool for data analysis.
The datasets to be analysed can come from many different sources, such as CSV files, databases, Salesforce, etc. Once data is loaded into ED, this amazing tool creates a “story”. What does this mean? When the “create story” button is clicked, the statistical algorithms behind the scenes begin learning from the data and identifying patterns. In other words, automated analytical model building takes place at this point.
Although the statistical methods used are quite intricate, and somewhat unique to this platform, their goal is to learn from the data and identify patterns. The proven techniques used in ED allow the system to learn from the data very rapidly; it is not unusual for models created on large datasets to finish running in only a few minutes. When we are looking at millions of rows of data with dozens of variables, it becomes clear why the machine excels at this task; similar computations performed by humans would take hundreds or possibly even thousands of hours to complete.
The story output of ED is delivered in a user-friendly manner that can be consumed and acted upon with great ease and great speed. On a practical level, the results that ED produces can be analysed and utilised in a number of ways, including:
- Displayed on Salesforce records.
- Included in Salesforce reports and dashboards.
- Visualised in Einstein Analytics.
- Employed to trigger Salesforce automation.
- Flagging important data trends to relevant users
2. What can Einstein Discovery do for your business?
The simple answer is, a great deal indeed!
It is all about the business. After all, Einstein Discovery, like any other AI platform, is simply a tool. Thus, “For most of us, figuring out how AI fits into our workplace begins with defining a problem that needs to be solved. (You don’t want to start with machine learning and make up a business problem to suit.) Only after you have clearly outlined the problem can you start to look at what solutions AI might offer.” (Goku Mohandas)
Do you have copious amounts of data sitting in various systems and providing little or no value to your organisation? Do you struggle to make sense of the terabytes of data that your business systems collate day after day? Do your disconnected teams and systems make it nigh impossible to make data-driven decisions at a group level? Then you can benefit from the power of Einstein Discovery!
Here are some practical examples of what Einstein Discovery can do for your business. This list is by by means exhaustive, but it should open your eyes to the possibilities.
- Predictive analytics around pipeline, forecasting, etc. It is one thing to analyse your data and identify trends around what has taken place. It is another thing entirely to analyse your data and see what could take place, or what might happen if you change certain variables. ED can employ ML and AI to make powerful, accurate predictions about your pipeline, cash flow, profitability, etc.
- Behavioural analytics with cohort analysis. That is, analyse millions of rows of events data (web site behaviour, social footprint, online interaction, app usage, etc) to facilitate behavioural understanding. Cohort analytics allows product teams to track the many different users within their platform and cater to their specific needs. Each group is known as a cohort and may have very different characteristics. Cohorts analytics show product teams where advanced users need more features, where less advanced users need more guidance, and where every cohort typically runs into trouble along their user journey. This use case incorporates churn and retention analysis.
- Sentiment analysis. Imagine being able to process millions of emails, posts, texts, and other sources of customer sentiment, then use this information to identify key sentiments among your client base. Well, you can, with Einstein Discovery. The results of this analysis can be employed for many purposes, such as case assignment, case escalation, retention analysis, etc.
- Identifying significant trends in your business and extrapolating what those trends could mean for you in the future. The larger and more complex your business becomes, the more difficult it gets to truly know what is happening and why it is happening. How good would it be to have an intelligent system analyse your data and identify important trends for you, then bring these trends to your attention in a way that is simple and actionable? Yep – ED does this.
- Supporting or refuting your understanding of your business, and proving or debunking the paradigms that you base your decisions upon, by analysing your data and exposing the naked truth without bias or prejudice. Similar to point four above, what you think you know about your business drives your decision making. What if you and your colleagues are wrong, even partly so? If your business paradigms and assumptions are flawed, your decisions will be faulty. ED enables you to support or refute your understanding by analysing your business data and exposing the truth, warts and all. It processes very large amounts of data, from various sources, in order to give business understanding, in a way that would be impossible without the aid of AI. Numbers don’t lie.
- Combine ED with marketing automation to create a powerful and adaptive system for marketing, service, and customer experience. Understanding and utilising your data to empower your marketing automation system is, well, really difficult. ED simplifies this process so that data-based conclusions can be fed back in to your CRM to supercharge your marketing automation.
- Couple Einstein Discovery with Einstein Analytics to build advanced trending and predictive analytics. When you augment your analytics system with the discoveries produced by ED, you create a powerful, augmented analytics platform that creates incredible value from business data.
3. How does this work in real life?
You are concerned about your compliance with government regulations, both in the past, and in the future. However, you have so much complex data on this subject that you would hardly know where to begin in the process of understanding and applying it. How do you find out if you have been compliant? How do you know where to begin the process of remediation and mitigation? What does your data tell you about future compliance risks? How can you make sense of so much information before it is too late? ED can make this happen and give you confidence that you are on the path to compliance. It can empower your risk mitigation strategy.
Imagine that you want an understanding of what new users on your app need to help them go to the next level and become loyal customers. The amount of data that you have around this is staggering, and overwhelming. There is simply no way to process this much data in Excel! So, what do you do? You could hire a team of data scientists and have them perform complex feature engineering and data modelling, then give you a result in a few months. Or, you can implement ED, prepare your data, perfect the model, and produce meaningful results and get a great ROI in a matter of weeks.
Another example is enabling a detailed and accurate understanding of the primary factors that influence your sales cycle and pipeline. Unfortunately, what you think is driving and controlling your sales cycle may not actually be doing so. You might assume, based upon personal and corporate experience, that the deal amount significantly influences the time it takes to close deals. However, this may not be the case, and another entirely different variable might be responsible for shortening or lengthening this critical time period. How can you find out? You have the data – you just need a way to process it, analyse it, identify the patterns, and make some solid, data-based conclusions. ED can enable you to do this!
What if you have a nagging feeling that some of your key business assumptions might not be entirely correct? Good luck with sitting down your C-suite and persuading them of this with your charm, wit, and intelligence. Opinions are a dime a dozen, and they are not easily relinquished. How can you find out it your assumptions are flawed, and how would you go about convincing your colleagues if this is the case? The answer is in your data. Once ED analyses your data and tests your hypotheses, you can correct any faulty thinking, build leadership paradigms on the truth, and move ahead with confidence.
Finally, who doesn’t want the ability want to forecast future business based upon years of existing data? How good would it be to be able to accurately model your business and predict various outcomes depending upon virtual levers that you pull within a simulated business environment? This would not only help you rule out faulty strategies, but it would help you to focus your efforts, teams and resources upon those plans that are more likely to succeed. ED can enable this with relative speed and ease.
“We live in a time of almost unimaginable volume of data creation; the rates really are quite mind boggling. Fortunately, we also live in a period of exciting advances in both the computing power – and the automation of statistical machine learning techniques. Together, these provide us an amazing super power with which to harness the insights, wisdom, and value of all that data. And for that, we thank you Einstein.” (Darvish Shadravan)
Many companies are sitting on a gold mine, but they never mine it, or get value from it – in fact, it costs them a great deal of money to maintain. I refer to the untouched gold mine of business data, where a wealth of information is neglected because businesses lack the tools and talent to mine it, process it, and get value from it. Times, though, have changed – powerful tools like ED can be utilised by citizen data scientists in order to empower data-based decision making. The few business that realise this and adopt such tools are moving ahead of the pack, and the rest are getting left behind.
Most businesses do not have the in-house skills to rightly deploy Einstein Discovery. If you do not have a team member with the time, aptitude and desire to implement ED, it may be wise to engage a consulting partner who can deliver you a timely ROI with minimum fuss and delay.
The results just might be awesome…
Resources and references:
- Understanding the practical applications of business AI
- Einstein Discovery….Machine Learning Or Just What We Do For Fun?
- An executive’s guide to AI
- What Fortune 500 Companies Really Need to Know About AI
- Machine Learning in Finance – Present and Future Applications
- How artificial intelligence can deliver real value to companies
- Twelve Predictive Model Data Prep Tips
- Moving Beyond Data Visualization to Predictive Analytics
- Data Prep Essentials for Automated Machine Learning
- Predictive Model Data Prep: An Art and Science
- The Growing Impact of AI on Business
- Demystifying Salesforce Einstein