A few years ago, almost every sales conference, article or webinar was bound to include these magic words: Big Data. The mirage of large and diverse data sets that could be turned into valuable information, with little or no effort, seduced business people everywhere. And for good reason. In theory, Big Data held the potential to transform the way modern companies operate. With multiple techniques developed to mine Big Data for vital, actionable insights, a true business revolution seemed possible.
Today, most of those techniques have lost power. But not all. Predictive analytics remains favored among consultants and service companies alike.
It has endured, in part, due to the abundance of historical data we have access to, the perfect fuel for prediction. The rise of cloud computing and the on-demand availability of almost infinite computing power have also led to its increased popularity and use.
But will predictive analytics turn out to be an incremental step in our quest for a more stable market or will it be a major step towards the future?
A Harvard Business Review article credits predictive analytics for a Harley-Davidson dealership’s 2,930% increase in sales leads. How did they do it?
By making use of the dealership’s CRM database and by analyzing existing creative content such as headlines and visuals, an AI-driven marketing platform isolated a set of characteristics and behaviors of potential customers.
Those who had completed a purchase or placed an item in their cart, along with those who had spent enough time on the website were grouped into small segments. The platform tried to find out which media and content worked best for each segment and drove leads that resembled them to the store. This process not only wielded the expected results, but also revealed a larger, potential market.
In other words, the Harley-Davidson dealership used a software-as-a-service platform that combined existing historical data, specific algorithms, and pattern recognition to predict and assess future outcomes. This is the very definition of predictive analytics. This one example among the many that show predictive analytics works. Given the right circumstances, it can be a formidably precise instrument.
For those of you who lead and manage sales, the question is, can predictive analytics be effective when applied to sales performance management? If so, when and how can it benefit your sales department?
1. When Does It Work?
When your goals are clearly set.
For predictive analytics to help you meet your goals, you’ll need to set those goals with clockwork precision. Narrow them down to a single question, if possible: what type of event or behavior from the past should be replicated or avoided in the future?
For example, if you’re running the sales department, generating revenue is one of your main concerns. Using predictive analytics, you can crunch large amounts of historical data about the deals your company has won and about those it has lost. Doing so will help you understand which of your current deals are most likely to close. Using the information, you now have and, if available, the power of compensation calculators, you can lead your sales reps to their highest earning opportunities. Thus, you can focus on the deals you can win and that will deliver the highest ROI.
However, revenue might not be the only thing on your mind. As a sales manager, you might also want to know who your top performing reps are. By using predictive analytics, you can identify which reps’ profiles are best suited to sell specific products in specific locations. You can also follow your reps’ improvement patterns to see if their success might be replicated for lower performing teams. If you have a large sales force under your leadership, predictive analytics could even help you detect the periods when your sales representatives are most likely to quit.
The uses of predictive analytics are not, of course, confined to a single industry or to a specific type of goal. Financial institutions, for example, can use predictive analytics to reduce risks or to better identify potential clients. Given the right data and the right circumstances, a healthcare company could even assess the likelihood that a competitor will launch a product at a certain time.
Predictive analytics can be systematically used to help you reach long-term objectives, such as retaining valuable customers or maintaining a steady revenue stream. It can also be used ad-hoc, for a specific goal, such as detecting if a specific event will happen again. Just like Uber can predict the destination of its returning customers, your business could, for example, anticipate the possibility of a sale happening in a specific territory. Any of the described scenarios requires that a clear target be set from the get-go.
When the market allows it.
As a process, predictive analytics works best in markets where you can establish clear parameters, based on measurable variables. Therefore companies from the pharma, telecom, and financial services industries, along with insurance companies and online- based businesses, have achieved the most success leveraging predictive analytics. Companies that rely on selling expensive, specialized, equipment are quickly following in their footsteps.
Like all markets, such markets can also be unpredictable. However, even when stability is an issue and long-term success may vary, these markets still offer measurable variables that allow active businesses to establish clear goals and focus on specific problems. Sometimes, even being able to measure variables such as growth, market share, or profitability can be enough.
To see what type of market you’re competing with, you have to ask yourself: are there any market rules that all major players follow? Does my market allow me to identify and isolate any variables useful to my model? Are values such as customer retention and employee engagement fully measurable, or are these only estimable? Is it possible to build models to quantify these variables, and if so, will the same variables be usable in larger algorithms or models? If you can get clear answers to these questions, you’ll find out if your market is ready for predictive analytics.
When your company size requires it.
The success of predictive analytics in any company depends on permanent investment and a steady stream of data. To justify large-scale use of the technique, your company needs a significant portfolio or at least a couple of large ROI deals. Also remember that models will have to be continually updated through machine learning to adapt to market changes, customer requests, and external influences.
Furthermore, when your company’s complexity hits a certain threshold, being competitive is no longer enough: you need to innovate constantly. Since the very definition of “competitive” changes with each passing quarter, your company must be able to continuously invest in its prediction techniques. Therefore a large company will find it easier to adopt predictive analytics.
Also, if you’re just beginning your journey to the top, you have to understand that predictive analytics can only help you grow if you already have a set of previous high-profile deals you wish to replicate. If you’re simply testing the market, other techniques may be more appropriate.
2. How Does It Work?
For predictive analytics to work its magic, Smart Data must be available. Smart Data is Big Data that’s been proficiently sifted through by your own teams or a service provider, cleaned of gaps and seasonal variations, and uniformly formatted. You’ll require a wide set – a couple of years’ worth – of complete and usable data to get dependable answers to your questions.
Also, remember that large volumes of data require equally large processing power. Having the data is only one part of the story. Your technology must be able to handle it.
A high-accuracy model must be established.
Once the data has been centralized and the required parameters set, the variables you choose to follow must be combined into a model capable of assessing future probabilities in a reliable and results-driven way.
While some data can be studied directly through statistical analysis, models are usually the preferred solution for predictions (what could happen?) and forecasts (how likely is something to happen?). Your model will follow a set of parameters (be it sales, new clients, or closed deals) through the years of available data and try to find the patterns most useful for your business to recognize and repeat.
A predictive model should be flexible, adaptable, and contextual. It should also be capable of handling tweaks and improvements, and of integrating new data. You’ll need skilled data analysts who can find patterns and commonalities in existing data and who can understand such a model and its algorithms.
Efficient and repeatable algorithms should be selected.
When trying to understand the science behind the magic, there’s a chance you’ll stumble upon terms such as decision trees, time series, regression, posterior probability, random decision forest, and gradient boosting, to name a few. These might seem confusing at first, but they’re well-known within the data analysis field and easily defined.
Regression analysis is one of the most common statistical approaches in predictive analytics. This technique requires evenly distributed, continuous data and establishes a relationship between a dependent variable and one or more independent variables.
For example, a financial institution could try to analyze current and historical payment data and establish a connection between future fraud attempts and independent variables, such as the device or region from which the transaction originates. Which is exactly what PayPal did in 2016 when trying to stop fraud attempts before transactions were processed, according to this TechTarget article.
While the financial giant probably used multiple PA techniques, it’s safe to say that regression analysis on its own could be of great use to a sales department. When applied to SPM, this algorithm can uncover new patterns and relationships between variables, leading to new insights and perspectives throughout the business.
Testing should be a constant, not a variable.
While all the techniques mentioned above can easily help you predict certain results, correlations between certain data sets may arise unexpectedly. It’s important to constantly test and learn from your parameters, algorithms, and models. Predictive analytics is an ever-evolving, ongoing process. As soon as new data is available, existing models should be updated and fine-tuned. You should constantly calibrate your models to take advantage of new data sets out there. To streamline the process, techniques such as data sampling, which allow analysts to use representative subsets of data in their predictions, may be used. This saves you from studying the entire data set with each new operation.
Automation is necessary, but expertise is what you’re really looking for.
Working with large databases requires both a high level of automation and human supervision. However, this is not the aspect that will occupy most of your analysts’ time. They will be confronted with a much bigger issue, one that can affect all types of analytics: the diversity of data, especially the diversity of data formats. Before a machine can learn or predict, the data has to be standardized. In many cases, this process requires human intervention.
3. How much should you use it?
To answer this question, you should first make sure the problem you’re facing or the objective you’re trying to reach, once solved or achieved, will save you money, time and effort. As mentioned above, predictive analytics can lead to exceptional ROI. If your company fits the bill for incorporating the technique, the long-term profits can fulfill your expectations.
Second, even if you don’t immediately see results from using predictive analytics, you’ll gain access to a source of structured knowledge that will enable you to position your organization for future growth. If you plan on using the technique long-term, you should also look into the collateral benefits that this knowledge may bring: access to such knowledge can also fuel other types of analytics and predictions.
4. What’s the Next Step?
If predictive analytics is already working for your company, you might want to see what the next business analytics “trick” is. After finding out what could happen in the future, most managers will naturally want to know what their business can and should do about it. This is where prescriptive analytics, a set of techniques that allow you to test potential solutions and possible scenarios, comes in.
If you’re unsure about using prescriptive analytics, chances are you’re not alone. A 2015 Gartner research quoted by Robert Hetu in his article, “Retailers Increasing Predictive Analytics Capabilities,” shows prescriptive analytics was employed by less than 5% of the companies from the research group. The results aren’t surprising: since prescriptive analytics relies on decisions and solutions that stem from the results of predictive analytics, you need to get the latter right before the former can amaze you.
Both predictive and prescriptive analytics, along with standard descriptive analytics, are part of what we can call the “analytics continuum.” Descriptive analytics uses historical data to describe, report, or explain what happened. Predictive analytics aims to find out if and when an event will occur again. And prescriptive analytics tells you what to do to repeat or prevent specific behavior.
Predictive analytics is neither an illusion nor a marketing promise. It’s a practice that will most certainly become standard. And the best part about it? All you have to do to reap its benefits is set the platform up, wait for the model to work its magic, and… let it blossom.
Measuring Sales Team Effectiveness Through Analytics
Which sales KPIs are most relevant to evaluating the success of new sales hires? And how do you use coaching to improve their performance? Watch this on-demand webinar to find out more.