Menu
Predictive Sales Analytics

The Secret to Sales Acceleration: Predictive Sales Analytics.


Can salespeople predict the future?

This question is not as crazy as it sounds and in fact the answer is moving closer and closer to yes everyday.  The predictions though don’t come from a crystal ball – they come from a computer and we are going to go on a journey to show how it’s possible and how you can start taking advantage of predictive sales analytics.

Time to play some Powerball? – Random vs. predictable events

Firstly, what do we mean by predict the future?  Do we mean that we can predict the next winning Powerball numbers?  Well unfortunately not because it is a completely random event.  The ping pong ball machine is not influenced by previous numbers that have been picked.  This is similar to every time you flip a coin the odds are 50/50 and that never changes (assuming a fair flip).  Even if you flip heads 9 times in a row the odds of the tenth flip is still 50/50.  The coin doesn’t remember the previous 9 flips.  What is unlikely is the combination of the previous 9 flips all being heads not the odds of the 10th flip.  Thinking about the entire set of events is very important but more on that in a bit.  

So we can’t predict the outcome of a lottery but we are interested in predicting people not ping pong balls anyway.  The good news is people are a very different case because they have memory and are a summation of their experiences.  We don’t have to have a philosophical debate here (which is fun) we only have to agree that people are in fact not random (like the coin or ball).  Their behavior is mostly rational and as a result can be predicted with ‘some certainty’ if given enough data.

You are your Starbucks order – Behavioral data makes prediction possible

Now the statement ‘with some certainty’ is important.  Can we predict exactly what coffee you are going to order at Starbucks 100% of the time?  No we can’t because even Starbucks customers change things up every once in awhile.  We can however predict with fairly high certainty what you will order and it’s based on your purchase history.  The simple solution is just assume the customer will order the same thing they ordered last time.  But it can get much more sophisticated based on the amount of data that can be analyzed.  A more data driven approach will look for the most common beverage you have ordered. Then layer in time of year, temperature (iced latte?), location and the predictions can get more specific and more accurate based on more available data.  There is some crazy math and data collection behind this but you don’t need to worry about that.  The point is prediction with fairly high certainty is possible given enough data and a specific behavior.  You also need something called a prediction model but that will be the subject of a future blog.  

Big retailers are making big investments – Predictive analytics drive revenue

It turns out that an increasing number of retailers do exactly what I have described above and are getting better at it everyday.  There was a controversial case a few years ago where Target could/can predict when a customer was pregnant and could tailor their marketing efforts appropriately.  It became a New York Times article and lots of discussion and controversy ensued.  The basics of the story is based on a customer’s purchase history and then married up with their prediction model they were able to know with some certainty if someone was pregnant.  What followed was coupons from Target for various Mom and baby products with the objective of establishing new purchasing habits for the newly required products.  The results for Target were impressive because they knew a life changing and purchase changing event ahead of their competition.  Often ahead of other family members.  They predicted the future based on a statistical probability informed by data.

What about the individual B2B sales professional?

Things these big retailers have in common is they have huge amounts of customer data, large technical teams, people on staff with titles like ‘Data Scientist’ and significant budgets. Most sales professionals don’t have these luxuries so the question is can we get predictive analytics into the hands of more ‘normal’ users?  Can sales professionals use predictive analytics to predict interesting things about their customers and prospects to help them be more successful?  The answer is yes but requires some new skills and a bit of new technology.  Both are pretty straightforward given the reward.

The rewards are about significantly increasing sales results by predicting where sales professionals should focus time and effort to further increase their odds of closing a deal.  No two prospects are the same once you have some data driving your decisions.  Focusing on the most important ones first, will improve your results over time.  To do this well you need data about your customer and prospects and you need a framework to help you prioritize your time.  That’s where the computer part comes in.  Once you have those pieces in place, you ability to predict will improve everyday.

It all about probabilities

In the Starbucks example it’s fairly easy to see that the accuracy of the prediction is very dependent on the amount of data that we have about the individual.  It’s a probability that gets stronger as we know more about the individual and their purchase history.  More data and better models will get us ever closer to 100% certainty of predicting their next Java experience.

The same applies in the B2B sales professional’s case.  Data is the differentiator and then it needs to get applied to a predictive model that will drive priority decisions.  Sounds complicated?  It doesn’t have to be.  You can start very simple and get more sophisticated over time.  The most important thing to realize is data collection is absolutely critical.  You cannot analyze what you don’t have and you are essentially flying blind without data.  The kind of data can actually vary a fair amount but for purposes of this article we are going to focus on buyer behavior data.  This can be the easiest to get and also some of the most useful in terms of predictive analytics.

A discussion about probabilities is absolutely critical to understanding predictive analytics in general and it’s not as complicated as it sounds.  The most important point is that pretty much everything we do is governed by probabilities and its possible to maximize your outcome by understanding those probabilities and making decisions to maximize the result in your favor.  Despite my pithy title, we are not predicting the future directly but rather predicting where to spend time that produces the best future.  That distinction is important.  Every decision that you make can be optimized by thinking about probabilities.  

  • What is the best email messaging to send?
  • What is the best time to send a message?
  • What is an optimal response time?
  • When should a call my customer?
  • What role should I target?
  • Who should I call first?
  • Is this customer actually interested?
  • And so on…

The decisions a seller makes in answering and acting on all of these questions can produce very different results.  Our goal is to find the best results.  Now if we just had a way of making this easier for sales professionals on a daily basis that would be great.  As it turns out we are making great progress on just that.

A sales process is a series of mini-milestones – Where to start?

Most salespeople are well aware of the principle that closing any deal (getting the big ‘yes’) is actually a series of smaller yeses on the journey to the big one.  This is essentially a series of mini-milestones that can more easily be measured, tested and optimized.  Think of each mini-milestone as the behavior we want to measure and predict just like we would try to predict the Starbucks order.  Break down your sales process into simplified steps and implementing predictive analytics gets easier on almost every level – process and technology.

Initially, in many sales processes, the first mini-milestone is just having a conversation (getting an appointment) with the prospect.  Tons of inside sales teams are just focused on ‘getting the appointment.’  This is no different to what you are likely doing today, the difference will come in how you need to track it and make decisions based on the tracked data.  You need to develop some predictive analytics skills around each mini-milestone. Another way of saying this is we need to maximize the conversion rate of each milestone and use predictive analytics as one of our guides.

So our first analytics goal is to get more appointments at the front end of our sales process.  That’s pretty valuable stuff or almost any organization.  Let the games begin.  We need to create a series of tests against that list of contacts in the form of some structured communication which is primarily email and phone (voicemail).  Here is the general approach.

Sales Lead -> Measurable process -> Structured results -> Prioritize Lead -> Repeat

One of the data points that you will actually be evaluating in this process is your lead data quality.  This is a critical and often overlooked problem in sales teams.  The quality of their lead data is terrible but all leads are treated equal.  This is horribly inefficient because calling bad numbers for example is just as expensive as calling good ones.  It’s actually worse because calling bad numbers is demoralizing for your team.  So using a data driven approach you will start to tell the difference in lead quality and that will start drive your lead prioritization.  You will get bad leads out earlier.

Creating a structured and measurable process

Generating good analytics data, requires some consistency.  For example, if you use random messaging (voicemails and emails) in all of your customer communications it is almost impossible to measure what’s going on and what’s working and what is not working.  Interestingly, this can be a fun and engaging exercise with your team that will reward them and also start to teach them some of the basics of the importance and power of predictive analytics.

One great way to start to build your analytics muscle is to have a competition with your team on the messaging that they use.  Essentially encourage the team to continue doing what they are doing but start to track the results at a very granular level.  For example, each rep uses their own email templates but you start tracking the results of those templates.  

You will need software to help here so this is done automatically and ideally stored directly in your CRM system.  This will create some interesting but fun competition between your team members but you will start to get insight into what messaging approach works best.  Think of it as A/B testing but across your entire team – A/B/C/D/E….testing.  Once you have a winner go ahead and reward that team member but everyone in a sense wins because they should all start using the improved messaging.

You should do the same testing with your voicemails.  In the case of voicemails, it is useful to make sure that your team is using voicemail scripts or they use pre-recorded voicemails.  Either is fine as long as you have some consistency in the messages that are delivered.  Then in a similar way to the email case you need to track effectiveness, reward a winner and share the messaging with the rest of the team.  The report below is an example of this information that can be gathered simply by starting to track your team’s communication.  From this report it is clear that the email with the subject ‘Prospecting Re-imaged’ has done well for us and should be used as one of our outbound messages.

This may seem basic but very few organizations do it well.  More importantly it is laying the foundations of a continuously learning organization that can take advantage of predictive analytics in the future.  In this example, we communicated with customers, tracked that communication, stored it and then used it to determine the best messaging.  This is the same basic pattern that we will use over and over but just get more sophisticated in the data and sales goal we are looking to optimize.

Template Stats

While this message testing is going on (measuring the open and click rates of sales messages) something else is happening at the same time.  That data can also be correlated with specific prospects.  This is engagement data tracked at the individual user.  This is very valuable because it will start to provide insight on your most interested customers.  So while you are testing your messaging you are also starting to prioritize your lead list.  You can start to group them based on their behavior (replied, clicked, did nothing) and based on that information it possible to create a more tailored follow up.  For example, you can create a simple workflow that has your reps only call customers that have opened or replied to your messaging.  This starts to filter out bad contacts and also gives your reps context for the call.  They should use structured messaging in their voicemails at this step as well.
Predictive Stats

In this small example, start calling from the top of this list which is prioritized based on the actions of the recipient.  The customers that have already shown some intent by engaging with your content are also the customer that have the highest probably of continuing to engage.  It doesn’t guarantee that they will engage, just that they are more likely to do so.  Each turn of the communication crank provides more prioritized leads for your reps increasing their effectiveness.  No more guessing or treating your lead list as flat or equal – they aren’t.

Predictive Forecasting

We have barely started to refine our sales process to be more data driven and therefore more predictive in the previous examples but it illustrates the process.  You will be able to create better messaging and prioritized leads without turning your team upside down.  The benefits for the frontline rep starts to become clear but what about sales management and that difficult job of forecasting.  Is it possible to use predictive techniques to forecast or as some sales leaders call it ‘guesscasting’.  

As mentioned above, the general approach to using predictive analytics is to create mini-milestones throughout your sales process and use data to help prioritize decisions around achieving the result.  Built into this milestones, needs to be markers for the customer actually closing or being at the right qualification level to make it to your forecast.  One simple example is was a proposal shared with the customer, did they view it and specifically did they view the pricing page?  It is then possible to build a fairly simple and data driven model that says if they prospect achieved certain mini-milestones and also viewed the pricing page they are likely enough to close that they should be in the forecast.  Notice this approach is almost entirely data driven based on the actions the prospect has taken and not the sometime over optimistic view of the sales person.  Optimism is a wonderful thing but can wreak havoc with a sales forecast.

The latter parts of the process is where predictive analytics gets to most complicated but also can deliver even more value.  This is where you can start to compare the executing of each mini-milestone and how they relate to each other.  Most importantly for customers that actually closed what path did they take.  This information can then be fed back into the system to get smarter with each iteration.

Making the case for Predictive Analytics

Predictive analytics is a tool.  An organization that has the ability to implement and use predictive analytics is the real goal.  An organization that has processes capable of structuring and measuring customer interaction and continuously learning from customers is the one that will win in the long term.  

The decision to start down the path with predictive analytics does not have to be a big one.  There are many small steps that can be taken at low cost and low risk to start to see if an organization can truly benefit.  The rewards can be substantial not only in terms of revenue growth but also getting to know the customer better and even getting that information over to customer success and engineering organizations.  Today there is very little learning that flows from sales into those organizations, yet sales is in one of the best positions to learn from the customer.

The real winner though can be the sales organization.  Predictive sales analytics holds the promise of delivering much more prioritized actions and therefore better results.  All that’s needed is to get started.



Post a Comment

*