Building a new software category is a battle. Want a few war stories?
In honor of Dreamforce 2015, I thought I'd let you in on my top 3 sources of scar tissue and sleepless nights since Dreamforce 2014. I hope a glimpse of how the sausage is made is helpful and interesting.
First, when I say battle, I don't mean against competitors. I mean the brutal effort to wake up every day and figure out how to create an end-to-end solution loved by reps, loved by managers, and loved by execs. We sell predictive sales analytics and so far the results have been better than we expected—but, wow, has it been hard.
So, it's time to open up about some things we figured out along the way.
Predictive is super hard
It's not the math—or not just the math—that makes predictive hard. What makes predictive hard is (a) making it fit the work of sales pros with a quota to hit and (b) making it fit across different companies, sales models, and verticals, extracting what's different about each one.
Predictive starts with data. We started with CRM data because it's the traditional system of record. But with a team full of veteran sales folks, we knew that CRM alone wasn't enough. CRM contains what reps have time to say happened, but it doesn't capture what reps and customers really do.
So we added the real source of truth: email and calendar activity. What's obvious is emails and meetings take two to tango. Email exchanges are with customers and meetings are with customers. Your email and calendar is the best way to measure both rep and customer engagement. Only with deep analytics on email and calendar can you know how engaged customers are and what rep activity patterns are most effective. It's one thing to hope that the customer is interested when a rep says, "I'm all over that account." It's another to know that the internal champion replies to every email and the exec sponsor replied within 48 hours of the redlines being sent over. If you consider predictive analytics for sales that doesn't learn about customers by analyzing email and calendar, it's more than having one hand tied behind your back—it's wearing a blindfold.
OK, you're now giving the right data to your killer data scientists. Home free? Not even close. Prediction in enterprise software is still early. So making customers comfortable using predictive to boost forecasting and selling requires three things:
1. Be right
I got an email a few days ago from a VP of Sales saying, "The fact that you called the number more accurately than I did is scary." Truth is, it's not "scary," it's required. And hard. Other companies claim to "call the number" by simply using weighted averages based on, say, "same time last year" or "same time last quarter" comparisons. That's not predictive. It's more like a glorified spreadsheet. You have to compare the communication flow on each deal to that of similar past deals won or lost.
A related lesson: Sometimes, when you're sweating to make the quarter, being right at "calling the number" isn't as critical as being right with 95+% accuracy about whether each individual deal is on track or at risk—early enough to take action before the quarter close.
2. Be clear
People hate black boxes. Trusting an algorithm without knowing how it works doesn't fly with sales leaders. So while we run complex algorithms, we show in a clear, simple way how and why deals are at risk. So simple that any sales exec can understand it!
3. Be fast
Nothing cuts exec excitement about predictive like hearing about a 4-6 week delay to "get your data and refine our algorithms." We've got it down to 2 minutes because that's what people need. On a personal note—this may sound strange, but one of the things I'm most proud of is our ability to deploy with no expensive, time-consuming professional services. We turn it on. It works. Some customers don't believe us before they log on. But the moment they do, BOOM! Seeing is believing.
Prescriptive is easier
Prescriptive recommendations—who to call, when to demo, or when to re-engage with an old customer—seem like magic. But once you get predictive right, prescriptive is not that hard. I know—that's like saying, "Once you've studied for the mid-term, acing the test is a snap." But it's true. Once you know which cadence of meetings, document flows, relationships, etc. means a deal is likely to close, the prescriptive algorithms you need to design are not that difficult. (OK, I'm not the one writing those algorithms, so let's just say "conceptually" not that difficult.) Essentially, you work backwards from the pattern needed for success to figure out what actions can get an at-risk deals back on track or what actions can move Best Case deals into Commit.
If anyone talks about a prescriptive system—something that tells your team how to perform better—before nailing their ability to predict whether deals will close or not, run. They are taking a shortcut based on hunches or conventional wisdom. Your team, product, and customers are unique and you need to have confidence in the predictions first.
And just like with predictive, make sure any vendor can prove themselves on the big three above: be right, be clear, and be fast. Do you really have time for wrong, confusing, and slow? I say that with a smile, but really, isn't that the alternative? Missing any one of those three means a waste of time and money.
Real-time visibility is as important as predictive
Predicting the future and getting prescriptive about what next action will increase revenue is dramatic and will be a part of every sales leader's future. But nothing fires up the pattern-matching experience of an exec or manager like a real-time understanding—with a single glance—of what's happening in every deal, what's changing, and how engaged the customer is. Real-time visibility equals speed: better coaching in time to make the upcoming meeting successful, better resource allocation in time to make the quarter, better forecasting based on up-to-the-minute deal insight.
With CRM, visibility into deals has been challenging for a long time. After 7 clicks, you'll see the latest view of whatever your rep remembered to add. But that's nothing like a mothership engine watching for changes as they happen—in CRM, email, calendar, and more. An engine able to display deal progress (or lack of it) in a crystal-clear dashboard or send an immediate alert when a deal in Commit is looking shaky.
I've said it before that if someone is trying to recommend what your team should do without being able to predict the future, they are selling snake oil. It's similar: If someone claims they "call the number," without letting you drill down into the details, changes, and trends in a deal, they are just hoping all their errors cancel out.
That's my top 3: predictive is hard, prescriptive is easier, and real-time visibility is king.
Put these together and the result is totally predictable: better forecasting for execs, more effective coaching to drive revenue for managers, and higher close rates for reps.