Why predictive models are better than a simple average

By admin

June 14, 2013

0 Comments

During a recent sales call, we were asked if instead of our predictive modeling app could they use “average days to call” as a metric? They were referring to our cfTIME app for Salesforce, which predicts time to next purchase and then determines the right time to contact the customer to increase the probability of a sale.

So, you could look at two definitions for average (didn’t you know, in statistics there are two of everything, that’s how we cover our bases!):

Average buying cycle (ABC). This can be defined as 365 days divided by number of orders. If there are 8 orders in a year, then 365/8 = 47 days. Note that you can also choose a calendar year or trailing twelve months.
Time to Next Order (TNO). This is the expected days to next order. One approach is to average out the previous few orders. So let’s say the last three orders were 25, 10 and 45 days apart. The average of this is (25+10+52)/3 = 29 days.

Now, the answer to this question has two parts:

Technical feasibility
Build vs. buy investment

Technically, our models predict at an accuracy of 80%. That means at any given point in time, 80% of the customers will buy within the window we predict for them.

But if you used an average, assuming a normal distribution, here is what you will find:

Predictive-Model-versus-Average

As you can see, only a small percentage of customers fit within an optimum time window for a sales contact. Customers whose buying cycle falls short of this window are often your most valuable customers. And customers who are above this average are less valuable customers who will take up a disproportionate share of your resources and will likely not yield a positive return.

Next is the question of ROI for an app such as cfTIME. What will it cost to build vs. buy? When you buy an application, the app price is a transparent cost. However when you build, there are a lot of hidden costs that cannot truly be accounted for. Employee time, hardware and software costs, efficiency (or lack thereof) are some of the costs you have to take into consideration. Further, there is an opportunity cost when your marketing vice president is spending a lot of time on something that is less effective at moving the needle in terms of sales.

A predictive model is, quite simply, an outcome-based exercise. We are either predicting more sales, more orders, higher average order sale or more categories. If a model does not deliver against those outcomes – which can be measured by applying rigorous methods – it can be modified to improve the outcomes. A simple “average” cannot deliver this.

We definitely support using simple metrics in many cases. Averages are directionally good, but if misused operationally they can have devastating consequences – ones that you may or may not be able to measure until it is too late.

Sales Managers: Are you a geek at heart? Do you like to drive results through numbers?

By admin

October 9, 2012

0 Comments

Over the course of our careers, we’ve had the privilege of working with dozens of sales managers. In many ways, managers hold the key to the success of their sales reps. They know each and every rep well, understand the customer and prospect mix each rep has, the market that reps are operating in, the relative difficulty of making the quota, etc. They capture and impart this knowledge during their one-on-one discussions with reps.

Such tribal knowledge is often not institutionalized throughout the sales organization. The result is lack of repeatability, consistency and scalability of results – which means they are starting over with each new rep, team and division that they move into management.

Certainly factors like compensation, incentives, coaching and product training play a role in this success too. But we see distinct traits in the sales managers that are attracted to work with us – they use numbers, statistics, data and models to get the best performance from their sales teams. Here are a few traits that we’ve seen, some challenges that they’ve faced and why they’ve chosen to work with us:

TRAIT: They desire to “do something different this time.”

As ideas like call blocks, teaming and Hawaii trips get suggested, they realize these are maybe one-off, not trackable or simply not proven to bring enough of the performance they need.

CHALLENGE: They are looking for a scalable solution they can point to every rep and say “do this, it works, did you do this?”

REASON: Predictive analytics provides that scalable, stable and tangible call to action that every rep can do. By using lead generation solutions liked scored, quality leads, managers can ensure increased sales productivity for every rep.

TRAIT: They play with large spreadsheets – a lot.

These managers often have downloaded the entire customer data from Salesforce, pivoted it, summarized it and they’ve looked for patterns to share with the reps.

CHALLENGE: They know they are on to something, but this is on top of everything they have to do. And although they have great Excel skills, they are running out of time … or memory on their laptops!

REASON: Predictive analytics gives them all the information they need on their fingertips. Instead of trying to do the mathematics and analysis themselves, they can now see the patterns forming and be left to do what they do best: manage and coach the reps.

TRAIT: They constantly seek new data.

Sales managers with a quantitative leaning know that data they have may not be enough to make the sales cycle quicker and certain.

CHALLENGE: They are buying leads from various sources, appending demographic data, mixing in product usage data and contact summaries to create a broader picture. They tend to look at social media data too. With so many sources to keep track of, analysis of their sources and quality control can become quite problematic.

REASON: Predictive analytics can give them the broader, larger picture, compiling the data, trends and statistics from all sources.

TRAIT: They focus on “the middle 60%.”

They leave their top 20% of sales reps alone, and they know who their bottom 20% are, but they want to improve results for everyone else. They look to analytical methods for improvement.

CHALLENGE: They know there is a cyclical effect that helps the entire organization, and these middle 60% are the difference between making the numbers this month vs. making numbers consistently. They believe in incremental change. The sales managers we work with want to improve x% of the reps, get x% more customers this month, increase average order by x% and penetrate x% more categories. They want consistent gains across many dimensions that are both scalable and sustainable.

REASON: Using predictive analytics tools to score leads, to gauge customer timing and to stay on top of buyer cycles can reduce variability and give the middle tier reps a big productivity boost, thus creating consistency across all reps, almost like making “the rest like the best.”

Incidentally many of these sales managers are also in growing companies, and they’re hungry for even faster growth. They act as a buffer between understanding the math and delivering a simple message to the reps.

Most of all, they are very inquisitive and inherently believe that an analytical approach does work. They have to believe this, in order to give us even 10 minutes of their precious selling time. That makes all the difference.

Inside sales: Playing by the numbers

By admin

October 4, 2012

0 Comments

One thread that ran through the sessions at AA-ISP’s inside sales conference in Chicago last week was: 1384137682395.

That’s right. 1384137682395.

Dollars. Percentages. Rates. Ratios. “Batting averages,” even.

Numbers.

The language of numbers, thinking analytically and driving decisions based on numbers was a shared language throughout the conference. And numbers don’t have to be complicated; many lessons were simple but still powerful for sales productivity. Here are a few we heard:

People need to receive an average of 6-7 lead nurturing contacts by marketing before they are sales ready.
The close ratio of “buying signal leads” versus “tire kicker leads” is 8:1. Buying signal leads request pricing, demos and trials. Tire kicker leads download white papers and attend webinars.
LinkedIn messages can return a response rate 3x more than email; LinkedIn InMail messages can return a response rate 30x more than email.
Email stretches out sales communications and sales cycles from what could have been 5 minutes to 5 days or 5 weeks or more. Don’t hide behind email. Have conversations.
Test different strategies for cold calls and initial conversations. One strategy may deliver 130 product demos while another strategy may deliver 390 product demos.

Implementing tests and processes that produce these numerical insights, and then using the numbers to guide sales actions, leads to increased effectiveness and productivity which are then measured with, you guessed it, obviously more numbers.

This is at the core of what we know at Valgen, that human choices and behaviors lead to numbers that can be identified and used to decrease costs, improve the productivity of sales and increase revenue.

A Valgen client shared how this was proven within his inside sales environment during a breakout session at AA-ISP Chicago. Our partnership resulted in 70% fewer lead duplicates and a 40% increase in lead conversion. We gave you a sneak preview of the lead generation lessons that led to those results in a recent blog post.

Three ways to improve outbound lead generation

By admin

September 25, 2012

1 Comment

The first step to improving outbound lead generation is to better understand pipeline activity. An analytical approach can help sales managers to create efficiencies in sales rep activities, resulting in improved lead generation outcomes:

Sales Activity

Lead assignment to reps – Placing a manageable quantity of leads in each sales reps’ queue on a timely basis so they can start calling, qualifying and closing deals without being overwhelmed.

Lead disposition by reps – The “why” of determining which leads are not qualified or worth pursuing is just as important as converting a qualified lead into an opportunity.

Time spent on non-sales activity – Creating action lists that can be used to prioritize, assemble and validate actionable info

How can analytics make reps more efficient with these sales activities?

Lead assignment: Reps disqualify leads at a faster rate than they qualify them. The assignment of new leads can be automated based on total current active leads as a ratio of closed leads, and the relative volume of leads compared to other reps.

Lead disposition: Disposition reasons that are clear, accurate and consistent can yield valuable info to fix specific problems. For example, if the reasons show bad data, go back to the vendor and get updated records or a credit. If leads are not in addressable market, then rank the highest such records and suppress them when buying future leads. When there is no current opportunity with prospects, maintain communication (with permission) and incorporate statistical models so you can reach similar prospects earlier in their buying cycles.

Time spent on non-sales activity: Reps need help prioritizing. Analytics can help in sorting, filtering, appending and synthesizing valuable info across prospects, accounts and transactions, and then present actionable information that can be used in a script. For example, we have created an app within Salesforce that combines account data with prospect data to give confidence and talking points to sales reps.

So how does this produce predictable, scalable sales growth?

By incorporating analytical methods, you reduce variability and create consistency across all reps – sort of like “making the rest like your best.” This is the biggest benefit to sales managers because reps can then follow what’s proven to work, consistently.

Next, sales reps can make analytics work for them. Any analytics process is only as good as the feedback it gets. By giving timely feedback, results can be improved continuously over time.

Finally, analytics can leverage big data and make it relevant to the reps, at each and every interaction with prospects. Analytics take care of the heavy lifting, so sales teams can focus on achieving sales outcomes.

By 1.) removing variability, 2.) providing feedback and 3.) leveraging big data, you can produce scalable sales growth.

Using temps to clean sales data – do’s and dont’s

By admin

September 19, 2012

0 Comments

We have been in organizations that have tried to use temps for sales data cleansing, and now have clients that do this. We know why you do it, and what you seek to achieve. But often we see this endeavor is not done well and as a result, doesn’t get you where you want to go.

We offer database and sales analytics software and services. Temp cleaning is not what we offer. We do clean customer data not as a stand-alone service, but to add more value so that our predictive models are more accurate and actionable. We use structured cleansing and matching processes, B2B databases, rigorous profiling and analytical validation procedures.

But here we’ll share our observations and tips about using temps for your benefit. This is because we are passionate about data, and have seen many instances where data cleansing was not done right, and later we had to work harder to fix it. We sincerely hope you find these tips valuable.

IT, Sales and Marketing should inform each other before data cleansing starts. Just like you wouldn’t dig a trench in your yard without calling Julie – we hope! – the group that initiates this activity must be transparent and inclusive of the others. It’s not about “who owns the data,” but “who knows the data.” Depending on the fields or tables, this can be IT, sales or marketing.
Give each temp clear instruction to follow. They must be very familiar with the instructions and also the nuances as to specific fields and circumstances, because there will be an unavoidable level of subjectivity. Input sample records and validate if they were corrected as per the instructions.
Determine which fields are appropriate for temp cleanup. Usual fields for temp cleanup are contact (name, title, company, country), market (industry and employee size), transaction data (i.e., missing values), and internal (rep name/IDs, validations etc.). But the most value from manual cleaning is internal data that requires knowledge of your processes. Postal and market data can be appended through other sources more consistently and cost effectively.
Recognize even the same fields from different sources are not all the same. Fields like SIC Codes (industry) are different – sometimes significantly – based on the compiler’s algorithms and information. Avoid mix and match if you can get high coverage with one.
Use multiple methods of validation (in-process and post-process). Address fields can be run through CASS and NCOA for a fraction of the cost than, for example, dialing to get the right ZIP Code. Incorporate validation by temps, reps and external sources.
Define the way forward. Before you start correcting the data, think through and set guidelines for future corrections – frequency, volume as well as expected evolving standards.

So, yes you can use temps, but be cautious in your approach, scope, consistency and validation. Also factor in the cost of alternate methods including out-of-pocket, time (opportunity cost), anticipated future costs and coverage trade-offs. There is no one size fits all.