Predictive Analytics Frameworks for Revenue Planning

Project Description

The Ask:

We need a consultant to apply predictive analytics to their core business (Talent Platform) for revenue planning purposes.  The core business generates revenue by seconding out our attorney’s to client locations where they do work to support then-house counsel of those clients ranging from maternity leave covers to M&A transaction support documentation. Their core business generates revenue from “engagements” spanning the course of days, months or years.  The engagements end date is flexible and can end early or “extend” to a future date at any given time per the client’s needs. However, there is a wide range of engagement LTV. The firm needs help predicting how their current (and future / to be won) engagements will perform based on historical analysis of past engagement outcomes (using the profile / characteristics of those engagements). 

We have put together initial research/data that the consultant will build off. The final deliverable will be a board-ready presentation that outlines the predicted future behaviorof the “book” (current engagements / mixed with assumption on future “to be won” engagements).

The Details:

1-2 week framework starting mid to late May 2017

The Audience:

The executive leadership team and business planning team will use these outputs and assumptions to support outward looking planning models

Business Concerns

The issue: Our “book of business” is a portfolio of active engagements at any given time with an estimated end date. Based on the estimated end date we can forecast how much revenue we will generate from those active engagements in the future.

However, the variance between the estimated end date and the actual end date is quite large. On average our engagements extend ~+40% against the initial estimate. While we see a consistent variance at a portfolio level if we were too look at an individual engagement level we would see a wide range of variances. Some engagements end earlier than expected and some engagements go +5X the initial estimated time frame. Our business planning team wants to do a “double click” and apply predictive analytics to each engagement to predict what the actual revenue from these engagements will be on our current “book of business”.

Why is this important?

Understanding how our book of business will perform for the rest of year (and in future years) is a mission critical piece of information for our business team in annual planning and determining sales staffing needs. We will have to build new business assumptions on top of our current book of business which then drives staffing and cost investments to support corporate growth objectives. In other words, whatever revenue is not generated from our current book must be won by sales teams to drive revenue growth.

What does success look like?

There are three main outputs of this exercise.

1. A predictively analytic framework on our forward book of business that predicts our 2018 carryover revenue performance by engagement within +-5% degree of accuracy (total book forecast vs. actuals accuracy)
2. A stand-alone model/segmentation of categorized engagement types with extension multipliers that can be refreshed going forward and used in future forecasting efforts
3. A list of proposed data points that are currently not being captured that could increase the accuracy of this framework in the future

Data Assets

● All data is in excel

● Customer segmentation and existing customers/prospects data

● Historical actuals – engagement-level performance against +6,000 engagements from 2014-2017 ytd

● Engagement performance is captured in our “book of business” reports which includes

     ○ The business team (sales team) that won the engagement and is responsible for managing the end date estimates

     ○ The primarily legal specialty of the attorney working the engagement

     ○ The client name

     ○ The engagement name

     ○ The resource name (attorney name)

     ○ The estimated start and estimated end date (actual end date included when               engagement has ended)

     ○ Revenue generated per month

     ○ The billing type of each engagement (is it a fixed fee per month, day / or is it estimated hours to be worked (hourly) or estimated days (daily) rate)

     ○ The estimated utilization (hours, day, fixed fee) per month for each engagement

     ○ The geographical location of the engagement (driven by business team attribute)

Data Analytics

Historical approach: Historically, the main reason we would run an outlook on our book of business was to predict “carry over”. Carry over is the amount of revenue we have in each year that is generated from revenue sold/new sales in prior years. Towards the end of a given year, during business planning for the following year, we would need to assume how much carry over we would receive and then we would be able to make a statement on how much additional revenue (new business) we would need our sales teams to win. Our approach was to triangulate an answer using different methodologies.

First, we would look at historical carry over performance and use that historical range and apply that to the current year (with assumptions on how much our current year book would grow by the end of the year / usually using our corporate revenue forecasts).

Second, we would look at an engagement by engagement level and determine the average age of each engagement and apply a predicted age (push out engagement with end dates that fell short of the historical average duration) which resulted in an assumed revenue outlook for carry over. Lastly, we would apply a top down adjustment to the carry over forecasts for any large (outlier) engagements that would move the needle and we have better line of sight to via communication with the salesperson managing that engagement.