ToolsGroup-Going Back to Its SCP Roots

Last fall we had the chance to reconnect with ToolsGroup during its North American User Forum in Boston. An expert in demand-driven supply chain planning (SCP), the company offers sophisticated software that analyzes demand history across multiple dimensions, enabling companies to obtain more accurate forecasts and inventory targets for driving outstanding customer-service levels with less global inventory in their supply chains. ToolsGroup’s solutions span key SCP areas such as demand planning (including demand collaboration, demand sensing, and promotion forecasting) as well as inventory optimization (advanced inventory modeling, product mix optimization, multi-echelon postponement/staging optimization, lot size optimization). These innovative and advanced technologies enable enterprises to improve and automate their planning processes.

ToolsGroup has more than 200 customers worldwide (in 31 countries) and one of the highest customer retention rates in the industry. The company has experience working with manufacturers, wholesale distributors, and retailers in a wide range of demand- and inventory-driven industries. ToolsGroup’s vertical focus hasn’t changed since its inception in 1993 in Europe: distribution-intensive verticals with challenging demand or inventory problems. The company’s three biggest verticals are consumer products, wholesale distribution and retail, and aftermarket parts.

ToolsGroup typically focuses globally on upper tier customers (over $2 billion [USD] in annual sales). “Upper mid-market” companies ($200 million [USD] to $2 billion [USD]), long a mainstay in Europe and other parts of the world, have now also become a focal point for the United States (US) market. Many of ToolsGroup’s customers run on SAP and Microsoft Dynamics AX enterprise resource planning (ERP) platforms. ToolsGroup systems have also been integrated with JDA Software (including former i2 Technologies and Manugistics), Oracle, and other legacy ERP and/or supply chain management (SCM) systems.

ToolsGroup is headquartered in Boston, Massachusetts, and in Amsterdam, the Netherlands. The company also has offices in major cities throughout Europe and distribution partners around the world. The company operates out of five offices in Europe, three in North America, and one in Cape Town, South Africa. ToolsGroup partners represent the vendor in other countries, such as Sweden, Israel, and Poland, while a new representative in Japan is translating the product interface into Kanji. 


ToolsGroup Over the Years
ToolsGroup was created 18 years ago to build demand and inventory software solutions for aftermarket parts planning. At that time, SCP was based on a variety of advanced planning and scheduling (APS) solutions, all based on outdated modeling technologies (1960s Scientific Inventory Theory and Forecasting methods) and “deterministic” optimization algorithms. Deterministic algorithms, in informal terms, behave predictably: given a particular input, they always produce the same output, and the underlying machine always passes through the same sequence of states.

The same approach characterizes many systems that are still used today, despite the inherent weak approximations. The lack of modeling accuracy and the reliability of this approach forces planners to spend a lot of time in tuning parameters and choosing options, carefully reviewing calculation details, and finally manually overriding a large portion of the system’s suggestions.

The aftermarket (service or spare) parts environment, with its many stock-keeping unit (SKU)-location combinations, high percentages of slow moving products, and very heterogeneous demand and supply behaviors across products, has required highly automated systems, with advanced self-tuning capabilities and statistically more rigorous modeling.

These requirements have now become common to many other vertical sectors, due to a number of factors (richer product ranges, more granular distribution network control, shorter time buckets) that have created the so-called “long-tail” phenomenon. Long tail refers to the statistical property that a larger share of a population rests within the tail of an exponential probability distribution than observed under a “normal” (or Gaussian distribution), which is symmetrical in nature.

In a vast majority of industrial sectors, planners have now to keep under control a much higher number of SKU locations, all characterized by a more volatile stochastic behavior, which is rather difficult to model. The typical “normal” (Gauss) distribution assumption can nowadays be used on a relatively small percentage of cases (from less than 5 percent to about 25 percent, depending on the sector and business characteristics), with the rest of the SKU locations requiring much more advanced statistical modeling capabilities.

For this aftermarket, probabilistic, randomized, or stochastic algorithms are more appropriate, as they determine whether a given number is prime and has a small chance of being wrong. For example, if the stock level at this location is X, then the service level (likelihood of no stockouts and missed deliveries) should be Y (say, 97 percent). Certain simplistic approaches (such as Croston’s method, or similar) designed to deal with the lumpy/intermittent demand behavior of items in the “long-tail,” can only provide some degree of protection against forecast instability, but it has been clearly proven that they can’t provide the modeling accuracy needed for reliable inventory modeling and optimization.

The use of such systems perpetuates the need to manage SCP in the traditional way: a sequence of scarcely integrated functions (demand forecasting, inventory management, and replenishment planning)—which, in general, are very manually intensive and prone to generate an undesirable degree of “bullwhip effect” (an unnecessary inventory buildup upstream) in the supply chain nodes. While they use some statistical elements here and there, there is no integrated statistical modeling of the overall supply chain.

What makes the ToolsGroup approach differentiating is that it is based on a fully integrated statistical modeling framework, which, for each supply chain node, provides a reliable statistical description (full probability distributions) of: the “flows” behavior (for each inbound and each outbound link), the inventory position at the node, and the performance provided (“stochastic” service levels and service times).

So, right from the beginning, ToolsGroup built solutions that modeled the demand and inventory planning processes without taking statistical shortcuts. The company’s first standard SCP software products were released in 1996. ToolsGroup has since expanded to other verticals, while maintaining rigorous statistical modeling as the backbone of the solution, encompassing multiple optimization functions and addressing demand and supply volatility.


SCP Redux
Many users and SCM software providers understood the problem of dealing with increasing demand and supply uncertainty and first tried to to get around it with the following advanced add-on modules:

  • Inventory optimization—Leveraging sophisticated Operations Research models to determine optimal inventory positions and replenishment lot sizes in a supply network.
  • Demand sensing—Looking for additional “visibility” on the network and trying to detect statistical patterns in the demand signal to better forecast demand at the SKU-location level.
  • Promotion modeling—Analyzing millions of promotional events via a statistical representation of promotions and advertising.

In 2003, ToolsGroup expanded to the US and saw the inventory optimization (IO) market become a large proportion of its local business. Nowadays, with consolidation and stand-alone IO providers being subsumed by SCP providers, SCP has once again become the majority of ToolsGroup’s business. Indeed, as seen by the recent acquisitions of Optiant by Logility and Logic-Tools by ILOG (and then IBM), the stand-alone IO market is being subsumed by SCP (SmartOps remains independent on paper, but most of its revenues come via the SAP Enterprise Inventory Optimization by SmartOps solution). 

IO is (and always has been) a subset of ToolsGroup’s SCP solution, but over the past two years or so, ToolsGroup’s new corporate positioning, particularly in the North American market, has been one of transitioning from IO to its broader SCP solution. Though already a large and competitive area, SCP continues to grow, particularly in the upper mid-market, and better represents the vendor’s solution footprint and unified global product message.

ToolsGroup competitors are no longer pure-play IO vendors (as most have folded their business), but full-fledged SCP vendors like JDA, Logility, and Demand Solutions. Logility, for example, claims recent growth owing to the integration of Optiant’s IO capabilities into the Logility Voyager Suite.


Statistically Laden and Responsive SCP
Most of the deterministic APS models use either general rules or bolt-on IO tools as a "Band-Aid.“ In contrast, ToolsGroup has always statistically modeled a comprehensive end-to-end supply chain solution with each function modeled statistically. The company believes that controlling volatile supply chains requires highly reliable and scalable statistical models. The statistical model is the backbone of the solution, encompassing multiple optimization functions and addressing demand and supply variability. ToolsGroup’s “full-on” approach (see figure) enables service level–driven planning that aligns all supply chain constituents and eliminates silo-based planning.

ToolsGroup’s solution suite, dubbed SO99+ (standing for “Service Optimizer 99+ Percent”) includes an analytical approach to understanding demand volatility and its impact on demand forecasting, demand sensing, and inventory optimization. Forecast accuracy and demand volatility consistently come up as the #1 issue in the vendor’s user surveys. Therefore, the vendor is streamlining efforts to get a better idea of the true demand signal, and align and synchronize it back through the rest of the upstream supply chain.

These demand signals and volatility call for a profitable supply chain response. A vendor can easily muster up a customer response with excessive inventory and morale- and energy-busting firefighting tactics, but a profitable response entails understanding the sources of variability and planning for a quick and effective answer. Statistically modeling demand volatility across multiple dimensions (line-order quantities, order-line frequency, forecast error) enables ToolsGroup to handle long tail “lumpy” demand (along with less complex fast movers), very high service levels (e.g., over 99 percent), vertically-integrated or multichannel distribution, and new product introductions (NPIs) or seasonal changeovers.


Discussion with ToolsGroup’s Marketing Expert
What follows now is an enlightening discussion with Jeff Bodenstab, vice president (VP) of marketing for ToolsGroup. Bodenstab is a computer software marketing executive with a unique track record of helping to build five young businesses into successful enterprises. Prior to ToolsGroup, in the late 1990s and early 2000s, he served as VP of marketing for North America at i2 Technologies, the world's largest SCP software company at that time, VP of industry marketing and VP of marketing for the automotive business unit, where he helped grow the business from a startup to more than $100 million (USD) annual license revenue in 4 years. Bodenstab holds an MBA from Harvard Business School and a BS in industrial and systems engineering from Lehigh University.

TEC: What are ToolsGroup’s particular capabilities and what pain points do they solve?
JB: Time and time again, user surveys indicate that forecast accuracy and demand volatility are the top supply chain concerns for most companies. To respond profitably means understanding the sources of variability and planning for them appropriately. We tackle the demand volatility problem head-on. Analyzing the demand characteristics, such as the shape of the demand distribution, allows ToolsGroup to address demand volatility better than anyone else.

If you have seen the movie or read the bestseller book “Moneyball,” you can appreciate how we can compare ourselves to the nerdy assistant manager who comes in with the computer models and shows the team how they can create a better roster less expensively by using a highly statistical approach. At ToolsGroup, we statistically model demand volatility across multiple dimensions such as line-order quantities, order-line frequency, and forecast error to understand it better. So I guess you could say we are the nerds when it comes to understanding demand volatility and its impact on demand forecasting, demand sensing, and inventory optimization. 

TEC: How would you describe the importance and differentiation of your Machine Learning Engine (MLE) in plain English?
JB: Machine learning is a cool new technology for solving some vexing problems.

Let me provide one good example in the area of forecasting trade promotions. Advertising and promotions are the main activities affecting the baseline demand and demand variability of most consumer packaged goods (CPG) companies. CPG companies spend up to 12 to 15 percent of their gross revenue on trade promotions and related trade funds for retailers.

In a problem such as modeling trade promotions, determining their actual impact or “lift” is daunting. A large number of variables with complex interactions are buried in huge amounts of data with a high degree of noise. Even with substantial expertise, it’s usually not possible to understand the correlations among the variables, especially at the more detailed operational level.

MLE extracts knowledge about which variables most affect demand, and produces a set of simple intelligible rules, easily understood by the user. It makes it possible to recognize the shared characteristics of promotional events and identify their effect or “lift” on normal product sales. The tool identifies which variables most affect demand, and produces a set of simple intelligible rules. Moreover, it identifies each rule’s relevance and importance, its range, and its cutoffs. In other words, machine learning is just a great technique for solving a really tough problem.

In the case of forecasting trade promotions, the benefits accrue to a wide range of departments of an enterprise:

  • Marketing—Enables model promotions and media events to support at a detailed level customer plan definition and key account team budgets.
  • Key Account Team—Allows campaign operational planning in which promotions and media events are allocated to particular accounts (mid- to short-term planning).
  • Supply Chain—Able to satisfy promotions and media uplifts with timely production and stock deployment to achieve the target service levels.
  • Finance—Enables a robust and precise analysis of campaign profitability.

TEC: I recently attended an event where Bob Ferrari of Supply Chain Matters gave a great presentation on the SCM trends, i.e., moving from historic to more predictive analytics in near real-time, cloud & composite apps, sense & respond, social & mobility, etc. (see related blog post). How is ToolsGroup stacking up along those lines and themes and with what capabilities?
JB: Some of those trends, such as mobility, are a bit outside the scope of our applications. We don’t offer a handheld supply chain optimizer yet, but maybe someday. The one I will focus on here is sense & respond, because we offer a unique contribution, taking a very different approach than any other SCP vendor that we know of. We promote a highly automated approach to SCP that we call “Powerfully Simple.”

“Powerfully Simple” is based on the concept that systems are more effective if they are smarter, with manual effort kept to a minimum and planners able to take a low-touch approach. Our software is designed from the ground up to be used less as decision support tool and more as an intelligent, automated solution. It employs powerful technologies, but keeps the smarts (“rocket science”) behind the screen, transparent to the user. It makes the planning easier, because the system is able to do more work on its own.

This approach reduces tedious adjustments and manual overrides, allowing planners to focus on higher value-added activities. As one of our customers articulates extremely well “Let your people do business, not statistics.” To hear one of our customers quickly summarize their experience click here, here, and here for video testimonies.

The goal is to have a reliable and stable (self-adaptive, even “black box”) SCP system that would automatically run on its own without continuous manual intervention. It would generate more sustainable results and reduce overhead and ownership costs. And the supply chain would become calmer as the bullwhip effect is tamed.

We believe smarter and more automated with a high-frequency response is the next generation in SCP. Predictive key performance indictors (KPIs) draw planner's attention to the most pressing supply chain problems. Along similar lines are intelligent "probabilistic" exceptions that are "pushed" to stakeholders.

A demand sensing engine is the “man behind the curtain” of sorts. It imports fresh daily demand data (from different streams of data) and senses changes in the demand signal by running comparisons to expected patterns. It then analyzes the statistical significance of the change using consumption logic at the SKU level, and redefines residual demand forecast accordingly.

TEC: Which products, regions, and verticals have been most active of late? What do you foresee in 12–18 months—more of the same or not?
JB: Northern Europe grew rapidly for us in the last 3 years, about doubling in size. We see the US strengthening significantly in 2012. All of the verticals remain very stable.

TEC: What is your message to companies where finished goods multi-echelon inventory optimization (MEIO) is not the top pain? In other words, how do you compete with broader SCM and/or ERP suite providers (that offer strategic network optimization, sub-assembly operations, etc.)? How do you plan to grow in an IO market that is slowly disappearing as an independent niche (while SCP/forecasting is also highly contested)?
JB: That's a good question. By bringing unique statistical skills to SCP, we offer capabilities that appeal to many types of companies. Here are a few examples:

  1. A statistically-based solution is ideal for addressing supply chains with a mix of intermittent or “lumpy demand” items alongside fast movers. As we discussed, intermittent demand is very typical not only in aftermarket parts and slow movers, but also in markets such as consumer goods, where replenishment is at a much more granular level, by SKU, day or week (instead of month), and customer channel. At that granular level, fast movers start to look lumpy and medium movers become intermittent. So we see a lot of opportunity in those verticals that have traditionally been strong for us.
  2. Vertically-integrated or multichannel distribution is becoming more common and e-commerce providers represent a good opportunity for us, because we can reliably model real-world demand behaviors for multiple SKU/channel combinations.
  3. Our technology is well suited to supply chains with many NPIs and seasonal changeovers, so for instance, fashion and apparel have been good verticals for us. 
  4. Finally, as a statistical approach to understanding demand volatility is really critical for achieving very high service levels (e.g., over 99%), we do very well in this environment. Food and beverage and pharmaceutical are good examples of two verticals that demand such high service levels.

TEC: How are your rapid inventory rightsizing (RIR, see related blog post), long tails, and sales inventory optimization planning (SIOP) initiatives doing and jiving with the latest strategy?
JB: RIR was a product offering specifically designed for the recession. We have incorporated many of its capabilities into our consulting solutions, but we don’t actively promote it as a product anymore.

As we discussed, successfully managing “long-tail” inventory is still one of our strong suits. If you checked out the RS Components video mentioned earlier, you saw the success one of our customers confronted with this problem. SIOP is an important, ongoing part of our SCP business. 

TEC: How far is your Demand Collaboration Hub (DCH) offering from a full-fledged sales and operations planning (S&OP) system that includes financial folks, product development folks (NPIs), etc.?
JB: DCH addresses the demand collaboration step in the overall S&OP process. It helps organizations transform the traditional S&OP process to “SIOP” (as most traditional processes do not really attempt to optimize inventories). 

DCH currently fully supports collaborative demand planning across different stakeholders, enabling organizations to share and update demand plans among sales, marketing, finance, product development, as well as with suppliers and customers. Sales teams can collaborate on account level activity and also share potential new deal information, which can significantly improve demand planning. Product development and marketing can forecast new product launches and promotions, and more. DCH is a key part of our solution for many of our customers implementing an S&OP process.

Further Reading

A Modern Tale of Long (Supply Chain) Tails — Part III. September 12, 2008.
How to Plan and Manage in Uncertainty and Volatility? Based on Reality and Facts, Duh! – Part 2. November 29, 2011.
IBM & ILOG Matrimony: Good for BPM, Uncertain for SCM? — Part 3. October 10, 2008.
Optiant Going to a (Much) Better Place: Logility – Part 1. April 15, 2010.

comments powered by Disqus