How HP Visualizes its Supply Chain using Geographic Analytics
How can you make strategic supply chain decisions faster and more effectively? For HP, one answer lies in a technique called Geographic Analytics
High-performing companies will embed analytics directly into decision and operational processes, and take advantage of machine-learning and other technologies to generate insights in the millions per second rather than an “insight a week or…
- Big Data and Analytics Focus in the Travel and Transportation…
- Embrace Convergence
- Performance Testing for Mobile Applications
- Defend your Business Against the Dark Art of Social Engineering
- All Resources
HP has served freight and logistic companies for more than 45 years—delivering innovative solutions with greater effectiveness and efficiency, at lower risk. We help support critical transportation operations through an event-driven decision support process and better information management.…
- Company Quicklook
When HP launched its biggest supply chain transformation in history in 2010, it was evident that the challenging multi-billion year-to-year savings targets would be difficult to reach in the traditional way.
This transformation required new approaches to deliver substantial improvements across businesses, targeting cost reductions as well as structural streamlining in the form of partner, network and site consolidations.
To meet these challenges, HP had to enhance its analysis toolbox to deliver supply chain projects faster and with better alignment across businesses and regions. (For more on the transformation initiative, see “HP’s Supply Chain Transformation Challenge.”)
One of the successful approaches developed to enhance our supply chain analytics capabilities is what we called “Geographic Analytics” (GA), a visualization technique that we describe more fully below. GA works. It reduced the time required for network optimization projects by up to 50 percent. In addition, projects driven by this approach were often better supported by the business groups.
The reason: Executives became involved much earlier than they would have in traditional, purely data-driven supply chain analysis work.
Although designed to support supply chain network optimization, Geographic Analytics has applications far beyond its original purpose. We received major interest in this technique from groups both inside and outside of HP, including from sales and after-sales organizations, enterprise risk management functions, and from supplier relationship managers. Each of these groups has a regular need for visualizing and analyzing location data for a variety of purposes ranging from studying after-sales networks to risk management of the supplier base.
What Is Geographic Analytics?
Let us now take a closer look at what Geographic Analytics is. In simple terms, GA is the visualization of network information on a map in order to drive supply chain optimization.
To get started, you map the relevant locations of the network, such as the distribution centers that you want to consolidate. Next, you add basic background information such as hosting business group, square footage, and volume information applicable to each location. Lastly, you apply a smart directory structure that allows you to categorize the locations and quickly filter them to give selective views of the map as needed.
In any analysis project, the data collection, cleaning, and interpretation are very time-consuming. The key is to avoid analysis paralysis. Instead of feeding comprehensive data sets into complex analysis tools, GA first aims to visualize only the most basic information—with the goal to provide the shortest possible “time-to-insight” to all stakeholders involved in the project’s decision-making process.
Many supply chain assessments have “intangible” framing conditions that are disruptive to any purely data-driven optimization solution. It is very important to capture such conditions as soon as possible. If they are overlooked or detected late, they can lead to significant rework and frustration for everyone involved.
For example, there might be regulatory requirements or tax conditions that overrule the data-driven optimized solution. Furthermore, there are often infrastructure constraints not reflected in traditional optimization software packages, such as local traffic congestions that make certain areas of a city unfavorable for heavy-traffic warehousing.
Displaying basic information on a map to involved stakeholders and experts, such as analysts and subject matter specialists, helps to identify and capture such conditions upfront. This method uses their intuition to determine the project’s course, as we describe below.
When the key information is visualized on a map, approaches to a supply chain problem evolve almost automatically:
- Densities of locations highlight major demand areas or inefficiencies in the structure—for example, too many locations in a small area.
- Additional data visualization techniques, such as simple traffic light indicators (for example, sites with high levels of inventory colored in red, sites with low inventory levels in green), guide both analysts and stakeholders to key areas of focus.
- Publicly available infrastructure information shows access to highways, airports, seaports, and railways.
Surprisingly little data is actually needed to determine the further course of the analysis. Often, hypotheses can be agreed on and unfavorable solution options can be dismissed before a more detailed analysis takes place.
The simplified approach is the starting point for the deep-dive into the remaining solution options. This usually requires a significantly smaller data set than what would have been needed without GA. An intended side effect is that at the end of the initial analysis, stakeholders and experts supporting the project will already be reasonably well aligned. Instead of being confronted with a solution coming out of a “black box” at the end of an analysis, all sides are involved from the start in the decision-making process.
Overall, GA speeds up traditional data-driven approaches because much less data is required. The reason is that relatively few pieces of information—once displayed on a map—are sufficient to obtain most of the background information from the involved parties. We call this “harvesting the intuition behind the problem”, a step that can significantly simplify data-driven analysis. This approach ensures that the analysis is steered in the right direction from the start as depicted in Exhibit 1.
How GA Contributes to Business Decisions
We have assembled a collection of examples where Geographic Analytics can greatly benefit supply chain and business decision making. These examples relate to network optimization, network flow optimization, risk management, and after-sales service. For each, we describe the analytical focus and outline which parameters need to be defined to get substantial benefits.
One of the supply chain manager’s most important strategic tasks is to streamline the supply chain network. In many global organizations, cross-docks, warehouses, and distribution centers have sprung up like mushrooms over the years. The result is often a jumble that needs to be cleaned up to achieve a lean, top-notch supply chain.
The traditional approach to addressing this problem has been to conduct what is known as a “center of gravity analysis”. For such an analysis, an expert—often an external consultant—is called in. This individual (or individuals) collects data, loads current sites, flows, inventories, transportation costs, and other data into a complex software tool that almost magically determines the optimal locations for your network.
The center-of-gravity-analysis approach has its drawbacks, however. The data-gathering and data-cleaning effort is time consuming and can be costly. Moreover, it is often difficult to put the recommendations produced by the software tool into action. And if you subsequently find that some of the implicit framing conditions do not apply to your situation, you are faced with the possibility of redoing the whole analysis.
This is where GA can expedite and streamline the process. Exhibit 2 shows a sample network of HP services and distribution sites in Singapore before we conducted a comprehensive consolidation analysis. Data was scarcely available; in fact, at the beginning of the project only the current locations and their space requirements were known.
Yet even with this minimal data set, GA immediately highlighted a large part of the solution:
- HP had four areas of presence: North, East, Central, and West. Stakeholders as well as experts agreed that the optimal solution would fall into one of these areas.
- The main cost driver within Singapore was rental cost rather than labor or transportation costs. A closer look revealed that those costs were significantly higher in the East and Central areas than in the West and North.
- For HP’s major export and import activities, the West was favorable because of its proximity to the harbor.
From these three points, we could outline the overall strategy: Consolidate as much as possible in the West. Instead of undertaking a general and open ended site-consolidation-at-optimized location analysis, we broke the problem down into several small problems that were much easier to communicate, understand, and solve. Our approach was to analyze what prevented each site to be moved to the West.
One could argue that this site consolidation example is a relatively simple one. However, the GA techniques used in this example apply similarly to much larger network optimization problems. If the GA approach does not solve the problem completely, it can effectively reduce the problem down to a handful of strategic questions that require significantly less data and effort than traditional approaches.
Network Flow Optimization
We can take the network optimization a step further by adding product flows between sites. From an analytical point of view, the network flow optimization problem is even more challenging than pure location consolidations.
Transportation flows typically have physical constraints that prevent consolidation, namely the incompatibility of transport means—for example pallets vs. slip-sheets, less-than-truckload vs. full-truckload shipments, containers vs. trucks, and so on. Therefore, it makes even more sense to visualize transportation flows prior to a deep-dive analysis.
Exhibit 3, which is illustrative only, depicts transportation flow visualization, in this case for EMEA (Europe, Middle East, and Africa). The sample flows depicted show different products in different colors. The arrow thickness corresponds to yearly shipping volumes.
Analysis of the display of the flows quickly points to potential optimization areas:
- Major flows go from Eastern Europe into several Western European countries. If these flows move mainly via LTL, there might be an opportunity for freight consolidation in Eastern Europe. This could involve full truckload shipments to the West and subsequent cross-docking or advanced “milk-run” routing to the final destinations.
- There are flows for product A from Central into Eastern Europe, and for product B from Eastern to Central and Western Europe. This points to possible backhaul trucking opportunities for the two products, noting that backhaul rates are usually significantly cheaper than one-way rates.
- Truck volumes to the North and South display little consolidation opportunities, so they can be treated as a secondary priority.
Risk management and contingency planning have become major considerations for supply chain managers in today’s ever leaner—and thus ever more vulnerable—networks. The Japan earthquake and Thailand flooding in 2011 served as reminders of these new focus areas.
Some of the most pressing related questions are:
- For natural disasters, which areas in the network are most vulnerable? Or, if disaster has already struck, how is the business affected?
- Where should supply sites be positioned to mitigate risks from natural disasters for the company?
These questions can be addressed by mapping the supply sites of your network as conceptually shown in Exhibit 4. Suppliers are mapped together with an assessment of the risk they pose to the supply network. Red signifies a high risk of failure from this supplier, while green shows a low risk. The assessment could be based on your company’s volumes with this particular supplier, or on the number of products affected should this supplier fail to perform.
Through this mapping exercise, areas of high risk become apparent—prior to any more data-intensive analysis. In a subsequent step, in-depth risk analyses can focus on these areas.
Service level agreements (SLA) concerning spare parts are a crucial point of any contractual agreement between service provider and customer.
Customers will pay a premium for the service provider’s commitment to deliver and exchange parts within a specified timeframe. Therefore, from a service provider’s perspective, one of the supply chain manager’s most important jobs is to match customer service level requirements with supplier capabilities. Over-commitments, such as a promised turnaround time that cannot be met, lead to customer churn and penalties. On the other hand, under-committing means that a higher service commitment at the same cost could have been provided to the customer.
Several questions are relevant here:
- For a new customer, is the current network capable of fulfilling the service levels that this customer demands? And what is a realistic reaction time?
- Are there “hot spots” in the existing network where SLAs are often missed? Should the network be upgraded?
Geographic Analytics can be of great help in answering these kinds of questions, as illustrated in Exhibit 5. The left picture displays a “Frontier map” with a traffic light system characterizing historical turnaround times. The mapping provides a good overview of the service levels that can be realistically offered in a certain area.
The right map, which we call a “SLA Hotspots Map”, identifies areas where SLAs are frequently missed. For example, these could be a sign of infrastructure bottlenecks from chronically congested access roads. This type of mapping also helps to identify areas for additional service center locations.
A Five-step Approach to GA
In our work at HP, we have identified five steps to get a Geographic Analytics project up and running successfully:
1. Create a Site Database
The first step is to create a database with the locations you want to map. This will be the basis for all GA. Databases can be created on an ad hoc basis for the problem at hand, or more systematically as a permanent repository that can be used for multiple projects.
The real power of GA gets unleashed with a permanent site database because of these advantages:
- Ready-to-use data allows quick strategic assessments.
- Existing databases permit easy sharing with other interested groups, which is the key for widespread use.
- Permanent databases come at a comparatively low cost. The main work lies in setup of the database, whereas ongoing maintenance requires very little effort.
2. Decide How to Represent Sites
Data representation will determine to a large extent how intuitively the data can be used. In this phase, quality trumps quantity.
The icons denoting your locations play a vivid role; therefore, the time and effort spent to set them up smartly is well invested. Icon shapes, acronyms, colors and sizes should be intuitive. Strike a balance between simplifying and conveying sufficient information. Our recommendation is to limit the icon representation to two or three levels or dimensions to keep a narrow focus.
A proven approach is to use icon shapes for denoting the site type (for example, to distinguish manufacturing and distribution sites), acronyms for the different business groups, and different colors to represent a “traffic light” system.
A good traffic light system is intuitive and it establishes the key measure of the assessment. Sites with high inventory, high failure rates, or high costs that require further attention would be identified in red; those with only moderate issues in yellow; and the best-performing sites in green.
Organize your data with a filtering structure that lets you dynamically choose which groups of sites to display.
This allows you to adapt the mapping addressing different areas of analysis one at a time. Typical examples are filters for business groups, location types, or the traffic light colors.
3. Select the Software
There are several tools, both commercial and free, that translate site databases onto geographic maps; an internet search for “mapping software” will lead to thousands of hits. In order to provide flexible support for HP’s business groups, the authors developed an HP-own mapping solution.
Whatever tool you use, make sure it is able to:
- Translate addresses into geo-coordinates if the latitude and longitude are not already included in your site database.
- Provide infrastructure and background information, such as city names, country borders, highways, railways, seaports, and airports.
- Customize the network representation through directory and filtering structures, different symbols and colors, and displays of additional relevant site information.
4. Align with the Stakeholders and Conduct an Initial Assessment as Early as Possible
As soon as you have the information mapped, show it to your involved stakeholders and experts supporting the project. Do not worry if the site data is not yet final. In our experience the first discussions will foster a common understanding of the current network and will channel the attention to the solution—independent of 100 percent data accuracy at that point. In addition, an early alignment helps to get the site data into a clean state.
Set clear and realistic targets for the first stakeholder meetings. First, confirm the network to be analyzed. Second, determine the direction of the analysis by questioning the current network setup, identifying hot spots, and defining potential solutions and hypotheses. Third, poll the stakeholders on their decision criteria for the overall solution.
From this foundation, derive those scenarios that will be evaluated in the subsequent analysis, while ruling out others because of their unfavorable physical, operational, regulatory, or other conditions. So, for example, a need for a seaport access would rule out inland locations, or import/export conditions or other trade restrictions would make certain countries unattractive.
5. Derive the Recommendation
GA will not always be sufficient to pinpoint the final solution, but it will significantly reduce the options to a few core scenarios that can then be analyzed with less effort. The optimization of HP’s Brazil services and distribution network is a good example.
GA did not give the complete solution, but it revealed the following simplified results that set the direction for it:
- Brazil was divided into five major regions.
- The actual network of eight locations was focused around the three fixed production sites in three sub-regions. For each of the production locations, an associated warehouse and distribution center was needed.
- Hence, the target was to consolidate to three main locations plus potentially some additional satellite cross-docks.
- Finally, the decision-driving factors for additional satellite cross-docks were established. The key factors were demand/volume in the satellite areas, transportation costs, and regional tax levels.
Additional data analyses ultimately led to the complete solution.
Key Lessons Learned
Our projects involving GA yielded some valuable lessons that we would like to share with other supply chain professionals.
Do not underestimate the lead time required to create a clean site database. Creating a site database requires considerable time. Aligning on the data fields is the easier part; verifying all sites, addresses and geo-coordinates is a far bigger challenge.
You will be surprised how much you did not know about your sites. For example, how many sites exactly are in your network? This may sound like a strange question, but our experience—at HP and at other organizations—shows that sites can be overlooked. In one of our projects, a zoom-in on street level revealed the company logo on a building that was in question. With a geographic display of sites, disputes over the existence of a site can be settled in no time.
Typically it takes 6 months to 12 months from the start of data collection until the data is reasonably clean and complete. Maintenance to keep the site database up to date is very low. A quarterly update cycle usually suffices because most businesses change locations infrequently.
Be lean on data collection and display. The key to GA is speed and simplicity. It is a common mistake to request large amounts of data at the beginning of an analysis for fear of missing a crucial data point. In our experience, less is more when crafting a timely and usable recommendation. Always remember it takes time to collect and process each additional data point, and this keeps you from aligning the already available material.
Simplicity in the data display also is key. A simple traffic light system helps people grasp the state of a network at a glance and will focus further discussions. Intuitive icons provide additional insights. Conversely, fine-tuned graphics overloaded with complex symbols will distract rather than enlighten, making it more difficult to achieve quick success.
Leverage publicly available mapping information. Many geographic mapping tools allow the display of additional information (“overlays”) that can provide valuable background information for decision making. Use them!
Commonly available overlays include:
- Road information such as expressways, access roads, and street names.
- Infrastructure hub information—for example, harbors, airports, and railway stations.
- Distance information such as linear and street distances.
- Environmental information on weather conditions, earthquake zones, hurricane routes, and so forth.
Start with a prototype. When people use GA, they love its simplicity, transparency, and power. When a project is viewed in retrospect, the advantages of GA generally become indisputable. However, it may prove difficult to get the first buy-in for the investment to set up the site database and the mapping tools. To overcome this hurdle, we recommend using a prototype to display what is achievable through GA. With such a prototype, it becomes much easier to secure support for a site-database and a mapping tool.
A Tactical and Strategic Asset
Flexibility, transparency, and speed make GA a great instrument to support strategic and tactical supply chain decision making. Geographically visualized data is easy to understand, thereby facilitating faster hypothesis derivation and decision making. It reduces the number of solution options by “harvesting the intuition behind the problem”—that is, by capturing the factual expertise of the parties involved, before heavy data-driven analytical tooling enters the scene.
While GA is an excellent method to support geographic optimization, it is meant to complement—not to replace—traditional data-driven analytics. GA substantially narrows the scope and data requirements for supply chain problems while carving the path for further data examination. The project direction can be established quickly and dead-ends can be ruled out early.
Our experience at HP confirms that Geographic Analytics supports the collaboration and alignment between operational and strategic managers. In short, it is a valuable asset in modern supply chain management.