Finding the Cure

There are good reasons to carry more inventory than you need. You might want to stockpile parts or materials expected to get more expensive, or keep rare parts on hand to improve customer service. Generally speaking, however, inventory is a physical representation of the battle for your share in the value chain. The less you can carry, the better your own position.

The problem with a general rule like “carry less inventory” is that it’s almost useless as a foundation for improvement. Open-ended guidelines also tend to become problematic. “Carry less inventory” isn’t a great mission when you run into production delays.


A popular tactic is to narrow definitions to make the guidelines more focused – “target reduction of low-velocity inventory,” or “lower contracted inventory commitments with the top 20 vendors” – but it seems that in a data-driven age, we should have the tools to build more specific strategies. What specific steps can we take, and what can we expect to change?

A recent development in analytics, data unification, promises to make these specific, value-driven prescriptions possible. For example, one of the biggest challenges with inventory management is carrying duplicate or interchangeable part inventories, often because companies don’t know what they have. Putting that information in view makes it easy to say “eliminate these specific duplicate parts in five months to save $3,000,000 a year.”

With the data in hand, the specific goal and tactical path become obvious. If you didn’t know exactly how messy procurement data was, consider that these simple measures are out there to find, and the vast majority of companies have no plan to take them on.

Silence In, Silence Out

Most inventories are held at a site level, and the data is disparate, making it challenging to get a central view. Further, the same part will have different names and measures.

There is no easy way to know when you’re carrying ten units of a single part while you think you’re carrying five. There are too many questions and the knowledge to answer them is just as distributed as the data itself. Is “Poly 51mm pipe” the same as “PVC stand. 2×14 inch?” Which would you eliminate, or would you keep both just in case?

Analysts in most companies need to excavate and refine data for each study, which means they have to first identify the right question, pull in as much data as they can find around that question, and merge the data – all before any actual analysis can be done. Under ideal circumstances, this is just barely possible, and certainly not replicable. In fact, by the time the work of getting data together is done, the data is often outdated. It’s why companies lean on open-ended guidelines and struggle to put a value on them.

Par for the Course

At the highest level, data unification brings together information about data from various sources. It is particularly suited to the challenge of inventory management, which needs access to ERP systems, PDFs of contracts, inventory tracking systems, sales spread sheets, external data on commodity pricing and much more. With data unification in place, data regarding spend, vendors, materials and inventory can be treated as a single resource. Further, once data sets have been rationalized, any analysis is easily repeatable or even automated.

This means that an inventory analysis can begin with a general survey of inventory trends, from which trouble spots can be found and specific fixes identified. It’s less of a hunting expedition and more of a trip to the grocery store. You can decide what to do based on what’s there and valuable to do, rather than pinning hopes and overhead on a single idea.

Collect and Combine

The first step to unifying data is cataloging data sources. Most analysts have access to about ten percent of the company’s data.

A good catalog is more than just a ten-fold increase in data awareness. The effect of incrementally more data is geometrically better analysis. Free data catalog tools are available that help analysts discover, organize and understand inventory-related data. It is important to get the tools distributed throughout the company so data owners can easily plug their sources into the system.

From there, analysts can evaluate fields in each relevant source, to establish how data from each source relates to data in other sources. This data connection process is no small feat. It requires intelligence that is spread around the company, and a global analysis like this could require dozens or hundreds of sources.

Companies are relying on a combination of machine learning and expert sourcing to get this done. Machine learning tools can identify data sets that have strong correlation based on data similarities, past matches and other factors. Where machine learning and the analyst’s own judgment fall short, a trusted network of experts are called in to clear up issues. This, too, can be aided by machine learning, having software help identify the best expert for each question.

Once the data is clean and ready for analysis, a handful of tests can uncover specific, value-based recommendations for improving inventory management. For example, companies might look for changes in terms over time across all contracts to note trends in inventory requirements. They could evaluate the costs of shipping parts from a central warehouse versus keeping inventories of locally-sourced parts at each location. This is all in addition to more tactical analysis, which has likely never been done on the long tail of lower volume parts that represent a substantial portion of overall inventory. Just eliminating duplications or restructuring outlier contracts can result in substantial and easily identified savings.

Data unification is the quickest path to uncovering specific goals for better inventory management. It turns broad ideas into achievable campaigns that deliver known value. Instead of a blank canvas, you can color by numbers and get museum quality work every time.

Matt Holzapfel is a Product Marketer at Tamr where he focuses on procurement solutions.