Vendor Managed Inventory – KPIs and Metrics

In the last two article about VMI, we took a look at VMI in general and some points regarding the successful implementation of VMI. Now, it’s time to have a more detailed look at one the things that do fundamentally change with VMI – KPIs and metrics.

We all measure our own and our suppliers On-Time delivery performance. This ranges from rather simple approaches comparing planned and actual delivery dates, to sophisticated ones taking everything, from changed delivery dates and quantities, over goods receipt inspections and actual unloading dates, to quantity and delivery date tolerances into account. This topic alone is worth an article, or multiple, of its own. For now, so, the important thing to understand for VMI is, that none of that matters anymore. Well, at least not in the traditional way.

The reasoning behind On-Time-In-Full / OTIF

No matter how sophisticated, or even over-engineered, your KPI system measuring your suppliers is (and yes, over-engineering is a thing in that case as well), there is an underlying reasoning behind it. Driven by traditional MRP-logic, however you approach OTIF, dates and quantities are based on your Material Requirements Plan. In the first step calculated automatically, it is the planners that fine tune these calculations. They make sure that goods are always available when needed and planned, and no excess inventory is built up. Whether it is done systematically, in the form of statistically calculated buffer and safety stocks, or manually / intuitively by the responsible planners, supplier performance factors into that calculation (safety stock is, again, a topic for itself, so).

As a result of this approach, your planning is reliable as long demand isn’t fluctuating too much, and for our topic at hand, suppliers deliver on-time, in-full and without any quality issues. This gives suppliers very limited freedom and flexibility.

The basic assumptions, and this important to note when considering VMI, are:

  • Your planning is always correct
  • All your suppliers have to do is adhere to the plan
  • Any deviation from that plan causes a problem down-stream

Regardless of how many different factors you consider when calculating OTIF, these three assumptions, either consciously or not, are always there. The question is, are they truly relevant? How often did a missed late delivery of, say, two days, cause any trouble? The first instinct would be to say “all the time!”, right? But do you really know? Or is it just because we have a tendency to remember the cases where it was a problem, while all the cases where it wasn’t show up in the regular OTIF and supplier reviews as “underperforming” suppliers? Because, without another KPI that monitors bad availability and a KPI system that links these two, there is no way to actually tell.

Supplier Performance und VMI

It doesn’t matter who owns the inventory, you or your supplier, when talking about performance measuring under VMI. The crucial difference between VMI and other procurement approaches is, that your supplier is responsible for availability of stock. And in order to do so, he needs a forecast from you. Guess what the two VMI KPIs should be?

Forecast Accuracy

Forecast accuracy is measured the same way for VMI, than it is for other purposes in your organization. The different ways how this can be done is, maybe, again something for different post. For VMI, let’s just have a look some principles regarding forecasting:

  • Forecasts are always wrong
  • Aggregated forecasts are better
  • Short term forecasts are more accurate than long term forecasts

The first point results, for VMI and in general, in a certain amount of buffer inventories. The huge benefit of VMI is, that, contrary to traditional procurement where both, suppliers and customers keep buffers, only one buffer stock is needed. The last point is best solved with rolling forecasts on a weekly basis covering, say, the next 16 weeks. Long true production lead-times can change the forecast horizon and frequency. True production lead-times are the relevant parameter for VMI, procurement lead-times are, to a large are part, influenced by the limited flexibility the supplier has to schedule his production runs. I remember one case in which production of a economically feasible batch was 3 weeks, pre-VMI procurement lead-time amounted to almost 5 months.

The second point is an issues, so. Usually, aggregating demand happens by aggregating the demand for customers in a certain region for a certain product family. VMI is implemented for single products / SKUs. Aggregation has to happen on a different level, where exactly depends on the individual case. Aggregating demand for spare parts and non-spare part demand, by region and for periods are all feasible solutions.

How do you measure your accuracy? I would suggest at the highest possible level of aggregation and the shortest feasible periods. That could be the complete demand for part A, spares and internal demand, for all customers in Europe on a weekly basis. This forecast is the relevant one for your supplier, it is reasonably aggregated and allows fast reactions and mitigation in case anything goes wrong. The calculation is simple enough:

(Actual Demand – Forecasted Demand)/(Forecasted Demand)

Positive deviations indicate a heavy demand exceeding the forecast, negative deviations the opposite. For VMI purposes, this is more then enough. Additional deep-dives are then conducted when need, e.g. to identify the root cause for a certain demand spike.

Using a rolling forecast raises the question which forecast to use. In our 16 weeks horizon, every period has 16 forecasts. The closer you get to the period in question, the more accurate the forecast gets. Personally, I would aim for the two or three week forecast, depending on the production lead-time. The one week forecast is usually to short term for your supplier to react in any meaningful way.

Supplier Performance / OTIF

The fundamental change happens measuring on-time delivery. All other performance parameters, quality, pricing and so on, are not really impacted by VMI and should be measured as before.

OTIF on the other hand, simply doesn’t exist anymore. Why? Because the is no time or quantity anymore to measure against. Sounds scary, right? So how do you measure your suppliers performance then? And what do we measure OTIF for in the first place? The answer is availability.

Under VMI, a supplier is responsible to assure availability based on the forecast. Suppliers want to be protected against excess inventory, and customers against shortages. This is achieved by defining minimum and maximum inventory levels. Using the forecast, and the on-hand inventory as well as scheduled deliveries, gives us an inventory projection. Now, it is the suppliers job, to keep inventory levels between minimum and maximum levels over the full horizon. And this, looking at the past instead of the future, is also the KPI replacing OTIF under any VMI agreement. Simply divide the number of days in a period, ideally one month, in which a supplier was above the maximum agreed inventory, or below the agreed minimum level, by the number of days in that period. For added transparency, I suggest to measure days below minimum and above maximum separately. Days above maximum hurt your warehousing capacity and cost, days below carry the immediate risk of shortages. And it is better to have that extra level of transparency.

VMI Performance = (Days per period within agreed inventory levels)/(Days per period)

VMI Performance Shortage Risk = (Days under minimum level per period)/(Days per period)

VMI Performance Excess Inventory = (Days above maximum level per period)/(Days per period)

It is important to note, that the granularity is defined by the days in a given period. In the case of months roughly 3.3%. This means, that we need an additional KPI, to properly appraise our supplier.

What is the worst thing to happen? Not just a potential shortage, but a true shortage. So the final VMI KPI we use, is the number of actual shortages happening during a given period. In this case, we want better granularity than 3%. So we use number of orders in a given period as a basis instead of days. And again, we have to split that KPI into two different metrics:

  • Actual Shortages on days with potential shortages
  • Actual Shortages during the full period

The first one is a sub-set of the second one. Why are we splitting that KPI? Because the root-cause, for actual shortages on days with enough on-hand inventory, is most likely not your supplier, but something different. But if it is not the supplier, why are we looking at it at all? Because as customers, our planners’ job is to make sure material is available, and not to simply monitor suppliers. So we use both metrics, they are closely related, the same person has to act upon them and the database is the same as well.

The KPIs are calculated as follows:

Shortages on days with potential shortages = (Number of shortage orders on days with potential shortages per period)/(Number of orders per period)

Shortages during period = (Number of shortage orders per period)/(Number of orders per period)

Both metrics increase granularity, and help us to identify any availability issue in a VMI agreement. Together with forecast accuracy, they also increase the context of your KPI system by an order of magnitude.


I am a fan of using the regular metrics as the basis for any contractual penalties or bonuses. Using a different set of parameters for that, just makes it a lot harder for everyone to achieve great performance. In our VMI case, we use the KPIs we previously discussed, as contractually relevant KPIs. Doing so means nobody is surprised when any performance issues result in bonus or malus claims during a review. How a VMI agreement can be structured and which details have to be included and covered will covered in another post.

Leave a Reply