Integration between an ERP and a planning & scheduling application like frePPLe remains a magic subject surrounded by myths, mystery, and smoke.

This is quite unjustified — the integration is far from complicated if you approach it correctly.

Here are our 5 best practices learned the hard way over many integrations we have worked on over the years.

  1. Technically, it’s straightforward and trivial.
  2. Your ERP data extract needs to be a complete and consistent snapshot.
  3. There is rarely a need to plan more frequently than daily.
  4. Separate execution and planning horizons.
  5. Stick to the planning capabilities of an APS software.


1) Technically, it’s straightforward and trivial

It really is. If you are facing technical issues downloading or uploading the data, something is very wrong.

Most of your efforts should go in correctly understanding the context and meaning of your data in both systems and making the correct mapping from a business point of view.

We have created a template for an ERP integration in which we extract the input for each frePPLe object with a query: see the documentation or the source code.

You can upload the resulting CSV-formatted files in the cloud edition of frePPLe or, if you run frePPLe locally, choose to directly load the data in your database. It can be easily extended to different technologies or communication protocols.


2) Your ERP data extract needs to be a complete and consistent snapshot

A planning application needs an accurate and complete vision of your current status.

There are two kinds of data elements to be interfaced:

  • Static information which doesn’t vary too often. Elements in this class are items, customers, resources, suppliers, bills of materials and routings. Some of these tables are rarely updated (for example the location table) and it might not be worth writing interfaces for them. Manually populating them in frePPLe is enough.
  • Dynamic information which changes from day to day. Elements in this class are inventories, open sales orders, open purchase orders, open manufacturing orders (work in progress), etc.

For the static data, you need to create a mapping between the ERP data model and frePPLe.

We typically start from identifying the data that needs to be populated in frePPLe from the frePPLe data requirements and then work out the mapping with the ERP data.

For dynamic data, the data extract should be as complete and consistent as possible. It represents a snapshot of your supply chain status. Think of it as a picture taken at a particular moment of time.

“Complete” means we need all relevant demand and supply elements. “Consistent” means that all the data elements have to be in sync, without duplications and black-holes.

The examples below describe some situations of duplications and black holes from real-life cases.


Example 1: Visibility of material during reception

Visibility of material during reception

A purchase order was placed with your supplier, and you receive the ordered material in your warehouse. The material may spend some time in quality control process before it is booked in inventory.

To obtain a consistent picture in the data extraction we need to see a supply of 100 pieces of item A at every moment we take the picture.

At moment 1 we see an open purchase of 100 pieces of item A. At moment 3 we see that the inventory of item A has been increased with 100 units.

What do we see at moment 2?

If we see both the purchase order and the in-quality-control stock, we are double-counting the supply. The planning system will incorrectly reduce future purchases for the item as a result.

If we see neither the purchase order or the stock, we have a black hole and miss some supply. The planning system will incorrectly react to this by generating a (potentially urgent) new purchase order.

A consistent snapshot requires that we see only one of these. A bit of thought and a correct filter for the data extraction is all it takes to do it right.


Example 2: Visibility of in transit material
Example 2

Imagine we ship material from our warehouse A to our warehouse B. If these warehouses are in different parts of the organisation for accounting reason, it is not uncommon to register a sales order to a dummy internal customer in warehouse A, plus a purchase order from a dummy internal supplier in warehouse B.

The purchase order and sales order may be created at different moments by different people.
They can also be closed or change status at different moments by different people.
And how do we see material that is in transit between both warehouses?

The simplest solution to assure a clear and consistent snapshot of the current status is to consider either one of the transactions as the leading one and filter out the other.


Example 3: Partial data refresh is not a good idea

Refreshing only a part of the dynamic data is never a good idea. All of the data elements are linked with each other, and we need a consistent and correct picture across all data elements.

For instance, refreshing inventory data without refreshing the list of open purchase orders isn’t correct. If you close a purchase order, it affects the inventory on hand. If we see the new inventory plus the purchase order that was still open yesterday, we get an incorrect snapshot.

Similarly, refreshing inventory data without the list of open sales orders isn’t correct.


3) There is rarely a need to plan more frequently than daily

It is a common misconception that you need to replan very frequently to adjust for every event that changes the planning input data.

Even for short term planning and scheduling, a daily replanning process is sufficient for the majority of businesses. In reality, today’s plan is already communicated to the shop floor and is being executed.

Interfering with actions that are in progress or about to be started is not easy. Continuously revising the plan just leads to unnecessary nervousness and confusion.

It is a better practice to hand out the execution plan for the day and leave the shop floor some room to self-organize the details and self-adjust for small variances and minor unforeseen events. Situations where the shop floor needs to contact the planner to know what to do next will occur, but they should be exceptions rather the rule.


4) Separate execution and planning horizons

A planning tool provides visibility of your capacity and material requirements for a time horizon that can be beyond what is relevant for the ERP. You need to feed the ERP with the plan outputs. It needs to start executing pretty soon.

In most cases, the remainder of the plan doesn’t need to be sent to the ERP at all. It can stay in the planning tool, and as time goes on, it will eventually roll into the horizon that is relevant to the ERP.

After all, an ERP system is very transaction-focused and registering all going-in activities. The role of a planning system is to provide clear visibility and analysis transparency on future plans beyond transactional details.


5) Stick to the planning capabilities of an APS software

A planning tool is not the right place to implement all your reporting, tracking, and monitoring needs. One should avoid abusing the APS database as a generic data repository.

Example: reporting on historical data

It is meaningful to use an APS to display the average purchase orders value per location in the upcoming months. But you wouldn’t use an APS to display the average stock on hand in the past months.

Indeed, the former is nothing else than the computed plan visualized in another metric. The later report would require that you stock a huge amount of data to be able to build the desired report, overloading the database with data that should rather be used in an ERP or BI tool, thus extending the workflow duration.

Johan De Taeye

Johan De Taeye

Johan is an engineer at heart. He worked 10 years for i2 (now jda), then another 10 years optimizing supply chain management for his clients. He founded frePPLe in 2011.