Data integration should start with definitions. Acceleration's technology director walks you through the process.

Integration has been an industry buzz word for quite some time now and typically used as a trump card in conversation, guaranteed to lift someone' eyebrow. It hides the complexity surrounding what it actually involves. Saying; "yes, we can integrate to xyz …" is the tip of the iceberg and usually fails to define what is included in that implied integration.

It is sad to see companies intentionally staying away from the details just to win business initially and then later disappoint the client when tangible facts are exposed. Often this realization comes much later, when the client wants to achieve something based on that perceived integration.

Definition: the good, the bad and the ugly
To ensure that everyone is talking the same language, proper definitions of various points need to be agreed upon. Stating that you can integrate is simply the start.

For example, coffee seems to be a central part of our lives these days (much like integration). Saying you can integrate is like the idea of having coffee. In reality though, there is a lot more to the whole process.

Are you going to take the easy route, and use instant coffee? It's quick and easy, and for the most part it works, because it is essentially coffee. It's not good coffee, but coffee none the less. In comparison, a lot of integration is accomplished relatively easily, yet often fails to provide the full scope of business objectives to be accomplished.

If you want good coffee, you'd probably want percolated or espresso coffee. In that case, the type of bean and the roasting process becomes important too. This quickly complicates the process until you get the best taste possible: the perfect cup.

To get to the best integration, you need to clearly define what you want to get out of the process. The more you define the required goals -- and how these will be used -- the more likely you are to get to the right solution as fast as possible.

The devil is in the details
The process of integration should preferably take the following steps:

Define who is involved. By definition, integration is between at least two parties or systems, but often includes many more. Make sure you know who all of these parties will be.

Define each party's participation. Each party in the integration has a role to play. These roles can range from being aware of data changes for reporting to participating in the system. Identify what each party is expecting from the process. Preferably, one of these parties also needs to take the lead, so that any issues can easily be resolved.

Define the data. The essence of integration is to share and merge data between parties. There needs to be better clarification, beyond simply mentioning the data that needs to be exchanged (i.e. saying "you need to export your usage statistics"). The format of the data has to be clear, as well as how it will be transferred and accepted. Preferably only exchange the required data, as transferring too much will put an unnecessary burden on your resources.

What happens next?
Most data needs to be changed or reformatted prior to or just after transferring it. You need to state how and when this type of transformation will take place. There may be additional follow-up processes involved after the data has been incorporated.

• Frequency. How frequently do systems need to exchange for the data to be relevant? Implementing something in real-time, or as fast as possible is not the answer. If the relevancy of the data is purely for a monthly report, update the data in a suitable manner. However, the size of the data transfer may dictate that more frequent, smaller batches make more sense.

• Checks and balances. This point is often ignored, and is the root of most issues further down the line. Define how you will validate the integration, as well as how you can follow up if you do have any problems.

Automation: the other pitfall
Integration is not always the only answer. Preferably, you want the various systems to automatically integrate to each other. Automation should not be set up and then forgotten. Periodically, you should report on integrated processes (even more at inception), to make sure that everything is happening the way it should be.

Final thought: it can be good
Integration and automation can be lifesavers if properly implemented. Set up correctly, they can save costs, as well as improve the effectiveness and overall perception of your services and company. Above all considerations, when used properly, both will improve the bottom line.

Filip von Reiche is technology director at Acceleration. Read full bio.