"Export Directly from Spoke Model" or "Import from Spoke to Hub and Export from Hub"


Hi Everyone,

This question has probably been asked before, but wanted to get opinions on the two options. I have heard hard-lined opinions that outbound data should always be brought into Hub and then exported to another spoke model or external system.

I've always taken a combined approach that focuses on mitigating the amount of unnecessary integrations while taking into account space and performance implications of where the data is housed.

Should outbound data always be in HUB or a consolidated Outbound Data Hub or is there merit to exporting directly from the Spoke if the below are true?

  • No negative impact to model performance
  • Data doesn't have to be combined with data from other spoke
  • Limited number of other spokes/external systems pull the data


  • Dikshant

    Model-to-model Import takes less time rather than exporting the data out of Anaplan. It depends upon if the dataset you're exporting is an incremental or full load along with frequency of the exports. But don't standardize it, i would say and take a wiser call based on all these factors weather to send it from spoke or from Data Hub based on different datasets.

  • @Taylor Moore

    If the data does not need to be consolidated with another data set (another model), just export directly from the model. This way, the exported data will always be up to date.

    If the data does need to be consolidated, think about creating an "Export Hub" model so you keep your inbound Data Hub clean which is only used for inbound data sources.

  • The one thing that will always be true is that every rule has an exception. 🤣

    In addition to the considerations you raised, here are a couple others to contemplate: A Hub might create a more controlled process, with clear line of sight to the data that was sent outbound regardless of changes to the "live" model. Additionally scalability may be increased as you can recycle some of the integration coding (same workspace/model ID across all exports).

    From my vantage, if the performance is not impact and there is no perceived benefit in staging the data in a Hub, then adding the hop ultimately only serves to create lag in the integration process. There are no hard and fast rules and the approach will ultimately come down to the business objective. Just my two cents. 😉