Project recommendations

The recommendations are based on common scenarios and input from implementation projects.

The purpose of these guidelines is to ensure stability and the best possible performance for your Commerce Connect implementation. Most projects, especially first-time projects done by development teams with limited commerce experience, should follow proven recommendations to avoid mistakes, technical debt, and performance and scalability issues.

Before making any exceptions from implementation recommendations, ensure that you have a solid understanding of the underlying framework. Deviating from the recommended way of doing things may create more work than necessary, and make future upgrades more difficult. See the Optimizely Commerce Connect system overviewOptimizely Commerce Connect system overview to learn more about the Commerce framework.

Commerce Connect projects often take more time than originally estimated due to the complexity of pricing, external integrations, and shipping and payment providers. Allow enough time for planning and design, and start small. Especially for first-time complex projects, it is recommended to do a proof-of-concept first, with a complete order process in place.

When altering workflows or other downloadable code added to your solution, make it a habit to comment and document the code in the project clearly. Well-documented code maintained in separated modules makes future system upgrades much easier.

Start early with the catalog modeling, since this can be complex and often requires more design than originally realized. Let the website catalog hierarchy guide your catalog structure, and limit the number of variants/SKUs associated with each product to avoid performance implications.

See Guidelines for working with catalogsGuidelines for working with catalogs for information about catalog modeling.

The built-in search API is useful especially for category landing pages and large catalog structures. It ensures better website performance, offloads database search and retrieval procedures, and provides paging and facets functionality. See SearchSearch.

The meta-field/model field search options are often misunderstood and misused. Two typical examples are:

You should have a clear strategy for caching. Either turn catalog caching on, or implement your own solution. Consider this especially for pricing, and warehouse and inventory updates. See CachingCaching.

Store helpers are useful, but you need to understand how they work before using them. You can download helper classes here (requires login). These can be modified to suit your needs, for instance by adding caching and other logic. As an example, the StoreHelper has no caching for prices, so this needs to be added.

Other limitations to helper implementations are, for instance, that CartHelper does not allow a variant/SKU to be added multiple times as multiple line items. Regarding ownership of business logic, for StoreHelper use the ReadOnlyPricingLoader instead. For CartHelper, download the code and make it "your own."

When implementing payment processing and shipping rate calculation, use the provider model. This makes them available in Commerce Connect and includes storage for configuration settings. See PaymentsPayments.

Modifying the database includes editing or deleting database tables and stored procedures. This can break functionality and prevent future upgrades.

The MetaImage metafield is used for storing images. If you use this metafield type in catalog meta-classes, it works for smaller, low-traffic websites, but is inefficient for larger websites as the output is stored in the database. Also, it is not accessible outside of an entry. However, the MetaImage meta-field performance was improved recently.

For example, avoid using GetDescendents to load everything recursively and then filter by something to find what you were looking for.

With complex solutions, where code is scattered across multiple components, it may not be easy to see how your system is loading content excessively. Everything on its own might look reasonable, but once those processes are combined, looped and repeatedly requested under load with a large content set, you may have problems.

In many cases, you can resolve this by doing the following:

There is no "one size fits all" recipe to avoid excessive content loading. And, you should not optimize prematurely. Think about what you are doing and how it might react in a production environment, with a bigger content structure and under load.

In addition, it is good practice to profile the application, even without evidence of performance problems. This is recommended even if the CMS is installed on a dev machine with no real load and runs along certain central routes/code paths. Finding something in the code that is easy to fix and makes the application 30% faster is a good feeling and good value even if the application is running acceptably in the production environment.