When finding files using CSOM, do folders improve performance?

Question

I know that, as a general rule, metadata should be used rather than folders in SharePoint. However, I haven't found anything that discusses the benefits of metadata over folders when it comes to finding files using CSOM.

I have a library that stores files that are associated with a non-SharePoint web form that pertains to customer feedback. I only need to find the existing files for a specific customer, or add new files for that customer.

If I have a library with 5000 files, and I want to find all files where Company = "Some Value", the inspection (either in C# or in SharePoint as it implements CAML filters) must iterate over all 5000 files in the library to find the result set.

If, however, I structure the files so that there is a folder for each company, then I can easily find all files where Company = "Some Value" by setting the FolderServerRelativeUrl and returning all the items the folder contains.

public void DisplaySharepointFiles(string folderUrl = null, int parentFolderId = 0, int level = 0)
{
    var sourceList = ClientContext.Web.Lists.GetByTitle(DocumentLibrary.Title);

    if (folderUrl != null)
    {
        Query.FolderServerRelativeUrl = folderUrl;
    }
    ListItemCollection items = sourceList.GetItems(Query);
    ClientContext.Load(items);
    ClientContext.ExecuteQuery();

    List<SharepointItem> folderList = new List<SharepointItem>();

    foreach (var item in items)
    {
        var spItem = CreateSharepointItem(item, parentFolderId, level);
        if (spItem.IsFolder)
        {
            folderList.Add(spItem);
        }
    }
    //do some unrelated stuff

    //subfolders & files
    foreach (var sItem in folderList)
    {
        DisplaySharepointFiles(sItem.ItemPath, sItem.Id, level + 1);
    }
}

Question

In light of my use case, are there benefits to avoiding folders and using metadata for locating specific files in a library using CSOM, or are folders less expensive in terms of performance?

Answer 1

As a high-level thinking on algorithms, the deeper the tree is, the longer the crawling algorithm is expected to run at least when compared to one-dimensional structure. If you know the "Company" as in your case, sure you can narrow the crawled content, but wouldn't the same apply if it's just a metadata, which can also be fetched with CSOM?