Updated: Sep 6, 2020
Departmental or workgroup shares have evolved over time, usually with little guidance or structure. Often, at the root level, we find file structures organized around:
Organizational units identified with acronyms. Organizational units and acronyms change over time, making indexing and classification less accurate.
Individuals’ names (Bob’s folder). After Bob leaves the organization, nobody will ever look in that folder again.
Other miscellaneous topics, entities, systems or projects that overlap creating duplicates and lost content.
The overall accuracy of the classification that comes from these types of shares can be improved with some reorganization. A high-level folder renaming will impact the auto-classification accuracy of every file below it in the hierarchy.
Basing root level, first, and second tier folders on a functional model will help in future efforts to organize content. A function is a combination of a verb and a noun like “receive compliance reports,” “plan conferences,” “initiate projects,” or “develop training.” It is not strictly necessary to adhere to the verb-noun combination (“conference planning” will also work), but you will notice that this functional approach will drastically improve overall classification accuracy: up to 95% accuracy compared to an organizational model (80%) or ad hoc/three letter initial model (65%). The reason is that content will be described better and subfolders will be more consistent. This is the key to getting better results with AI or auto-classification capabilities that leverage context as well as content.
As an example, in the Legal department of one organization, folders at the root level were labeled with three letter initials of the responsible attorneys. Because of this, each attorney and their team, set up their own subfolders. One chose “Contracts,” one chose “Contracts and Agreements,” one chose “AGMTS,” and so on. No single query or definition could easily capture the true intent of the foldering.
When setting up new folders or cleaning existing structures for content, keep the following guidelines in mind:
Think about how someone in 5 years, who has no history with the organization, will want to find content. This level of simplicity and clarity will help your classification tools and M365 as well.
Don’t set up a structure that easily provides for a single document to reside in multiple places. Folder structures work better if there is a mutually exclusive place to put things unless you can ensure the indexing allows for reorganization. If you have a folder for emails and a folder for specifications, an email about specifications will be difficult to locate consistently.
Use folder names that help identify a common index field for the documents placed in that area. Significant labor will be saved if a document automatically inherits the folder name as an index value.
In addition to foldering issues, workgroups are the best level to address a number of cleanup activities. Chances are many of the following items will be the responsibility of a limited number of individuals within the workgroup. You can get maximum cleanup benefit with minimum user impact of you focus this way. File analysis tools can report on and often fix issues like:
Odd characters – M365 sometimes has a hard time dealing with odd characters in file names.
Long file paths -- Tools to migrate content to M365 systems may also have a problem with files or folder paths that are too long. A simple report can help identify them.
Duplicates -- Around 20-30% of the files on a typical workgroup share are duplicated, usually with around three copies per duplicate set. As it turns out, 15% of the duplicates cannot be deleted or some larger collection of files will lose integrity since they are part of an application, database, or web-collection. Dealing with duplicates is not a one-size-fits-all solution.
Large files -- By themselves, large files are not bad. But if the file is unnecessary, the benefit of getting rid of it is larger than other sized files.