Plan requirement
| Subscription | Any plan |
| Also required | Quality Assurance or Workforce Engagement Management |
| Access | Admin |
Filters decide which conversations reach reviewers. Get this wrong and you review an unrepresentative slice.
Create one
- Click Conversations in the sidebar, then the plus beside Public filters or Private filters.
- Enter a Filter name and set the conditions. Turn on Show all conditions to see every condition type, or search the list for the one you are after.
- Click Apply Filter, and look at a sample of what it selects.
Conditions are combined with AND, not OR: a conversation has to match every one of them to appear. That is the reason a filter with four conditions returns nothing, and it is usually the third condition that did it.
Public is not quite everybody either. A public filter is visible to everyone in the account except workspace agents, so the reviewers get it and the people being reviewed do not.
A new filter starts at thirty days. The date condition on a new filter defaults to conversations created in the last thirty days. A filter that comes back empty is often looking at too short a window rather than at the wrong conditions, and a filter you built to cover a quarter is not covering one until you change that.
Editing, duplicating and deleting
All three sit in the options menu beside the filter. Edit changes the name, the visibility and the conditions; finish with Save changes. Delete removes it from the list.
Duplicate is the one with a twist: the copy always lands in Private filters, wherever the original sat. Useful for trying a variation without an audience, and a surprise if you went looking for the copy next to the original.
Random beats targeted, mostly
A filter that selects only escalated or badly rated conversations tells you what goes wrong and nothing about ordinary work.
Most of your quality lives in ordinary conversations, and that is what a random sample catches.
Where targeted filters earn their place
- A new agent, reviewed more closely for their first weeks.
- A specific process you have just changed.
- Low satisfaction ratings, as a separate exercise from routine reviewing.
Keep those alongside a random sample rather than instead of it.
Check the volume
A filter matching four conversations a week starves reviewers; one matching four hundred means they pick the short ones. Aim at slightly more than you plan to review.
Watch for accidental exclusions
A filter on one channel excludes the others entirely, and a filter on a tag excludes everything untagged. Both look like a full picture and describe a fraction.
Sample what it selects
Read five conversations it picked and ask whether they look like your normal work. That check catches most filter mistakes in two minutes.
Revisit after any change
New channels, new tags, a renamed group. Filters break silently when the thing they reference changes.
Comments
0 comments
Article is closed for comments.