[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-cost-per-resolution::en":3,"gloss-cluster-cost-per-resolution::en":23,"gloss-next-cost-per-resolution::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"cost-per-resolution","analytics","Cost per Resolution","Cost per resolution is the fully-loaded cost of closing one support case: total support spend for a period divided by the number of cases resolved in it. It is the metric that turns an AI support tool from a demo into a business case, because it is the one number that can be compared like for like against the human channel it is supposed to replace or relieve. Building it honestly means being strict about the numerator. Vendor subscription and per-resolution or per-token usage fees are the obvious inputs. The ones teams leave out are the ones that decide the answer: the human hours spent on escalated cases the assistant could not close, the ongoing work of maintaining the knowledge base the assistant reads from, the initial integration effort amortised over a sensible period, and the review time spent checking the assistant's answers while trust is still being established. A cost per resolution that counts only the invoice reliably understates the true figure, sometimes by enough to reverse the comparison. The denominator has the same definitional problem as deflection rate, and for the same reason: a resolution the customer did not experience as resolved is not one. Use a consistent resolution definition across both channels, or the comparison is meaningless — the usual mistake is to apply a strict standard to human agents and a loose one to the assistant. Two refinements make the number more useful. Segment it by case type, because AI economics are very good on high-volume repetitive questions and much worse on rare complex ones, and a blended average hides both. And track it over time rather than at a single point: cost per resolution usually improves for months after launch as coverage widens and the knowledge base fills in, so an early reading understates the steady state as reliably as a vendor's projection overstates it.","Cost per resolution is total support spend divided by cases closed. Counting only the vendor invoice understates it enough to reverse an AI-versus-human case.",null,[11,14,17,20],{"slug":12,"name":13},"active-user","Active User (DAU, WAU, MAU)",{"slug":15,"name":16},"deflection-rate","Deflection Rate",{"slug":18,"name":19},"total-cost-of-ownership","Total Cost of Ownership (TCO)",{"slug":21,"name":22},"usage-based-pricing","Usage-Based Pricing",[24,28,29,33,36,37,40,43,46,49,52,55],{"slug":25,"category":5,"name":26,"updated_at":27},"ab-testing","A\u002FB Testing","2026-08-24T02:46:38+00:00",{"slug":12,"category":5,"name":13,"updated_at":27},{"slug":30,"category":5,"name":31,"updated_at":32},"autocapture","Autocapture","2026-08-24T02:46:37+00:00",{"slug":34,"category":5,"name":35,"updated_at":32},"customer-data-platform","Customer Data Platform (CDP)",{"slug":15,"category":5,"name":16,"updated_at":27},{"slug":38,"category":5,"name":39,"updated_at":27},"guardrail-metric","Guardrail Metric",{"slug":41,"category":5,"name":42,"updated_at":32},"identity-resolution","Identity Resolution",{"slug":44,"category":5,"name":45,"updated_at":32},"multi-touch-attribution","Multi-Touch Attribution",{"slug":47,"category":5,"name":48,"updated_at":27},"novelty-effect","Novelty Effect",{"slug":50,"category":5,"name":51,"updated_at":32},"retention-curve","Retention Curve",{"slug":53,"category":5,"name":54,"updated_at":27},"sample-ratio-mismatch","Sample Ratio Mismatch (SRM)",{"slug":56,"category":5,"name":57,"updated_at":27},"seat-utilization","Seat Utilization"]