What extra metrics should I add next, and when should I update the baseline?

Start with the baseline signals you can observe consistently: discovery events over time; first-party retrievals aligned to approved facts; the proportion of questions resolved using approved knowledge; confirmed AI referrals or browser arrivals; and outputs that include visible citations. These form the defensible core you can show a client. Then plan progressive metrics that reveal deeper knowledge and usefulness, not guarantees. Track changes over time, such as the share of core questions covered by approved facts, the rate of new Q&A additions, and whether outputs stay structured. Present each metric as a simple chart with a short caption and a baseline period. Be explicit about limits (for example, caching can affect signals) and remind stakeholders that improvements reflect better knowledge coverage and answer quality, not guaranteed sales.