Start with the symptom
An evening slowdown can have several causes: other traffic in your home, a busy wireless channel, the destination service, or congestion somewhere beyond the router. A slow browser test alone does not identify the cause. Record whether the problem is low throughput, delayed game responses, dropped calls, or only one slow website.
Separate the local network from the wider path
On the same device, compare a wired connection to the router with Wi-Fi, if both are available. Disable the unused interface so the route is clear. Pause your own downloads and cloud backups for a quiet comparison, then record the normal busy condition separately. Do not interrupt essential traffic or change shared equipment without permission. A better wired result narrows the wireless path as a possible contributor; it does not prove the ISP is fault-free.
Keep a comparison log, not a universal benchmark
Repeat the same task at evening and quieter times over several days. Keep the device, browser, connection method, test destination, and approximate location consistent. Alternate the order of wired and wireless runs rather than always testing one first. Note other household traffic, negotiated link speed, the application symptom, and any test validity warning. Save unsuccessful runs too. Differences between destinations are useful context, not a head-to-head ISP ranking.
Do not infer a provider diagnosis from timing alone
A repeatable evening pattern is evidence to investigate, not proof of node saturation, a peering dispute, or intentional throttling. If you suspect traffic shaping during peak congestion windows, follow a structured protocol for detecting ISP throttling to compare encrypted versus unencrypted transfers systematically. A VPN changes the route and adds overhead; a different result with it does not establish intent. Missing replies in a traceroute can reflect diagnostic rate limiting rather than dropped forwarded traffic. Compare several services and share observations with support without presenting a browser estimate as a root-cause verdict.
Try reversible changes one at a time
If the symptom follows your own uploads or downloads, test a schedule change or an appropriate router queue-management setting. OpenWrt documents SQM as a combination of scheduling, queue management, and shaping, with CPU and configuration constraints. It may trade throughput for responsiveness and cannot repair every congested upstream link. Save the previous settings, verify compatibility, and restore them if service becomes unstable. There is no guaranteed flat ping.
Escalate with records and check the offer
Send your provider dated examples, the wired comparison, affected services, and whether ordinary traffic or test warnings were present. Ask what further checks they can perform. Compare a new plan using its actual terms and the provider's Broadband Consumer Label where available; price, typical speeds, data allowances, and network-management disclosures matter. Fiber is not immune to congestion, and switching technology is not a guaranteed fix.