Start with a reproducible symptom
At fictional Canal Workshop, calls break up in one room at 11 a.m., but the same laptop works elsewhere. Record time and time zone, location, SSID, BSSID, band, channel, width, client model and driver, application and traffic direction. Compare an affected client with a known-good one in the same place and time window. Repeat off-peak and under load: a reading taken after the room empties is not enough.
1. Separate Wi-Fi from upstream services
First validate association, authentication, IP address and default gateway. EAP/RADIUS, DHCP or DNS failures can feel like slowness even with healthy RF. Compare an approved internal HTTPS endpoint over Wi-Fi and wired access with comparable paths and policies. Internet testing adds WAN and remote-server variables; client link rate is not application throughput. These commands read state or send an authorized probe without changing settings. An open port does not validate HTTPS content, and a blocked ping does not prove a radio problem.
# Poste Windows, service de test fictif / Windows client, fictional test service
netsh wlan show interfaces
Get-NetIPConfiguration
Resolve-DnsName app-test.atelier.example
Test-NetConnection app-test.atelier.example -Port 4432. Coverage: compare RSSI and signal-to-noise ratio
A less negative RSSI is a stronger signal, but noise also matters. An RSSI of −72 dBm and a noise level of −91 dBm give an SNR of 19 dB if measured at the same receiver and time. This is an example, not an acceptance threshold. Compare several locations and orientations with a representative device. The controller may measure the uplink while the client sees the downlink; raising AP power does not fix a client that cannot transmit back. Check AP transitions as well. For persistent coverage problems, survey the site before moving or adding an AP.
3. Interference: identify what occupies the channel
High retries can accompany weak signal, collisions or disruptive transmissions; they do not identify a cause by themselves. Compare channels, width and audible neighbours. Wi-Fi networks on the same channel share airtime, while overlapping channels or non-Wi-Fi emissions can disrupt reception. High utilization alone cannot separate these cases. Where hardware supports it, correlate RF views and spectrum analysis with the exact incident time; measurements at an AP may differ from what the workstation receives.
4. Capacity: measure airtime during the busy period
Good signal with latency rising only at peak hours points toward a capacity problem without proving one. Collect utilization, retries, useful throughput, latency and loss by AP and band. Inspect slow-rate clients and shared channels; client count does not directly measure consumption. At Canal Workshop, 82% utilization at 11 a.m. versus 24% after the meeting, with unchanged signal, warrants checking contention. Verify that the AP uplink, switch, gateway and WAN are not saturating at the same time. Do not add an AP on the same channel without a reuse plan.
5. Cisco Catalyst 9800: start at the client and work upstream
Use a CLI account whose AAA role permits the required show commands. On the controller, confirm the IOS XE build with show version and check syntax in the matching reference. The fictional MAC represents the address actually used for this SSID, which may be private. In client detail, record AP, state, channel, signal and available counters; compare two timestamped samples rather than totals since boot. For a moving client, check mobility history and correlate with the incident. For a radio issue, inspect RRM measurements for the affected AP and band; then check switch port speed, errors and PoE. These commands start no debugs or captures.
! Catalyst 9800 — lecture seule / read-only
show version
show wireless summary
show wireless client mac 0200.0000.0033 detail
show wireless client mac 0200.0000.0033 mobility history6. Cisco Meraki: correlate client history with RF Spectrum
With a Dashboard read-only role scoped to the network, open the client details and align connection and performance history with the incident time. In Wireless Health or the available troubleshooting view, find the last successful step: association, authentication, DHCP or DNS as events allow. For an established session, compare signal, throughput and latency against a known-good device. In RF Spectrum, inspect the AP, band, neighbours and utilization. Views and analysis capabilities depend on model, firmware and access; some APs have a dedicated radio. Check the organization’s licensing model: MR Enterprise includes Meraki Health in the co-term guide, but advanced features and subscriptions have their own terms. For Catalyst, also confirm AP/IOS XE compatibility and applicable entitlements before planning a feature or upgrade.
7. Pilot a correction and define rollback
In read-only mode, record channels, widths, power, profiles and baseline measurements. RF changes are a separate step. If evidence shows contention linked to overly wide channels, try a narrower width in a pilot area with a compatible channel plan; for a coverage hole, validate AP placement instead. Change one variable in an agreed window and preserve WPA and 802.1X. Check whether a profile is shared before editing it. Repeat the same call in the same place under comparable load; check roaming and neighbouring areas too. If loss, latency or association worsens, restore the prior profile or setting and repeat measurements. A focused diagnosis helps choose the pilot; a broader wireless review remains useful when multiple areas or services are affected.