A device walking down a corridor should move from one access point to the next without anyone noticing. Usually it does. When it does not, the device clings to a distant AP at 6 Mbps while a better one is 20 dB stronger, and the user reports "the Wi-Fi drops when I walk to the office."
The decision belongs to the client. Everything the network does — k, v, and r — is advice.
What each amendment does
| Amendment | Name | What it provides |
|---|---|---|
| 802.11k | Radio Resource Measurement | Neighbor reports: a list of nearby APs and channels |
| 802.11v | BSS Transition Management | A suggestion, or request, to move to a specific BSS |
| 802.11r | Fast BSS Transition (FT) | Pre-authentication so the roam skips a full handshake |
| OKC | Opportunistic Key Caching | A vendor approach to the same problem, older than r |
k saves scanning time: instead of probing every channel, the client asks the AP "who is near me?" and gets a curated list.
v lets the infrastructure influence the decision — steer this client to that AP, or tell it the current one is shutting down.
r reduces the handshake cost by caching keys ahead of time. On WPA2-PSK with a slow client, a full reassociation can take 100–300 ms; with FT it can be under 50 ms. For voice, that is the difference between a click and a dropped call.
The roam itself
A client roams when its driver decides the current link is bad enough. Typical triggers:
- RSSI below a threshold (vendor-specific, often −70 to −75 dBm)
- Retry rate above a threshold
- Missed beacons
- A better candidate discovered during a scan
Because the triggers are vendor-specific, two laptops in the same room can behave differently. That is normal and not a configuration error.
Turning them on, carefully
802.11k and v are low risk and broadly supported. Enable them globally.
802.11r is where districts get hurt:
- Older IoT devices, printers, and some Chromebook models fail the FT handshake and simply do not connect.
- Mixing FT and non-FT APs on the same SSID can cause repeated failures.
- WPA3 transition mode plus FT has vendor-specific quirks.
Enable r per SSID, test with the actual device fleets (especially the oldest ones), and keep a non-FT SSID available during the pilot.
A failed roam is worse than a slow one
A client that cannot complete the FT handshake drops and retries, which looks like intermittent disconnects in one hallway. If you enable 802.11r and see that pattern, turn it off before you debug anything else.
Steering: the blunt instruments
| Control | Effect | Risk |
|---|---|---|
| Minimum RSSI | Disconnects clients below a threshold | Aggressive values cause drops, not roams |
| Band steering | Pushes dual-band clients to 5 GHz | Old clients may refuse and fail |
| Load balancing | Refuses associations on busy APs | Clients may not retry another AP |
| 802.11v steering | Suggests a better BSS | Clients may ignore it |
Minimum RSSI is the most misused. Set too high, it disconnects a client that has nowhere better to go. Set it at −70 to −72 dBm for 5 GHz, and verify that the device lands somewhere with better signal.
Worked example: a Chromebook down a hallway
Three APs, 60 feet apart, 14 dBm, −67 dBm at the room edges. A Chromebook walks the corridor:
- At AP-1, signal −55 dBm. No reason to roam.
- At the midpoint, AP-1 is −72 and AP-2 is −60. The client should move; many do at that point.
- If it does not, it continues to −78 and starts retrying frames, consuming airtime.
- With 802.11k, the client knew AP-2 existed and roamed at step 2 instead of step 3.
- With 802.11r, the roam costs 40 ms instead of 250 ms.
The fix for step 3 is coverage design plus k; the fix for step 5 is r, if the fleet supports it.
What breaks in production
- No secondary coverage. A device has nothing to roam to, so it holds on until it drops.
- 802.11r enabled globally without testing old devices, producing "some devices won't connect in the science wing."
- Minimum RSSI too aggressive — the client disconnects, fails to find a better AP, and reconnects to the same one.
- Different security settings per AP on one SSID, so every roam is a full authentication.
- Band steering on legacy clients that only support 2.4 GHz.
Checklist
- 802.11k and v are enabled everywhere; 802.11r is enabled per SSID and tested.
- Secondary coverage meets −72 dBm at every point a device might roam.
- Minimum RSSI, if used, is −70 to −72 dBm on 5 GHz and has been verified with real clients.
- The oldest device fleet has been tested against any roaming feature you enable.
- Roaming is validated with a walk test that measures packet loss, not by reading the dashboard.
Next
The Wi-Fi series continues with two advanced pieces already published: deterministic RF tuning for Wi-Fi 6 and for Wi-Fi 7. Read those once the basics above are in place.