“HO success rate dropped” is a symptom, not a diagnosis. Before touching a single parameter, I classify the failure. The classification follows the 3GPP MRO logic and it saves hours.
Step 0: Is it a parameter problem at all?
Rule these out first, because no offset tuning will fix them:
- Missing neighbor relation, or a PCI collision/confusion
- Preparation failures: admission control reject, X2/Xn or transport issues
- Execution failures: RACH problems on the target (preamble, root sequence, UL coverage)
If preparation and execution look clean, move on to the tree.
Branch 1: Too late
The UE loses the radio link in the source cell before the handover happens (or while it is happening), then re-establishes in the target.
- A3 offset too high, TTT too long, or hysteresis too wide
- CIO set against the neighbor that should be taking the traffic
- A coverage hole or a sharp SINR drop at the edge (corners, tunnels, high speed)
- Typical sign: RLF in the source cell with poor SINR just before the failure
Branch 2: Too early
The handover succeeds, then the link fails shortly afterwards in the target, and the UE re-establishes in the source.
- A3 offset too low, TTT too short, hysteresis too small
- The target looks good only briefly, so you get ping-pong and unstable edges
- Typical sign: a high ping-pong ratio and a short stay in the target cell
Branch 3: Wrong cell
The handover succeeds, the link fails shortly afterwards in the target, and the UE re-establishes in a third cell.
- No dominant server (pilot pollution or overshooting cells)
- CIO or neighbor priorities pushing UEs to a cell that is not the best one
- Missing neighbor relation to the cell that actually serves the area
- Typical sign: re-establishment into a cell that is not in the source’s neighbor list
Parameters behind the tree
- A3 offset, hysteresis, TTT, CIO
- Timers: T310/N310/N311 (radio link failure) and T304 (handover execution)
- Neighbor relations and PCI plan
- Coverage and tilt, since parameters cannot fix what RF should have fixed
My rule:
Fix neighbors, PCI and coverage first, and tune offsets last. Offsets only move the problem around if the RF plan is wrong.
Parameter names differ by vendor, but the logic is the same everywhere.
What is the biggest handover failure cause you see in your network: too late, too early, or wrong cell?
LinkedIn: 