A DDoS test report is the testing deliverable that records what a DDoS test proved, what it could not prove, and what should change because of it. The traffic is gone the moment the generators stop, so the report is the only part of the test still working a month later. It has two readers: the engineer who will fix what failed, and the person who decides which fix gets funded first.
What a useful report contains
The configuration that was tested
A result describes one configuration on one day. The report records which one: rule and policy versions at the edge and the WAF, the mode each control was in, the provider plan, the window with its time zone, and the generator source ranges.
Without that record, nobody can tell later whether today's stack is the one that passed.
Per-vector results, in four states
Pass and fail are too coarse. Each vector ends in one of four states, and the report should name which:
- Held. Legitimate traffic was still served at the top authorised rate. That means "at least this much", nothing more.
- Failed. Goodput fell below the agreed threshold at a stated rate, and the component that gave way, the layer of first failure, is named.
- Not reached. The test stopped before the target did, because the generators ran out of capacity or a stop condition fired. A limit on the test side is not a pass.
- Out of scope. The authorisation excluded it. It is recorded as an unknown, never folded into a clean summary.
A report that collapses these into green and red makes "not reached" look like "held", which is the most expensive misreading a test can produce.
A timeline from both ends
Attack start comes from the generators' send log. Detection, mitigation and recovery come from the target's own telemetry. The two only line up on a shared clock, so the report says how clocks were synchronised; a mitigation that appears to engage before the attack began is a clock problem, not a fast control.
The timeline also records what each control actually did. A rule that matched every attack request while in count mode logged the flood and stopped none of it, which is the block vs count mode trap showing up as evidence.
How to act on the report
Every finding needs three things written next to it: an owner, a fix, and a retest condition. The retest condition is the exact vector, rate and scope that must run again to close it. A finding without one tends to be closed by a ticket rather than by evidence.
Order the findings by how low a component failed against the attack you expect, and then by reach. A component many services share outranks a single endpoint even if it held to a higher rate, because its blast radius is wider.
Expect the next failure to move. Fixing the first layer that gave way makes some other layer first, so the retest should run past the old failure rate rather than stop at it.
Keep a separate section for findings about the test itself: a dashboard that went blank under load, a stop that drained slowly, a generator ceiling below the authorised peak. The next test inherits all of them.
A report is also dated the day it is signed, and every change after that date erodes it, which is why DDoS testing frequency is set partly by what the last report found. The report can only state numbers somebody was recording, so the outcomes it must evidence are best named before the first packet, in the reporting section of a DDoS test plan.