> ## Documentation Index > Fetch the complete documentation index at: https://docs.deepblocker.ai/llms.txt > Use this file to discover all available pages before exploring further. # Help desk and call-back results > How a help desk test is graded, what each verification label means, and what the call-back trap results page shows per person. A help desk test and a call-back trap have their own results views on the campaign page, in place of the per-person table. This page explains both. The standard table is on [Read your results](https://docs.deepblocker.ai/vishing/results/overview.md). ## Help desk results The view opens with one sentence, for example "2 of 6 help desk calls ended with a reset granted without verification. 1 more reset was granted after a verification step. Agents verified the caller on 3 of 6." Under it: how many calls an agent answered, and how many counted as failed under the campaign's rules. ### The three tiles | Tile | What it counts | | - | - | | **Resets granted without verification** | Calls where a reset was granted, or a code read out, and the agent did no verification first | | **Calls where the caller was verified** | Calls where the agent did any verification step | | **Not reached** | No answer, voicemail or the call could not be completed | ### Verification observed The grader reads each transcript and records the strongest verification step the agent took. | Label | What the agent did | | - | - | | **No verification** | Nothing. The caller was taken at their word | | **Callback to a number on file** | Offered or made a call back to a number already on record | | **Manager approval** | Required a manager to approve | | **Ticket or self-service** | Sent the caller to a ticket or a self-service route | | **Identity questions** | Asked identity questions before acting | **No verification** shows in red. Every other label shows in green. ### How a call is graded A call is **Failed** when the agent did something the campaign's failure rules count, without verifying first. The rules are set on the **Help desk** step: a reset granted or a code read out, a reset promised, or personal details confirmed. See [Run a help desk test](https://docs.deepblocker.ai/vishing/campaigns/help-desk.md). Otherwise the call is **Passed**. "Reset granted without verification" means both things on one call: a reset was granted and the label is **No verification**. ### The per-agent table One row per agent, by the name they gave, or **Unknown agent 1**, **Unknown agent 2** and so on. | Column | What it shows | | - | - | | **Agent** | The agent's name, who the caller posed as, and which cover story was used | | **Verification observed** | Every label seen across that agent's calls | | **Reset granted** | Resets granted out of calls, with **without verification**, **promised** and **confirmed details** counts under it | | **Detector verdict** | Calls caught by our detector out of calls scored. See [The "detector caught" column](https://docs.deepblocker.ai/vishing/results/detector.md) | | **Calls** | How many calls that agent answered | A second table lists every call: the number, who the caller posed as, the agent, the detector badge and the time. A **What happened** column puts it in words, such as **Code read out without verification** or **Refused (identity questions)**. **Details** opens the call. See [One call in detail](https://docs.deepblocker.ai/vishing/results/call-detail.md). ## Call-back trap results The view opens with one sentence, for example "3 of 12 people called the number back, 1 gave information." Under it is the typical time from message to call. It also says how many callers our detector scored and flagged. A real employee should not be flagged. ### The four tiles | Tile | What it counts | | - | - | | **Messages sent** | People who were sent the message | | **Called the number** | People who called the trap number | | **Gave information** | Callers who gave the AI what it asked for | | **Did not call** | Messages sent minus people who called | ### The per-person table | Column | What it shows | | - | - | | **Person** | The person, or their team when names are hidden | | **Message** | **Text sent**, **Email sent** or **Text and email sent** with the time, or **Not sent** with the reason | | **Called back** | When they called and how long after the message, or **No call** or **Not yet** | | **Result** | One of the labels below, what was disclosed, and a **Coached** badge when coaching ran | | **Detector on caller** | Our detector's verdict on the caller's voice, or **no call** | | Result label | Meaning | Counted as | | - | - | - | | **Called and gave information** | Called the number and gave something away | Failed | | **Called, gave nothing** | Called the number but held firm | Failed | | **Messaged, no call yet** | The wait is still running | Pending | | **Did not call** | The wait ran out with no call | Passed | | **Message not sent** | No address, opted out, or the daily cap | Not reached | Calling the number is a failure. Giving information is a worse one. **Details** opens the call for anyone who called. ## Problems > [accordion] The help desk view says No help desk call has been answered yet > > The desk did not pick up inside its opening hours, or the calls are still queued. Check the campaign status and the hours on the **Help desk** step. > [accordion] An agent appears twice > > They gave a different name on each call, or no name on one. Rows are grouped by the name given. > [accordion] The call-back view says No message has been sent yet > > Messages go out inside the campaign's call window. Check the schedule, and that the people have an address on the chosen channel. > [accordion] Someone is marked Did not call but says they called > > They called from a number not on their contact, so the call could not be matched. Such callers hear "wrong number" and nothing is kept. Add their number to the contact before the next test. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.