> ## Documentation Index > Fetch the complete documentation index at: https://docs.zeroruntime.ai/llms.txt > Use this file to discover all available pages before exploring further. # Review Results > Follow a run live, find the turn where your agent failed and why, apply the judge's fixes, and re-run. ## Runs The **Runs** tab lists every simulation run with its status, mode, pass rate, cost, and start time. Pass rate and cost appear once the run finishes. | Status | Meaning | | - | - | | **Running** | Conversations are still waiting or in progress. | | **Completed** | Every conversation has finished. | | **Stopped** | You stopped the run before every conversation finished. | Select a run to open it. ## A run The run page updates live while the run is going. The header shows the run's name, status, and mode, with **Stop run** while it's running, and **Re-run Simulation** and **Delete Simulation** after it ends. Below the header you see the run ID, the agent (select it to open the agent in the builder), start and end times, duration, and totals. ![Run page with the Delete Simulation and Re-run Simulation buttons, run details, summary cards for finished, passed, failed, average turns, elapsed time, and cost, the Results by scenario grid, the By persona breakdown, and the conversations table](https://cdn.zeroruntime.ai/zrt/docs-resource/agent-builder/run-detail.png) ### Summary | Card | Shows | | - | - | | **Finished** | Conversations finished out of the total, and how many are still to run. | | **Passed** | Passed conversations out of those judged, and the pass rate. | | **Failed** | Failed conversations, and the scenarios they came from. | | **Avg turns** | The average number of turns per conversation. | | **Elapsed** | Time since the run started, and about how long is left. | | **Cost** | The cost so far, and while running, about what it will cost when finished. | ### Results by scenario A grid with one row per scenario and one column per persona. Each cell shows that pair's result, and **What failed** shows the judge's summary of the latest failure. Use it to spot whether a failure follows a scenario (a gap in the agent) or a persona (a kind of caller the agent handles badly). **By persona** shows each persona's passed and judged conversations as a bar. ### Conversations Every conversation in the run, with its scenario, persona (and environment in audio mode), result, the judge's reason, turns, and cost. | Result | Meaning | | - | - | | **Waiting** | Queued. It starts as soon as a slot in the run is free. | | **Running** | In progress. | | **Success** | The judge passed the agent. | | **Failure** | The judge failed the agent. | | **Stopped** | The conversation was stopped, or couldn't be judged, for example because the agent never joined the call. Stopped conversations aren't charged. | Open a row's **⋮** menu to view the conversation, its scenario, or its persona, or to delete a finished conversation and its result. ## A conversation Select a conversation to open it. Use **Previous** and **Next** to move through the run's conversations. ![Failed conversation with the Failed at turn #6 banner, the audio waveform, the transcript with turn 6 outlined and its criterion and judgement, and the Evaluation panel with the scenario criteria and How to fix it expanded](https://cdn.zeroruntime.ai/zrt/docs-resource/agent-builder/conversation-failed.png) ### Result A banner at the top shows the verdict, the turn that decided it (for example *Failed at turn #3*), and the judge's summary of what happened. ### Transcript The transcript shows every turn between the caller and your agent, with tool calls. In audio mode, a recording plays above it, and the transcript follows the audio. The judge marks the turns that mattered to the verdict: | Mark | Meaning | | - | - | | **Failed** | The agent's reply broke the scenario's criteria at this turn. | | **Lead-up** | A turn that led to the failure. | | **Aftermath** | A turn affected by an earlier failure. | | **Passed** | A turn the judge checked and passed. | Each marked turn shows the **Criterion** it was judged against and the judge's **Judgement**. Use **Jump to** to go straight to a marked turn. ### Evaluation The **Evaluation** panel shows the scenario's success and failure criteria, so you can compare them with the transcript. Evaluation starts when the conversation ends. When the judge has suggestions, **How to fix it** lists the steps to turn the conversation into a pass. Each step has a title, an explanation, and often an instruction you can copy into your agent's [system instructions](/agent-builder/agent/system-instructions-and-greeting). The panel also shows the conversation's **Agent cost**, **Judge cost**, and total. ![Passed conversation with the Passed banner and the judge's summary, the transcript, and the Evaluation panel showing the scenario criteria and the agent cost, judge cost, and total](https://cdn.zeroruntime.ai/zrt/docs-resource/agent-builder/conversation-passed.png) ## Fix and re-run **1. Apply the fix** Update the agent in the builder, for example with an instruction from **How to fix it**. Test it with **Start Test Call** if you like. **2. Commit and deploy** [Commit](/agent-builder/versioning/drafts-and-commits) the change and [deploy](/agent-builder/deploy) the branch. A re-run tests the version that's deployed on the branch when it starts. **3. Re-run the simulation** Open the run and select **Re-run Simulation**. The agent, branch, mode, environments, and limits stay the same as the original. Choose which of the original scenarios and personas to include, for example only the ones that failed, then review the cost and select **Re-run N conversations**. **4. Compare** The re-run is a new run of the same simulation. Compare its pass rate with the previous run on the **Runs** tab. > [!TIP] > > To try a different mode, persona set, or limit, start a new simulation instead. A re-run only lets you pick from the original simulation's scenarios and personas. ## Stop a run Select **Stop run** on the run page, or **Stop Run** in the run's **⋮** menu on the **Runs** tab. The confirmation shows the run's progress, pass rate, and cost so far. * Finished conversations keep their results. * Conversations still waiting or in progress are marked **Stopped**. * You're only charged for finished conversations. ## Delete a simulation Select **Delete Simulation** on a finished run. This permanently deletes the simulation, every run of it, and all their conversations, transcripts, and evaluation results. It can't be undone. Scenarios and personas aren't deleted. They stay in your library for new simulations. Credits already used aren't refunded. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.