Skip to content

FAQ

Q: Why don’t I see Calls Analysis or COPC Evaluation in the analysis wizard? A: These analysis types are only available for Call Center Calls projects. Make sure your project’s type is set to “Call Center Calls” when creating the project.

Q: Can I run both Calls Analysis and COPC Evaluation at the same time? A: Yes. You can select both (and other analysis types) in the same analysis run. Each type generates its own results.

Q: How long does analysis take? A: Typically 1–5 minutes depending on the number of calls and their length. You’ll receive an email notification when processing is complete (if notifications are enabled in your profile).


Q: Will my existing recordings be analyzed when I turn automatic analysis on? A: No. Only calls uploaded afterwards are analyzed automatically. Analyze the backlog once with a manual New Analysis run.

Q: Why do I see one result instead of one per batch? A: Batches are folded into a single merged result per project and analysis type, labelled all analyzed calls, so you always have one view of every analyzed call. The individual batches are not listed.

Q: Can I change the questions or the rubric later? A: Yes, from Configuration → Configure…; the change applies to calls that arrive afterwards. Be careful with custom questions: answers already collected stay attached to the old question list, so the merged result would mix two question sets. The wizard warns you and tells you how many calls are affected.

Q: My template has hundreds of codes, but the merged Custom Questions result only shows a handful of categories. Why? A: A template’s codes are only used when the configuration was saved From template. If the questions were typed in by hand — or the configuration was saved again with the question source switched back to Enter questions — no codebook is attached, so the answers are grouped into categories derived from the answers themselves. Reopen Configuration → Configure…, choose From template, save, then press Refresh categories on the merged result to re-codify the calls already analyzed.

Q: A few calls are missing from the report. Why? A: Either they were never analyzed, or their analysis failed. The report leaves out calls that have been transcribed but not analyzed — typically calls uploaded while automatic analysis was off, which it does not pick up later — and a warning above the metrics says how many; analyze them once with New Analysis. A call whose analysis failed shows the reason under its status in the Files tab — most often Automatic analysis skipped — quota exceeded, meaning your subscription had no minutes left to cover it. The report also shows a banner counting these calls.

Q: Does automatic analysis work with SRT/VTT uploads? A: Yes. Transcript uploads are analyzed on arrival like audio recordings. Because they never passed through transcription, their minutes are charged at analysis time.


Q: The report says 0 calls, but I know the calls were analyzed. What’s wrong? A: Almost always the date range. The report defaults to Last 30 days and filters by the call date, so recordings with older call dates fall outside it. The report tells you when this happens and offers a Show all time button.

Q: Where do the contact reasons come from? A: Uspeech extracts a free-text topic from each call and groups those topics into categories. Projects with frozen predefined codes classify each call directly; open-ended projects are re-grouped daily or when you press Refresh contact reasons.

Q: Which hours count as the day shift? A: 08:00–20:00 by default; everything else is the night shift. Only calls with a real call date/time from your metadata are classified — calls placed by upload time are excluded rather than counted in the wrong shift.

Q: Is the PDF’s executive summary written by AI? A: Yes. It is generated from the aggregated report numbers only — the model has no access to transcripts and does not invent quotes or causes. If generation fails, the PDF is still produced with the tables and a note.

Q: How does Uspeech know which agent handled a call? A: From agent_id (preferred) or agent_name in your call metadata. If neither is provided, the agent name the AI heard in the conversation is used as a fallback — and metadata always takes precedence over it.

Q: Can I fix a call’s date or agent after it has been analyzed? A: Yes. Open the ⋯ menu on the file and choose Edit call metadata, or send a PATCH to the API. The report picks up the new values immediately, and the analysis itself — scores, topics, sentiment — is left untouched, so nothing needs to be re-run.

Q: Can agents see their own results? A: Yes. Add each agent’s email to your call metadata and invite them with the Agent role. They get a login with only their own calls, scores and transcripts, plus a personal report and PDF. See Agent Access.

Q: Can an agent see other agents’ calls or the team’s averages? A: No. Every figure an agent sees is calculated from their own calls. They have no agent table, no team-wide statistics, no project settings and no API access, and they can’t change anything.

Q: Why is an invited agent’s report empty? A: Usually the email in the call metadata doesn’t match the invited address, or the calls were uploaded without metadata, so their agent is only the name the AI heard. The report also covers analyzed calls only and opens on Last 30 days. See If something doesn’t look right.

Q: Does the report only find problem calls, or can it also show the best ones? A: It shows the best ones too. The Exemplary calls section lists the period’s best calls — no critical errors, most of the rubric applied, high scores on what applied — with the calls that reached their goal first. Generate best practices then turns them into three to five behaviors other agents should copy, each with a quote from the call. Both are included in the PDF export, ready to use in training.

Q: Where does “Goal reached” come from? I don’t run Calls Analysis. A: Mark the criterion that records the outcome — a sale closed, a payment promised, an issue resolved — as the goal of the call in your rubric. Its score then answers Goal reached for every call, including calls already scored, without running Calls Analysis.


Q: What happens if I don’t have any evaluation rubrics? A: You need at least one rubric to run COPC Evaluation. Create one in Settings > Evaluation Rubrics (COPC) before starting the analysis.

Q: Can I change the rubric after running an evaluation? A: Each evaluation stores a snapshot of the rubric used at the time of analysis. Changing the rubric afterwards does not affect past results. To re-evaluate with an updated rubric, run a new analysis.

Q: What does “critical error” mean? A: A critical error occurs when a call scores 0 on a criterion marked as “Critical Error” in the rubric. The entire call automatically receives a score of 0%, regardless of how well the agent performed on other criteria.

Q: What score is considered “passing”? A: The Dashboard uses 80% as the passing threshold. Calls scoring 80% or above (without critical errors) are counted in the Passing Rate metric.

Q: Can I use different rubrics for different evaluations? A: Yes. You can create multiple rubrics and select the appropriate one each time you run a COPC evaluation. This is useful for evaluating different call types (e.g. sales vs. support).

Q: Why did a call where the customer hung up after five seconds score 100%? A: Because the agent did everything that had become due in those five seconds. Criteria whose moment falls after the call ended are skipped and awarded their full points — an agent is not marked down for steps the call never reached. Anything the agent had a turn to do before the ending is still scored normally. Expand the call and the skipped criteria are marked N/A. See How Scoring Works.

Q: Why does one criterion show a much higher average than the rest? A: Usually because it is conditional and often does not apply. Calls where a criterion was skipped count toward its average at full marks, so a criterion that applies to a third of your calls will read high regardless of how agents performed on it. Check the Applied to column in the Call Center Report scorecard, or the line under its bar in Score by Criterion: if it applied to a small fraction of your calls, you are reading call mix rather than performance.

Q: What does “no data” mean in the Applied to column? A: Those calls were analyzed before Uspeech recorded which criteria applied, so they are counted apart rather than guessed: 47 of 47 · 17 no data means the criterion applied to all 47 calls that carry the information, and 17 older calls do not. A criterion shows — only when none of its calls in the period carry it. New analyses always do.


Q: Is there a limit on how many criteria a rubric can have? A: There is no hard limit, but we recommend 5–15 criteria for optimal results. See Tips for Best Results for rubric design recommendations.

Q: What format should my CSV file be in for importing criteria? A: A plain CSV with columns: id, name, description, max_points, is_critical, section, and optionally is_goal and excused_on_customer_hangup. No header row required. An .xlsx will not import — export it to CSV first. See Evaluation Rubrics for a complete example.

Q: Do I still need to write “does not apply if the customer hangs up” into my criteria? A: No. Early terminations, criteria whose situation never arose, and mutually exclusive criteria are handled automatically for every rubric. Leaving the text in does no harm, but it adds nothing. What you do still need to write is what earns full marks, and which calls each criterion covers — see How Scoring Works.

Q: Closing steps are docked when the customer hangs up during the close. How do I stop that? A: When the customer says “okay, thanks” and hangs up, the transcript reads like a normal goodbye, so the evaluator cannot tell the agent was cut off. Send who hung up first from your dialer as ended_by in the call metadata, and tick Excused if the customer hangs up before the farewell on the closing steps of your rubric. See Closing steps when the customer hangs up.

Q: If the customer hung up, is a critical criterion such as offensive language still scored? A: Yes. A critical criterion can never be excused by a hang-up. If the agent can be quoted using offensive language, the criterion scores 0 and the call 0%, whoever hung up first — including words said after the customer left, as long as they are in the recording and its transcript.

Q: Can I share rubrics with my team? A: Rubrics created within a team context are accessible to all team members. You can also create personal rubrics that are only visible to you.

Q: What happens if I delete a rubric that was used in past evaluations? A: Past evaluation results are not affected. The rubric snapshot is stored with each evaluation, so historical results remain intact.