What it is

Enterprise Reports is an enterprise package for operations leaders. It is off by default. Velaro turns it on for your account; each item below is switched on separately, and an item that is off is hidden and refused by the server.

What is in the package

1. Program, website and platform breakdown. Reports, Enterprise Reports, Programs. Compare every website, team, platform, device or entry page side by side: chats, missed chats, transfers, wait, first reply, QA score, QA coverage and negative sentiment.

2. Skill routing by entry point. Skill routing, Requirements. A requirement can apply only when the chat starts on a page whose address contains some text, when a query-string parameter has a value, when a custom value matches, or when a pre-chat answer matches. Requirements with no condition apply to every chat as before. The simulator accepts a page address and custom values so you can test before turning a requirement on. Skill routing narrows which agents on the team can take the chat; the team still comes from the website or workflow.

3. Individual agent scorecard. Enterprise Reports, Agent scorecard. Each agent by name: chats, resolution rate, transfers, handle time, first reply, AI score, supervisor-reviewed score and sentiment impact. Sentiment impact is the agent's net sentiment (positive minus negative) compared with every other human-handled chat, in percentage points.

4. QA coverage and coaching queue. Enterprise Reports, QA coverage and coaching. Shows the share of resolved chats the AI scored and supervisors reviewed, by week and by website, and lists low-scoring chats nobody has reviewed, per agent. A supervisor review in Quality QA takes a chat out of the queue. Coverage below 100 percent usually means a rubric samples fewer than 100 percent of chats, a monthly scoring cap was reached, or the chat had fewer than two messages; the report says which settings apply.

5. Sentiment by program, agent and topic. Enterprise Reports, Sentiment. Negative share with change against the previous period of the same length.

6. Keyword intelligence and topic radar. Reports, Keyword Intelligence and Enterprise Reports, Topic radar. Top phrases with growth, plus topics ranked by volume, growth, transfer rate and handle time.

7. Automatic tagging. AI Field Extraction reads each resolved chat and fills line of business, intent and state. With automatic tagging on, the line of business and intent also become conversation tags (for example Line: Auto, Intent: Claim) so agents stop typing them. Turn on Create tags in AI Field Extraction settings.

8. Weekly executive brief. Reports, Email Subscriptions, Executive Brief, Weekly. Delivered Monday morning in your time zone. With this item on, the email adds program comparison, QA coverage, the coaching queue, the topic radar with service opportunities and, if you have the Intelligence Suite, the health score and volume forecast. Use Email me this week's brief on the Weekly brief tab to test delivery without waiting for Monday.

9. Bonus insights. Enterprise Reports, Bonus insights. Six views: the QA coverage gap (manual versus AI review, with the reviewer hours it would take), next-best-fix opportunities, an early warning for escalation risk, an agent-by-topic skill-gap matrix, handle time versus quality outliers, and root causes of abandoned chats. Every dollar figure shows its assumptions so you can change them to your own.

10. Operations Briefing. Enterprise Reports, Operations Briefing. One run produces every report in the package from a single load of the data, plus one short AI summary that is cached and size-limited. The page shows its own AI cost, so the summary is never a surprise.

For how topics, themes and sentiment are chosen and combined in these reports, and every way to shape them, see the article "How Velaro determines chat themes, sentiment and topics, and every way to shape them". Year in Review and New-agent ramp are upcoming reports in this package and are not yet available.

Five more reports (items 11 to 15)

Each is its own item, switched on separately, and each sits on the Enterprise Reports page as its own tab. All five are inferred from the chat data the platform already holds; none is a survey. Each tab shows the caveats the server sends. The sections below say exactly how each number is produced, so you can explain it and its limits honestly.

Repeat-contact leakage (item 11)

Enterprise Reports, Repeat contact. Of the chats that closed in the range, the share where the same contact came back within the return window (1, 3, 7, 14 or 30 days; default 7). Only Resolved chats from identified contacts that ended at least one window ago are counted, and the report says how many were too recent to judge. It also shows how many came back on the same topic (as a share of all counted chats and as a share of those who came back), the median hours to return, how often a chat closed by a bot versus a human came back, and the split by agent, topic and program.

Repeat-contact leakage: what it can and cannot tell you

A return is a signal that the first chat may not have settled the matter. It is not proof, because people also come back with a new question; the same-topic figure is the closer hint. Contacts are matched on the contact record, so a visitor who never identified themselves is never counted as returning.

Demand vs coverage heatmap (item 12)

Enterprise Reports, Demand vs coverage. A day-by-hour grid, in the time zone shown on the page, of how many chats that needed a person arrive (queued for or assigned to an agent, or missed or abandoned) against the agent hours available in that hour. Chats the bot handled alone are counted separately and do not drive staffing. This report ignores the From and To dates: it always uses the last 1 to 12 whole local weeks you pick (default 6), averaged per week. Coverage means agent hours in the Available status (Busy and Away do not count), measured from status changes.

Demand vs coverage: gaps, forecast and limits

Pressure is chats per available agent hour. A cell is marked GAP when chats arrive and no agent is available, or when pressure is higher than an agent can handle; the assumed handling time (default 10 minutes) and chats at once (default 3) are shown on the page. It lists the biggest gaps (by missed plus abandoned chats), after-hours volume and a next-week forecast (the median of the same hour over the weeks shown). Website and team filters narrow demand but coverage is always site wide. It does not say how many sales or chats were lost.

Bot false containment and handoff regret (item 13)

Enterprise Reports, Bot containment. A chat is bot touched when an automation ran or the AI resolved it. Claimed containment is the share of bot touched chats that the bot closed with no handoff, Resolved, with no missed reason: what a bot dashboard would claim. True containment removes the bot-only chats where, within 72 hours, the same contact started another chat that reached a person or was about the same topic. Only chats at least 72 hours old are judged.

Bot containment: handoff regret and limits

Handoff regret counts chats handed to a person that ended badly: negative overall sentiment, a negative visitor message just before the handoff, a QA score below the site's low-score threshold, or missed or abandoned. It also shows the median minutes of bot time before handoff. Some handoffs are policy (refunds, legal, account changes) and not a bot failure. A returning visitor does not prove the bot answered wrongly. Anonymous visitors cannot be followed for a return.

Inferred customer effort index: how it is scored (item 14)

Enterprise Reports, Customer effort. Each chat gets a score from 0 to 100 added up from these signals: each transfer 15; each reopen 15; each visitor message that repeats an earlier one 10 (wording at least 70 percent the same); each email or order number the visitor had to give twice or more 10; a queue wait over 60 seconds 5; each visitor turn above the site median 2 (at most 10 in total); the same contact starting another chat within 7 days 10 (only for chats that ended 7 or more days ago); and each clearly frustrated visitor message 5 (at most 15 in total). The total is capped at 100.

Inferred customer effort index: what it can and cannot tell you

The page shows the median and the 90th percentile (never the average), the share of chats that could be scored, how much each signal adds, and the split by topic, program and team. Higher means more effort. It shows where customers are likely working hard. It cannot tell you how a customer felt, it is not a survey, and it must not be compared with a CES score. The repeat and re-ask signals need facts recorded from the chat while its text still exists; a chat without them scores 0 on those signals and the page shows the coverage percentage.

Promise ledger: where promises come from (item 15)

Enterprise Reports, Promise ledger. Only messages written by a human agent are read (not bots, automatic replies or system notes). A sentence counts as a promise when it commits the agent or team to contact the customer later, such as "I will follow up", "we will get back to you", "someone will call you", "a manager will reach out". Questions, offers ("can I", "if you like") and negatives ("I can't") are ignored. The deadline is read from the sentence (within 2 hours, today, tomorrow, by Friday, next week) and defaults to the end of the second business day.

Promise ledger: kept versus overdue

A promise is Kept when a human agent sends another message in the same conversation at least 5 minutes after the promise, or the same contact gets a new human-handled conversation, before the deadline. Past the deadline with no such follow-up it is Overdue; before the deadline it is Open. The kept rate is kept divided by kept plus overdue. Dismissed promises are not counted.

Promise ledger: what the customer sees and the limits

Managers and administrators see the promise sentence (emails, phone numbers and order numbers replaced with placeholders) and the contact, and can Dismiss a promise that was kept another way. Other users see the counts and the split by agent only. Detection uses fixed phrases: about 85 percent of what is listed is a real promise and about 60 percent of real promises are caught. Chat, email conversations handled in Velaro and transcribed phone calls are all covered; only contact outside Velaro (a personal mailbox, an unrecorded call) is invisible, so Overdue means check, not broken. Phone transcripts have no speaker labels, so a promise the caller makes can be listed; Dismiss it.

Year in review (item 16)

Enterprise Reports, Year in review. It builds one document that looks back over the last 12 whole calendar months in Pacific time (for example, in October 2026 it covers October 2025 through September 2026). Pick the dates if you want a different range, then press Generate. The document is built in the background, so the page shows a progress state while it works; when it finishes, a Download link appears. If the build fails the page says so and shows the reason, and nothing is downloadable until you generate again. Nothing is built or fetched until you press Generate.

Year in review: what it can and cannot tell you

Sentiment at the moment of a bot handoff and the bot-phase history are recorded from the day that capture went live, so months before that date show Insufficient data for those parts, never an estimate. Chat themes that have not been classified are shown as Unclassified rather than guessed. Like the other Enterprise Reports, the Year in review tab only appears for accounts that have the item, and a site without it gets a refusal from the server.

New-agent ramp: how new agents compare with experienced ones

Enterprise Reports, New-agent ramp. An agent counts as new when their first chat with a customer falls inside the dates you picked. For each new agent the report shows, week by week from that first chat, how many chats they handled, the median time to their first reply, their average QA score and their average chat rating. Next to it sits the same picture for experienced agents (anyone who handled chats before your start date), as the median across those agents over the same dates. Week 1 is the first seven days, the table stops at week 12, and a hire late in your range shows fewer weeks. An empty cell means there was no data that week, for example no scored chats or no ratings, and is never shown as zero. A week with only a few chats is a small sample, so read it with care. Test and demo chats are left out unless you tick Demo data only.

Year in review (item 16)

Enterprise Reports, Year in review. It builds one document that looks back over the last 12 whole calendar months in Pacific time (for example, in October 2026 it covers October 2025 through September 2026). Pick the dates if you want a different range, then press Generate. The document is built in the background, so the page shows a progress state while it works; when it finishes, a Download link appears. If the build fails the page says so and shows the reason, and nothing is downloadable until you generate again. Nothing is built or fetched until you press Generate.

Year in review: what it can and cannot tell you

Sentiment at the moment of a bot handoff and the bot-phase history are recorded from the day that capture went live, so months before that date show Insufficient data for those parts, never an estimate. Chat themes that have not been classified are shown as Unclassified rather than guessed. Like the other Enterprise Reports, the Year in review tab only appears for accounts that have the item, and a site without it gets a refusal from the server.

What the data can and cannot show

Chat text is cleared after archive, so anything that needs the words must be recorded while the chat is still readable. Sentiment at the moment of a bot handoff, the bot-phase history, and the frustrated-visitor-messages part of the effort score are recorded when a chat closes, from the point that capture is live. Chats closed before then show Insufficient data for those parts, never an estimate. Per-message sentiment uses a word list for English, Spanish, French, German, Portuguese and Italian and AI scoring for every other language (cached, with a daily allowance per site; a message past the allowance stays unscored). The overall sentiment label (one per chat) is separate. Never say a figure is available when the page shows Insufficient data.

The Enterprise Reports package and feature keys

Enterprise Reports is an enterprise-internal package, off by default, switched on by Velaro staff per account. Every item is enforced on the server: a site without the item gets a refusal, not just a hidden menu. If a customer asks for a report they cannot see, say it is part of the Enterprise Reports package and to ask their Velaro contact. Items 9 and 10 are Bonus insights and the Operations Briefing (one run of every report with one short AI summary).

Enterprise Reports feature keys

Program breakdown EnableEnterpriseReportsProgramBreakdown; Entry point routing EnableEnterpriseReportsEntryPointRouting; Agent scorecard EnableEnterpriseReportsAgentScorecard; QA coverage EnableEnterpriseReportsQaCoverage; Sentiment breakdown EnableEnterpriseReportsSentimentBreakdown; Keyword topics EnableEnterpriseReportsKeywordTopics; Auto tagging EnableEnterpriseReportsAutoTagging; Weekly brief EnableEnterpriseReportsWeeklyBrief; Bonus insights EnableEnterpriseReportsBonusInsights; Operations Briefing EnableEnterpriseReportsOperationsBriefing; Repeat contact EnableEnterpriseReportsRepeatContact; Demand vs coverage EnableEnterpriseReportsDemandCoverage; Bot containment EnableEnterpriseReportsBotContainment; Effort index EnableEnterpriseReportsEffortIndex; Promise ledger EnableEnterpriseReportsPromiseLedger; Year in review EnableEnterpriseReportsYearInReview.

Existing features these build on

Sentiment Reports, Quality QA scoring (the rubric is editable under Quality QA, Rubrics), Skill-Based Routing, Chat Intelligence themes, Report Email Subscriptions and AI Field Extraction keep their own plans and settings. Enterprise Reports adds the analysis on top; it does not change who can use them.

Keys for the New-agent ramp item: EnableEnterpriseReportsNewAgentRamp.

Keys for the five: EnableEnterpriseReportsRepeatContact, EnableEnterpriseReportsDemandCoverage, EnableEnterpriseReportsBotContainment, EnableEnterpriseReportsEffortIndex, EnableEnterpriseReportsPromiseLedger.

Turn it on or off (Velaro staff)

Open the site in Super Admin, Feature packages, Enterprise Reports and switch items on or off. Or run velaro entitlement set-override --site 30747 --key EnableEnterpriseReportsQaCoverage --value true --reason "...". To see what is on for a site: velaro entitlement enterprise-reports --site 30747.

Automation

velaro enterprise-reports program-breakdown --group-by website, or the Velaro MCP tool enterprise_reports with report set to program_breakdown, agent_scorecard, qa_coverage, coaching_queue, sentiment_breakdown, topic_radar, weekly_brief or weekly_brief_send_test. The five newer reports are velaro enterprise-reports repeat-contact --window-days 7, demand-coverage --weeks 6 (from and to are ignored), bot-containment, effort-index and promise-ledger (dismiss with promise-dismiss --id <id>); the MCP tool takes report set to repeat_contact, demand_coverage, bot_containment, effort_index, promise_ledger or promise_dismiss.

Dates and time zones: every Enterprise Report counts days in Pacific time, the same as the other reports. Sentiment at the moment of a bot handoff, and the "frustrated visitor messages" part of the effort score, are recorded when a chat closes. For chats that closed before this was recorded, and whose text has already been cleared, the page shows "Insufficient data" instead of a number.

Promise ledger: where promises come from. The report reads messages sent by human agents in the chat (not the bot, automatic replies or internal notes). A sentence counts when the agent commits to a later contact, such as I will follow up, I'll get back to you, someone will call you. Questions, offers, conditionals and negations are skipped. This uses fixed phrase rules, not an AI model, and runs when the chat closes (and as a background fill for the last 90 days). The deadline is taken from the sentence (within 2 hours, tomorrow, by Friday) or defaults to two business days. Kept or overdue is worked out when you open the report: kept means an agent wrote again in that chat at least 5 minutes after the promise, or the same customer had a later human-handled chat, before the deadline. Chat, email conversations handled in Velaro and transcribed phone calls are all covered; only contact outside Velaro (a personal mailbox, an unrecorded call) is invisible, so Overdue means check, not broken. Phone transcripts have no speaker labels, so a promise the caller makes can be listed; Dismiss it.