Product
doTask ChatbotOne inbox and an AI Agent for WhatsApp, Instagram, Messenger, web chat and emailAI AutomationsWorkflows that run themselves, without losing qualityAI-Powered CreativeSenior creative, faster, with AI in the loop
Performance
Google AdsSearch, PMax and YouTube, priced by the workMeta AdsFacebook and Instagram with creative that keeps winningDigital AdsMeta, Google and TikTok campaigns that convertSocial Media CreativeReels, carousels and content for every platformKOL ManagementInfluencer strategy and execution, end to endSEOOrganic growth that compounds month over month
Creative and brand
Video and Media ProductionBrand films, reels and content at speedBranding and StrategyPositioning, messaging, identity and guidelinesEmail CreationEmails people open, read and buy fromWeb DevelopmentWebsites and landing pages built to sellApp DevelopmentMVP to store-ready, fixed price and date
Try it now
doTask ChatbotTalk to the AI AgentAsk it anything about doTask — it answers from our own pages and hands over to a person when it should.Open the chat →Book a demo
Analytics

What the reports measure

Overview, human support, AI support, tickets, agents, sales, leads.

Zack Lam
Authored by Zack LamFounder, doTask
A shop counter with a customer being served
Analytics
Read it in the appThe same article opens from Help inside app.dotask.io, beside the screen it describes.

Recommended reports (top of Analytics → Reports; owners and admins): with bookings on, Bookings captured by the AI Agent (with how many it took in 90 days) and Booking value by service; with two or more people and ratings switched on, Satisfaction by agent; past 100 conversations in 30 days, Busiest hours. Open this report opens the builder on that template; nothing is saved until you press Save. A report you already saved that counts the same thing is not suggested. Template links are /analytics/reports/new?template=<name>, such as busiest_hours; older links with a number still open the same template.

  • Overview — conversations in the period, and how many the assistant resolved without a person.
  • Human support — first response and resolution times for your team.
  • AI support — assistant conversations, handovers, and model cost.
  • Tickets — email tickets by state and time to resolve.
  • Agents — per teammate: assigned, solved, response time.
  • Sales — orders on the connected Shopify store in the period; an order is assisted when its checkout email belongs to a contact who wrote to you in the seven days before, AI-assisted when the AI Agent replied in that window. Orders are read live from Shopify and never stored. Product links sent counts AI Agent replies that looked products up.
  • Leads — contacts created in the period with an email or phone, by channel, plus leads a Flow collected.

Every number is read from the same conversations the inbox shows, for the period picked at the top.

Custom reports

Analytics → Reports → New report. Pick what to count (conversations, messages, tickets, bookings, ratings, AI reviews, survey answers), the measure (a count, an average, a total, a share), what to group by (day, week, month, hour, channel, agent, service, topic…), filters, the period and how to show it. The result runs as you change it. Save it with a name; set “Email it” to weekly or monthly and add addresses to have it sent on Monday mornings or the 1st. Download CSV gives the rows.

Topics and quality

Analytics → AI support → Topics & quality. Once a conversation the AI Agent replied in has been quiet for half an hour, a short model call reviews it: the topic it was about, a 1–5 score for how the AI Agent handled it, whether the customer got what they came for, the customer’s mood at the end, and one line on what would have helped. Topics are named in plain words; give the reviewer your own list under AI Agent → Configure → Review topics to file under fixed headings. The review also shows on the conversation itself.

TipRead “What would have helped” weekly: the most repeated line is usually an answer missing from your knowledge sources. Each review is one small model call, billed under Usage as “review”; switch it off under Configure if you would rather not.
See also:All documentationFAQ