How the estimate works
- Each counted action credits its net minutes: the minutes the work takes by hand, minus the minutes you still spend on it with Coffee Chats, such as reviewing a draft or reading a report.
- Where a source gives a range, the estimate uses the low end. Every total is rounded down and shown as an estimate.
- Only finished work counts: an email that sent and did not bounce, a chat that was held, an application that went in.
- Totals are counted from the work still in your account each time you look.
- This page describes model version
ts-2026-10-27. When a minute value or a rule changes, the version changes and every total, past ones included, is recounted with the new model.
What counts
The key is the action’s fixed name in the model, so any figure can be traced to its row. The grade says how strong each figure’s basis is: A is government data, B a named survey or study, C a vendor’s figure, and D our own assumption. Seven of the ten actions rest on our assumptions, because no published figure we could check covers them. A figure we could not check on its source page is not used.
When each action counts
- People found for you: a contact Coffee Chats found for you, counted once an email to them sent and did not bounce. People you added, imported or who booked through your booking page do not count.
- People researched for you: a research report you spent credits on that finished. When chat prep for that person counts, the research is not counted again.
- First emails sent for you: a campaign’s first email to a contact that sent and did not bounce, once per contact per campaign. An email you send from the composer outside a campaign does not count.
- Follow-ups sent for you: each automatic campaign follow-up that sent and did not bounce. Follow-ups you send by hand do not count.
- Chats scheduled for you: a held chat booked through your booking page, or scheduled for you by Coffee Chats from an email thread. Chats you add yourself or that come in from your calendar do not count.
- Chat prep: the first held chat with each person whose prep was ready before the chat started. A later chat with the same person reuses that research and does not count again. Basic prep with no research on the person does not count.
- Call notes and summary: each held chat whose notes and summary were ready.
- Thank-you notes drafted and sent: a thank-you Coffee Chats drafted from your call that was then sent. A thank-you you wrote yourself, or only marked as sent, does not count.
- Applications tailored: a job you applied to that still uses a resume or cover letter Coffee Chats tailored for it. Improving your general resume does not count, and neither does a job where you replaced every tailored document with your own.
- Applications filled for you: an application the agent filled that went in, meaning the job is marked Applied or a later stage. A fill that was refunded does not count.
Where each number comes from
- People found for you (D). Our assumption: the low end of 10 to 40 minutes to find the right alum and a working email by hand. For comparison, The 2-Hour Job Search budgets about 50 minutes to identify and line up several referral conversations (as summarized by Black Enterprise, July 2012).
- People researched for you (D). Our assumption: 12 minutes to research one person by hand before reaching out. No source we found gives a per-person figure. For comparison, sales teams report 5 to 10 minutes for a light pass and 20 to 30 minutes for a deep pass that ends in a drafted message (FirstSales, on account tiering for outbound, August 2026), and about 6 minutes of research inside a 15 to 18 minute manual prospecting workflow (Unify, on cold email practice and the sales development workflow, July 2026).
- First emails and follow-ups sent for you (D). Our assumption: 3 minutes to copy your template, fill in the person’s details and send it by hand. A campaign sends your own template, or a draft you saved for that contact; the campaign itself uses no AI to write it. Writing the template is not counted.
- Chats scheduled for you (D). Our assumption: 8 minutes of back-and-forth email to agree a time, then sending the invite and link. For comparison, Doodle’s blog (November 2021) reports a study, an ETH Zurich thesis with 96 people, in which Doodle saved about 11 to 15 minutes per group meeting of 4 to 6 people. It gives no figure for one-to-one chats.
- Chat prep (D). Our assumption: 12 minutes researching the person, the same assumption as people researched for you, plus 3 minutes writing questions.
- Call notes and summary (D). Our assumption: the low end of 10 to 30 minutes to write notes, a summary and next steps after a chat of about 30 minutes.
- Thank-you notes drafted and sent (B/C). About 5 minutes to draft a work email: roughly 5 and a half minutes in a Slack-commissioned survey of 8,000 US and UK small-business employees (reported by Fox News, October 2023), and 5 to 6 minutes according to Missive (“Time spent on email”, updated April 2026). The estimate uses 5.
- Applications tailored (B). In Monster’s 2026 State of the Resume survey of 1,001 US job seekers, 36% spend under 15 minutes tailoring a resume for each application and 32% spend 15 to 29 minutes, so the median falls between 15 and 29 minutes (Monster’s January 2026 press release, with the breakdown as reported by Money Talks News). The estimate uses 15.
- Applications filled for you (C). Timed audits put a corporate online application at about 5 minutes on average (InFlight, reported by FedWeek, December 2022), and typically 6 to 10 minutes, most often 6 (Expr3ss, research on time to apply). The estimate uses 5.
What your time is worth
Dollar figures value your time at $15 an hour, our estimate of what a student’s hour is worth. That rate sits below the average hourly wage for bachelor’s-level interns, which NACE reported at $23.35 in March 2026, from its 2026 guide to intern compensation. If you pay for Pro, Analytics also compares that value with what Pro cost you over the same period.For schools
- A school sees time saved only as a total across the students who opted in to its sponsored seats, counted from when each student opted in and never from before the school joined Coffee Chats.
- Figures from fewer than a minimum number of students are withheld and shown as Private. Counts are rounded into buckets and hours are rounded down to whole hours, so a school’s total can read lower than the sum of its students’ own totals.
- Totals cover the students currently opted in, so they drop when a student withdraws.
- These are students’ hours, not staff time.
- The advising-capacity equivalent values the hours of chat prep and tailored applications at $30.93 an hour, the median pay for school and career counselors and advisors in May 2025, from the US Bureau of Labor Statistics Occupational Outlook Handbook. It compares capacity; it is not time a career office spent.
Known overcounts and undercounts
Each of these stays because closing it needs a record Coffee Chats does not keep.Overcounts
These are credited even though the real saving may be smaller.- A campaign email you rewrote by hand before sending is credited at the template rate.
- A campaign email CoffeeGPT rewrote before you sent it is credited at the template rate.
- A campaign first email built from a draft you wrote for that contact is credited like a template send.
- Call notes count for every held chat, and chat prep once per person, even if you never opened them.
Undercounts
This work is done but not credited, or credited low.- A follow-up that sent with no step from you still has the review time taken off.
- A job tailored before 27 October 2026 and filled by the agent later counts only as tailored.
- An application the agent filled that you submitted but never marked Applied is not counted.
- Chat prep regenerated after the chat is not counted.
- A contact you moved back to New by hand loses its credit.
- A job you marked Withdrawn, or moved back to Saved, is not counted, even if the application went in.
- Deleting a contact, a meeting or a job removes its work from the estimate.
What never counts
- Practice calls and mock interviews.
- Emails sent from the composer outside a campaign, including reply drafts and emails CoffeeGPT writes.
- Unsent drafts, failed sends and bounced emails.
- Chats that were cancelled, missed or have not happened yet.
- Applications that never went in, and applications where the agent had nothing to fill.
- Contact imports, calendar sync, reminders, reply sorting and safety checks.
- Building or improving your resume outside a job application.