the completion rates you have heard for online certificates, from about 3% to 85%, can all be accurate at once. They disagree mainly because each one divides by a different group of people and was measured under different conditions: everyone who registered, people who paid, people who finished at least one course, or students whose grade depended on finishing. Course quality explains far less of the gap. Before you compare two rates, check that they share a denominator and a condition.
The question you are actually asking
You are reading a proposal, a press release or a vendor page. One number says 3%. Another says 80%. You want to know which provider is better, or which figure belongs in your grant report.
Think of a completion rate as a fraction with a finish line on top and a starting line underneath. Everyone agrees roughly where the finish line is. The fight is over where the starting line was drawn.
The figures, each with its starting line
Here are the numbers that show up most often when people argue about certificate completion, with what each one actually divides by.
3.13%: everyone who registered for a MOOC. Justin Reich and José Ruipérez-Valiente analyzed edX courses from Harvard and MIT for their Science paper "The MOOC Pivot." As Inside Higher Ed reported in January 2019, "Among all MOOC participants, 3.13 percent completed their courses in 2017-18," down from nearly 6% in 2014-15. The starting line here includes anyone who clicked "enroll," including people who never meant to finish.
46% to 56%: people who paid. The same study found that "among the 'verified' students, 46 percent completed in 2017-18," compared with 56% in 2016-17. Verified learners were those who paid to take "verified" courses. Same courses, same platform, a different starting line, and the rate jumps roughly fifteenfold.
29.5% and 45.7%: one Google cohort, two starting lines. Jobs for the Future reported in March 2021 that across a community-college network offering the Google IT Support Professional Certificate, "2,902 of the 4,497 total students enrolled" had completed at least one course, and "1,325 had completed all five courses required to earn the certificate." Divide 1,325 by 4,497 and you get 29.5%. Divide 1,325 by 2,902 and you get 45.7%. Nothing about the learners changed between those two numbers. Only the starting line moved.
At least 85%: the certificate was half the grade. JFF's Google Career Certificates Implementation Field Guide (31 August 2023) describes South Texas College reaching "a completion rate of at least 85% in one semester" when finishing the certificate counted for 50% of a credit-course grade. The condition matters more than the course: students had a deadline, a grade and a teacher.
80%: a program's own claim. Merit America states on its eligibility page (19 March 2025): "We're proud of our 80% program completion rate." The page gives no cohort, timeframe or method, and the same page says applicants "must be a fluent English speaker, reader, and writer." Its newer Workplace AI Essentials track, built on the Google AI Professional Certificate, publishes no completion rate at all.
70%+: not a completion rate. Google's May 2025 impact report says "over 70% of our U.S. graduates" reported "a positive career outcome (new job, promotion, or raise)," based on a 2024 graduate survey. That figure starts after the finish line. People who did not finish are not in it.
The denominator decoder
Copy this into your review notes. Each row is one figure with its starting line and condition.
FIGURE STARTING LINE (DENOMINATOR) CONDITION SOURCE, YEAR
3.13% all edX MOOC participants free, self-paced, no stakes Reich & Ruiperez-Valiente, 2017-18 data
46% verified (paying) edX learners paid for a verified certificate same study, 2017-18
56% verified (paying) edX learners paid same study, 2016-17
29.5% all enrolled (1,325 / 4,497) college network, Google IT cert JFF, 2021
45.7% finished 1+ course (1,325 / 2,902) same cohort JFF, 2021 (derived)
>=85% students in one credit course certificate = 50% of grade JFF Field Guide, 2023
80% not stated not stated; fluent English req. Merit America, 2025
70%+ graduates only (survey) outcome, not completion Google, 2025A worked example with real numbers
Take the JFF cohort and imagine two providers reporting on it.
Provider A reports against everyone who enrolled: 1,325 of 4,497, or 29.5%. Provider B reports against people who completed at least one course: 1,325 of 2,902, or 45.7%. Provider B looks 16 points better with exactly the same graduates.
Now add a third provider, C, who reports only learners still active at the midpoint. We do not know that number for this cohort, so we will not invent it. You can see the direction, though. Every time the starting line moves later, people who left early drop out of the fraction, and the rate climbs. Researchers call this survivor math: you count the survivors and forget the people who never made it to the line you drew.
Three questions that decode any completion claim
- Who is in the bottom of the fraction? Everyone who signed up, everyone who showed up once, everyone who paid, or only people who reached some midpoint?
- What had to happen to count as finished? The full certificate, one course, a passing grade, or a survey response from a graduate?
- What was at stake, and who checked? A grade, a fee, a coach, a deadline, or nothing? Was the number audited, or self-reported by the program selling the service?
If a claim cannot answer the first question, you cannot use it to compare anything.
Why outcome surveys are not completion rates
Here is the catch with figures like Google's 70%. A survey of graduates tells you something about people who finished and chose to answer. It tells you nothing about how many people finished. A program can have a high graduate-outcome figure and a low completion rate at the same time. Put them in separate rows of any report and never average them.
What a funder should require in a proposal
If you fund or review training proposals, ask for these four things in writing:
- Two denominators, side by side: everyone enrolled on day one, and everyone who attended the first session.
- A definition of finished: for this certificate, the Google credential showing in Coursera by a stated date.
- The condition: deadlines, live sessions, grades or stakes, described in a sentence.
- The misses: how many people did not finish, and where they stopped.
A proposal that offers only one rate, with no denominator, deserves a follow-up question before it deserves a score.
Where the simple explanation stops being true
Denominators explain most of the spread, but not all of it. These figures span 2017 to 2025, different certificates (edX courses, Google IT Support, Google Career Certificates) and different learners. Community-college students, paying MOOC learners and adults on a free state license are not the same people. Stakes and support also change behavior directly: the South Texas result came from a design where finishing counted toward a grade.
So use the decoder to make comparisons fair, not to rank providers. And remember that no public completion rate exists yet for the Google AI Professional Certificate itself. Coursera's program page shows 1,632,553 learners enrolled as of 15 October 2026 and no completion figure.
Adjacent questions
Which denominator should a grant report use? Report both day-one enrollees and first-session starters. If your funder requires one, use the one it names and print the other beside it.
Does paying increase completion? In the edX data, paying learners finished far more often, but they also chose to pay. The study shows an association, not proof that charging people would raise your rate.
Are program-reported rates audited? Usually not. Treat an unaudited rate as a claim and ask for the method.
To see how ed.cafe plans to apply both denominators to every cohort, including the ones that fall short, read the Finish Ledger method.
Sources
- Study offers data to show MOOCs didn't achieve their goals, Inside Higher Ed, January 2019
- Grading Google's Certificates, Jobs for the Future, 25 March 2021
- Google Career Certificates Implementation Field Guide, Jobs for the Future, 31 August 2023
- Merit America program eligibility requirements, Merit America, 19 March 2025
- Workplace AI Essentials career track, Merit America, accessed 15 October 2026
- Google Career Certificates impact report, Google, 29 May 2025
- Google AI Professional Certificate program page, Coursera, accessed 15 October 2026
