Original research · Edition 1 · August 2026
They paid. They still don’t believe they can do it.
Ask an adult learner to rate their confidence out of ten and you will get a nine or a ten. Ask the same person, in the same form, what they are most nervous about, and you get something else entirely. Half of these students rate their confidence at 10 out of 10 and not one rates it below 7. In their own words, 39% say the thing they fear most is whether they can learn it at all.
Snapshot taken 26 August 2026 across 553 enrolled students, 456 of whom have a recorded payment. Coverage varies by field and every figure carries its own n. Composition and stated belief only: no placement, salary or employment-outcome claims. Free to read, cite and republish under CC BY 4.0.
50%
rate their confidence of finishing at 10 out of 10. Nobody rates it below 7 (n=62)
39%
asked in their own words, name their own ability to learn as their biggest fear (n=62)
56%
had never written a line of code at the point they paid (n=113)
62%
already held a degree, and 73% did not study computing (n=113)
Finding 01
On a scale, nobody is worried.
Self-rated confidence of completing the programme, asked at onboarding (n=62), and self-rated commitment to the goal (n=61).
- 10 out of 1031 · 50%
- 8 out of 1014 · 23%
- 9 out of 1010 · 16%
- 7 out of 107 · 11%
- 6 or below0 · 0%
Commitment to the goal
- 5 out of 553 · 87%
- 4 out of 57 · 11%
- 3 out of 51 · 2%
- 2 or below0 · 0%
Half the students put themselves at the very top of the confidence scale, 89% put themselves at 8 or above, and the lowest answer anyone gave was a 7. Not one person in 62 used the bottom six points of a ten-point scale. Commitment is the same shape: 87% choose the maximum.
Read on its own, this looks like a cohort of people who need no reassurance at all. It is exactly the kind of number a school is tempted to publish, and exactly the kind of number that makes onboarding surveys feel like a formality. If everyone says 10, the question is not measuring anything.
That is the first useful result here, and it is not really about us: a self-rated confidence scale, asked at the moment of commitment, appears to measure enthusiasm rather than belief. People who have just paid for something are not in a frame of mind to record doubt about it on a form, and a scale that runs from 1 to 10 but is only ever answered from 7 to 10 has thrown away most of its range before the analysis begins.
Finding 02
In their own words, the doubt comes straight back.
Free-text answer to 'what is one thing you are most nervous or uncertain about?', coded into one primary category per answer (n=62, the same people as Finding 01).
- Can I actually learn this?24 · 39%
- Will I get a job?8 · 13%
- Will it be worth it?8 · 13%
- Can I find the time?6 · 10%
- Nothing, or more excited than nervous6 · 10%
- Other or too vague to code10 · 16%
The single largest category, by a factor of three over anything else, is people doubting their own ability to learn the material. 39% describe some version of not being able to keep up, being a slow learner, not grasping complex ideas, or starting from zero and wondering whether that is survivable. Only 13% name the job market. Only 13% worry the money will turn out to be wasted. 10% say they are not nervous about anything.
The same 62 people produced both charts. The distance between them is the finding. A rating scale said this group was near-uniformly confident; an open box on the same form said four in ten are frightened of the thing they have just bought.
We think the honest reading is that both are true. Confidence in the decision and confidence in the ability are different quantities, and adults making a costly career change routinely hold the first without the second. They are sure they want this. They are not at all sure they are the kind of person who can do it.
The practical consequence is that the doubt does not get spent at the checkout. Anyone designing a reskilling programme, a corporate academy or a public retraining scheme who assumes the hesitant people are the ones who did not sign up has the population wrong. The hesitant people signed up too, and they are sitting in week one waiting to find out whether they are about to be exposed. This is the same barrier we measured on the way in, from a completely separate dataset of 325 applicants, in what stops people retraining for tech. Paying does not remove it.
Finding 03
Most had never written a line of code.
Self-described programming and data experience at the point of enrolling (n=113 of 553 completing this field).
- Level 0 - no coding experience at all63 · 56%
- Level 1 - finished beginner tutorials35 · 31%
- Level 2 - can build simple applications10 · 9%
- Level 3 - has deployed an application4 · 4%
- Level 5 - working developer1 · 1%
56% recorded no coding experience of any kind, and 87% were at level 0 or level 1 - nothing at all, or beginner tutorials only. Exactly one person in 113 described themselves as a working developer.
Put next to Finding 02, this stops being a surprise and starts being an explanation. When four in ten people say they are afraid they cannot learn it, they are not being irrational or performing modesty. They are correctly describing their starting position. They have nothing to reason from, no prior experience of getting stuck on a bug and then unstuck, and therefore no evidence about themselves either way.
It also sets a floor under what any beginner programme has to handle. A curriculum written for people who have already done a bit of Python at university is not addressing this population. The modal student here has never opened a terminal, and the first genuine encounter with confusion is the moment where the fear in Finding 02 either gets resolved or gets confirmed.
Finding 04
They are not uneducated. They are educated in the wrong thing.
Highest qualification and field of study at enrolment (n=113 each).
- University or bachelor's degree56 · 50%
- Diploma or technical school17 · 15%
- High school or earlier17 · 15%
- Master's degree or higher13 · 12%
- Pre-university10 · 9%
What they studied
- Something else entirely37 · 33%
- Computer science or IT31 · 27%
- Business, finance or management26 · 23%
- Medicine or health6 · 5%
- Natural sciences6 · 5%
- Art or design4 · 4%
- Social science3 · 3%
62% already hold a degree - 50% a bachelor’s and 12% a master’s or higher - and 73% did not study computing. The largest single field of study is “something else entirely” at 33%, with business, finance and management close behind at 23%. Median age is 28, mean 29, and the range runs from 17 to 62 (n=130).
This is what the national statistics look like from the inside. Malaysia has near-record-low graduate unemployment and record graduate underemployment at the same time, a pattern we set out with official figures in the underemployment paradox. The people in this table are the individual version of that statistic: degree holders, mostly in their late twenties, paying a second time for a second education because the first one did not route them anywhere.
Note also that 27% did study computer science or IT and are here anyway. A computing degree is evidently not, on its own, sufficient to feel employable as a developer, which is worth sitting with if you run a computing faculty.
English is a quieter constraint underneath all of it: 27% describe their own English as basic, 48% as professional and 24% as native (n=143). Roughly a quarter of this population is learning to program in a second language they do not feel fluent in, while also learning to program.
Finding 05
They buy the opposite of how they say they learn.
Stated learning preferences and support needs (n=62, multi-select) against the programme format actually purchased (n=520 of 553).
- Hands-on projects60 · 97%
- Videos40 · 65%
- Live discussion40 · 65%
- Reading20 · 32%
What keeps them on track
- Mentor feedback54 · 87%
- Accountability check-ins40 · 65%
- Solo progress tracking29 · 47%
- A study group28 · 45%
What they actually bought
- Part-time158 · 30%
- Self-paced129 · 25%
- Full-time, online88 · 17%
- Full-time, in person49 · 9%
- Job-ready programme30 · 6%
- AI-native programme27 · 5%
- Everything else39 · 8%
Asked how they learn best, 97% choose hands-on projects and 87% say mentor feedback is what keeps them going. Only 47% think solo progress tracking works for them, and 65% want live discussion. Two thirds say they would rather work with a small team than alone. 55% say they are motivated more by structure than by flexibility.
Then they buy the other thing. 55% of the 520 students with a recorded format bought part-time or self-paced study, and only 9% bought a full-time in-person programme - the format that most closely matches the mentor-led, live, team-based experience they say they want.
We do not think this is inconsistency. It is a constraint binding harder than a preference. Adults with jobs, rent and dependants cannot buy the format they would learn best in, so they buy the format they can actually attend, and the gap between those two things is absorbed silently by the student.
For anyone building online education, that gap is the whole design problem. Self-paced is what the market buys and mentor feedback is what the market says it needs, which means a self-paced product without live human contact is selling people the schedule they need and withholding the support they told you they need. It is also, plausibly, part of why completion is such a widespread problem in this industry: the format most people can afford in time is the format that gives them least of what they said keeps them going.
Finding 06
Money filters who gets in, quietly.
Number of separate payments recorded per student, among the 456 of 553 with at least one payment.
- One payment353 · 77%
- Two payments47 · 10%
- Three payments32 · 7%
- Four or more24 · 5%
77% settled in a single payment and 23% spread it over two or more, with 12% taking three or more instalments. Of the 170 students with a payment type recorded, 59% are marked as full payment.
This is the number that should be read against our other dataset, and it is uncomfortable. Among 325 people applying to retrain, 47% said they could only take free resources and roughly three-quarters said they could not pay a lump sum. Among the people who actually enrolled, most paid a lump sum.
The two facts are not in conflict; they describe a filter. The population that wants to retrain and the population that manages to is separated substantially by the ability to produce money at once, and the difference does not show up anywhere in a school enrolment table because the people it excludes never appear in one. Every dataset of students is, by construction, a dataset of people who could pay.
We are publishing it because it is the most useful figure here for anyone outside our commercial interest. If you are designing an employer-funded scheme or a public retraining policy, the constraint to attack is not motivation and it is not awareness. It is that the modal interested adult cannot produce the money in one go, and the market has quietly organised itself around the minority who can.
Finding 07
Coming from everywhere, heading for AI.
Current or most recent occupation and stated areas of interest (n=62 each); end goals, multi-select (n=113).
- A job outside technology42 · 68%
- Already in a technical role11 · 18%
- Currently studying9 · 15%
Technology areas they name
- AI45 · 73%
- Full stack30 · 48%
- Automation25 · 40%
- Data18 · 29%
- Backend10 · 16%
- Frontend8 · 13%
- Cybersecurity5 · 8%
What they want out of it
- A stable, full-time job70 · 62%
- Working remotely56 · 50%
- Freelancing or client projects38 · 34%
- Building their own startup28 · 25%
- Improving a current non-tech role17 · 15%
- Managing developers better9 · 8%
- Not sure yet9 · 8%
68% came from a job outside technology. The recorded occupations are not a narrow band: sales and admin executives, HR and compliance officers, mechanical and geophysical engineers, a tax consultant, an editor, a copywriter, a medical intern, a property agent, a barista, a driver, an airline ground operations manager. 15% were students and 18% already held a technical role and wanted a different one.
What they are aiming at is far more concentrated than where they came from. 73% name AI as an area they want to work in, ahead of full stack at 48% and automation at 40%. Frontend, the traditional entry point into this industry, is named by 13%.
On goals, 62% want a stable full-time job, but 50% explicitly name working remotely and a third want freelance or client work. Half of this population is not trying to join a local company at all; they are trying to reach a market that is not geographically bounded, which is a materially different ambition from the one most national reskilling programmes are designed around.
Limitations
What this data cannot tell you.
The two headline charts rest on 62 people. That is a real sample and it is a small one, and no amount of interesting structure in it changes that.
The survey is voluntary and recent. Programme format is recorded for 520 of 553 students and payments for 456, because those are transactional and cannot be skipped. The onboarding survey can be. 113 people answered the background questions and 62 answered the confidence and learning-style questions, which are the most recently added. Those 62 are therefore weighted towards students who enrolled recently, and nothing here should be read as describing the full five-year intake.
Self-selection runs through everything. These are people who chose one school, in one market, and paid for it. They are not a sample of Malaysians, of career switchers, or even of people who considered retraining. The comparison to our applicant data in Finding 06 is the most exposed claim on the page, because the two datasets come from different systems and overlapping but not identical periods.
Coded free text is a judgement. The fear categories in Finding 02 were assigned by hand, one primary category per answer, and several answers named two things. A different coder would move a handful of records. The gap between 39% and the next category at 13% is wide enough to survive that; a finding that rested on a five-point difference would not be.
We did not publish enrolment seasonality. The cohort join-date field looks like it shows a dramatic December peak. It does not: only 84 distinct dates exist across 507 records and one date alone carries 119, which is the signature of fixed cohort starts and a bulk data migration. The chart was tempting and the finding would have been false.
There are no outcomes here at all. No placement rate, no time to hire, no employers, no salaries. Those claims require a standard of evidence that internal CRM records cannot meet, and a school reporting on its own graduates has an obvious incentive problem. We would rather publish a narrower report that holds.
Finally, the obvious one: we sell the training this report describes. That is a conflict worth naming on the page rather than in a footnote, and it is why every denominator sits next to its numerator.
Methodology
How this was produced.
Source.The enrolled-students table from Sigmaschool’s internal systems, exported on 26 August 2026: 553 records, each one person who enrolled. Coverage by field: programme format 520, at least one recorded payment 456, payment type 170, English proficiency 143, age 130, qualification, field of study, prior experience and end goals 113 each, and the confidence, commitment, fear, learning-preference and occupation fields 62 each.
Instalment counting. The number of payments per student was derived by counting how many of the thirteen sequential payment columns hold a positive value, not from any declared payment plan. 456 students have at least one. Payment amounts, currencies and processors were read during the analysis and are not published.
Free-text coding.The 62 answers to “what is one thing you are most nervous or uncertain about?” were each assigned exactly one primary category, so the rows sum to 62 with no double counting. Where an answer named two concerns, the category is the one named first and most concretely. Occupation and areas of interest were coded the same way; the interest figures are multi-select and sum above 100%.
A time series we tested and dropped. First payment dates are clean - 343 distinct dates across 449 records, with no date carrying more than four - so programme format by year of first payment can be computed. We are not publishing it as a demand trend, because the menu of programmes on sale changed substantially across those years, and a chart of what people bought would be read as a chart of what people wanted. The two are not the same thing when the shelf keeps changing.
Privacy. No names, contact details, employers, addresses or individual records are published, and no verbatim free-text answers are quoted anywhere on this page. Several of the original answers disclose a disability, an employer or an industry specific enough to identify the person, so the report describes categories only. Every figure is an aggregate over the stated n.
Corrections and licence. If you spot an error, write to support@sigmaschool.co and a dated correction will be published here. Figures are released under CC BY 4.0 . Quote them, chart them, republish them - including commercially - with attribution and a link. We ask only that the sample sizes travel with the numbers.
Questions
Frequently asked.
What is the confidence gap?
The distance between how confident people say they are on a rating scale and what they say they are afraid of when asked in their own words. In this dataset, 50% of enrolled students rate their confidence of completing the programme at 10 out of 10 and none rate it below 7, yet 39% of the same people name their own ability to learn as the single thing they are most nervous about. The scale reports near-total confidence; the free text reports widespread self-doubt.
Are these applicants or paying students?
Paying students. Every record in this report is someone who enrolled, and 456 of the 553 have at least one payment recorded against them. That is the point of the report: these are people who have already cleared the money barrier and made a decision, so their remaining fears cannot be dismissed as the hesitation of people who were never going to start.
How many students had never coded before?
56% recorded no coding experience whatsoever at the point of enrolling, and 87% were at level 0 or level 1, which means no experience or beginner tutorials only (n=113 of 553 who completed that onboarding field). Only one person in the sample described themselves as a working developer.
Does this report contain placement or salary data?
No. There are no placement rates, no time-to-hire figures, no employer names and no salary outcomes anywhere in it. This is a study of who enrols, what they believe about themselves and what they buy. Outcome claims need a different standard of evidence than we can meet from these records, so we do not make any.
Why are the sample sizes different for different figures?
Because different fields are filled in at different stages. Programme format is recorded for 520 of the 553 students and payments for 456, because those are transactional. The onboarding survey is voluntary, so the background questions were answered by 113 people and the confidence and learning-style questions by 62. Every figure on the page carries the exact number it was calculated from, and none is presented as representing more people than answered it.
Can I cite or republish these figures?
Yes. Everything here is published under CC BY 4.0, so you can quote, chart or republish the figures, including commercially, with attribution and a link back. We ask only that the sample size travels with the number, since several of them are small.
The doubt is normal. It is also testable.
39% of people who paid were afraid they couldn't learn it.
The cheapest way to find out which side of that you are on is to build something. Six real projects, a live instructor session and the full programme e-booklet - one signup, no card.