Mark Desmaele
Founder and Chairman at TDSO
Low-income education systems have no shortage of teacher-training programmes. Almost all promise to train teachers, and most report impressive numbers: thousands trained, hundreds of schools reached, a whole province or country covered. Yet the learning outcomes in many of these same systems barely move. Children still leave primary school unable to read a simple text or handle basic arithmetic, years and grants later.
Why is it that so much teacher training produces so little change in the classroom, and what separates a programme that still matters in five years from one that evaporates the moment the grant ends? The research on this is reasonably settled, and it runs against the instinct that the best programme is the one reporting the largest number of teachers trained.
The most common way of training teachers at scale in low-income systems is the cascade model. A small group is trained centrally; they return to their regions and train others; those others train teachers in their districts; and the process repeats down through several tiers until, in theory, the training reaches every classroom in the country. In the research literature, the cascade is itself a form of “training of trainers”: it works by turning each level into the trainer of the level below.
Its appeal is genuine: the cascade is cheap: one central event seeds a national rollout. It is fast: a system can claim to have reached tens of thousands of teachers within a single budget year. It is administratively simple; it fits the reporting cycles ministries and donors work to, and it answers real political pressure to show wide coverage quickly. For a ministry with a small budget and a large workforce, it can look like the only affordable option. None of this is foolish. The difficulty is that every one of these strengths is about spread, and none is about whether anything changes in the classroom.
Not all cascades pour the same thing. The critical question is what flows down the chain, and how well each person is prepared to receive it.
In the weak and most common version, what flows down is a message about teaching, and the people asked to pass it on are ordinary teachers who were prepared to teach children, not to train adults. A selected group attends a workshop, then returns to train colleagues, who in turn train others. The trainer role is bolted onto teachers who never received preparation for it. The research is blunt about the consequence: reviews identify “insufficient trainer preparation” as a primary cause of failure, alongside the absence of any follow-up in the classroom. Teachers who “receive training in order to train others” transmit the content poorly, because passing on a summary of a method is not the same as being able to deliver the method (Rembe, 2006, as cited in the whole-schooling curriculum review, 2023). At each tier, the material is diluted; by the time it reaches a classroom, it is a faint echo of the original.
Case studies from very different systems report the same pattern. Reviews of cascade programmes in Kenya, Nepal, Cameroon, Malaysia and Botswana describe a long, one-way chain that distorts the message and barely changes what teachers do. A study of the Kenyan system put the causal link plainly: the multi-tier cascade “contributed to dilution of content being covered, hence leading to the ineffectiveness of the entire programme” (Cogent Education, 2016). There is a further irony the research notes. Because each tier is instructed rather than helped to practise, told what to do rather than shown and coached, the cascade fails on its own terms: the method by which it trains is the very method it is usually trying to train teachers out of. It preaches active, practice-based teaching through passive, one-way telling.
One review of non-formal education gave the mechanism its sharpest name. It described the approach as “spray and pray”: spray information at as many people as possible, and pray that something sticks. The characterisation is not new. In 1999 the science educator Jane Butler Kahle, testifying to the US Congress, described short-term, standardised sessions of this kind as “make and take” or “spray and pray” workshops, and drew the distinction exactly: they can work to deliver information or a discrete skill, but they usually do not improve teachers’ content knowledge, their classroom practice, or their pupils’ learning. That is the whole difficulty in one sentence. A cascade of this kind is, at its bottom, a system for distributing information, and it is built to do that well: it moves a message quickly and widely, and counts the people it reaches. But distributing information does not change teaching. Knowing about a method is not the same as being able to use it. Teachers do not teach differently because they were told about a better way; they change when they practise it, in their own classroom, with someone watching and helping them get it right, again and again until it holds.
This approach also persists for a reason that has nothing to do with pedagogy. A large share of teacher “training” comes from a commercial market: publishers, software vendors, programme providers, whose success is measured in books sold, licences renewed, and users registered. For a provider on that footing, reach is the product: the incentive is to place a resource in as many hands as possible, and whether teachers’ practice or pupils’ learning actually changes is, at best, a secondary concern. The result is what has been called the commodification of instruction, in which teachers are trained to use a product rather than helped to develop their own judgment, and “good teaching” quietly comes to mean good consumption of what is sold (Hibbert & Iannacci, 2005). Dissemination, in this sense, is not a neutral accident of the cascade; for much of the sector, it is the business model.
This distinction underpins the whole question, and everything else in this paper follows from it: the weak cascade optimises for the spread of information; what changes a classroom is ongoing guided practice.
The weak cascade also produces reach figures far larger than the change delivered: the reported figure and the delivered change can be very different things.
A typical headline reads: “this programme will reach 30,000 teachers.” Look closely, and the number is usually two very different quantities added together. A small first figure, the teachers trained directly, perhaps a few hundred, is real. A much larger second figure, the teachers those trainers are assumed to reach as they pass the training on, is not a result but a projection, resting on the assumption that each trained teacher will train several colleagues, that those colleagues will absorb and apply what they receive, and that all of this will happen without funding, preparation or follow-up. And that assumption usually fails.
A documented case makes the arithmetic. In a cascade run through the Open University of Greece, twelve first-phase facilitators were each asked to train twenty people, producing roughly 250 trainers, who were then recorded as having reached some 8,200 people (Karalis, 2016, as reported in Ndongfack, 2020). The headline figure for such a programme is 8,200. The number actually prepared with any depth by the original experts is twelve. Everything between the twelve and the 8,200 depends on transfer surviving several further relays, each unfunded and unobserved, which is precisely what the evidence says does not reliably happen. The large number is not a lie; it is a projection wearing the clothes of a result: the reported number adds something measured to something merely hoped for, and presents the sum as delivered. From the headline alone, there is no way to tell how much is real. The projections are often made in good faith rather than to mislead, but the effect is the same: the single most-quoted figure in teacher-training proposals, the number reached, is frequently the least reliable.
Guided practice is what works.
This is where decades of research converge. Training that distributes information without sustained, in-classroom follow-up rarely changes teacher behaviour or student outcomes. One-off workshops, however well designed and however many attend, are largely ineffective, because changing how someone teaches takes time, repeated practice, and feedback on their own attempts in their own classroom. The Learning Policy Institute’s review of thirty-five rigorous studies concludes that the development which changes teaching and learning is embedded in classroom practice and typically involves coaching or professional learning communities. The same body of work, associated above all with Linda Darling-Hammond, draws the contrast in plain terms: development that changes practice is sustained, collaborative, tied to the teacher’s actual work, and built on observation, modelling, coaching and feedback, everything a one-off workshop is not. A cascade of the weak kind is the structural opposite: a one-way broadcast that ends at the moment of delivery, with nothing at the point, the individual classroom, where change would have to take root.
The strongest single piece of evidence on what works is a 2018 meta-analysis by Kraft, Blazar and Hogan, pooling sixty causal studies of teacher coaching. It found substantial effects: coaching improved teaching practice by 0.49 standard deviations and student achievement by 0.18, large results by the standards of education research. Individualised, classroom-based coaching, observing a teacher, giving specific feedback, supporting deliberate practice of particular skills, genuinely changes teaching and learning.
But the same analysis yields the most important finding for anyone scaling up: effects shrink sharply as programmes grow. Average effects from large programmes are only a fraction of those in small, well-run ones. This is the crux. The intervention that works is intensive and human, and a thin, cheap cascade cannot preserve it as it spreads. Scale does not fail because the effective method is weak; it fails because the effective method is expensive to keep intact, while the cheap method that scales easily is the one that does not work.
The effective approach is not the cheapest, at least not at first. Building real capability, with preparation, classroom practice, coaching and measurement, costs more per teacher in the early years than spraying a message down a cascade. The evidence suggests deep change at cascade prices is not available.
The cost argument is different, and stronger, over time. A cascade spends a modest sum and leaves little behind; when the grant ends, the capability ends with it, and the next programme starts again from near zero. Capability built into a permanent institution costs more to establish but keeps delivering after the funder has gone, at no further cost, because running it becomes part of the institution’s ordinary work. Measured per teacher who actually changes practice, and across the years the capability keeps working, the more expensive approach is very often the cheaper one. The right question is not “what does it cost to train a teacher?” but “what does it cost to leave behind a system that keeps improving teaching after we leave?”
A few features consistently mark out the programmes that leave durable change from those that report large numbers and little else.
The first is trainer preparation. In durable programmes, the people who train teachers prepare for the trainer role over time and in the classroom; in the others, that role is simply added to teachers after a workshop.
The second is the number of relay layers. Each time the capability is copied between the trainer and the classroom, something is lost. The durable programmes keep that number small; the cascade multiplies it.
The third is follow-up. Sustained support in the actual classroom after the initial training is the feature the evidence associates most strongly with change. Without it, a workshop becomes the whole intervention.
The fourth is the reach figure itself. In the durable programmes, the number reported is teachers actually trained and supported; in the others, it includes teachers merely projected to be reached through onward, unfunded relay.
The fifth is where the capability lives when the grant ends. In durable programmes, it sits in a permanent institution that keeps delivering; in the others, it sits in individuals who disperse, and the capability disperses with them.
How TDSO applies this
TDSO’s model is built directly on this evidence, and answers each of those questions by design. It rests on three features that together separate it from the weak cascade: how trainers are prepared, how few times capability is transferred, and where that capability finally lives.
TDSO prepares teacher trainers through two combined channels: dedicated workshops and a three-year graduated apprenticeship in real classrooms. In the first year, the trainer observes TDSO trainers delivering the model with real pupils; in the second, they co-deliver alongside TDSO; in the third, they deliver themselves, coached by TDSO. The preparation is progressive, supervised, and located where the model has to work (in the classroom). This directly answers the failure the literature names as “insufficient trainer preparation”, and the coaching across years is itself the sustained follow-up whose absence the research identifies as fatal.
Secondly, where the weak cascade relays a single undifferentiated “trainer” role down through anonymous tiers, TDSO keeps two roles deliberately distinct. TDSO prepares the teachers and the teacher trainers separately. These are not the same capability copied down one tier; they are different, deliberately prepared roles. Distinguishing them is precisely what stops the model from becoming the diluting chain the research condemns.
The third feature makes the rest durable. The evidence warns that capability lodged in individuals does not hold, and a capability held by a visiting organisation or a temporary project office can never make that final shift, because the organisation is, by definition, external and temporary. Only a permanent body already inside the system can. This is why the home for the capability matters as much as the capability itself: a government teacher-education institution carries what a project cannot, legitimacy over the profession, authority over the careers and credits that give training consequence, and a mandate that outlasts any grant. TDSO therefore places capability inside institutions that persist and confer exactly these things. The teacher-trainer role is built into the TEIs; the higher trainer-of-trainers role is built into the TEIs and the government department, so the system can prepare its own trainers after TDSO has gone. Institutionalisation turns a successful programme into a permanent national capability, and it is why the model is designed to end: the funder pays for a transition into public ownership, not for a dependency that must be funded forever.
Teacher training fails to reach the classroom when it optimises for spread rather than change. Spraying information down a cascade reliably does one thing: moves a message to a large number of people and reliably fails at another: changing how any of them teach. Add to that a trainer role bolted onto teachers who were never prepared for it, a headline that counts exposure as impact, and a design that leaves nothing behind when the grant ends, and you get a full report and an unchanged classroom. It succeeds when it prepares trainers properly and in the classroom, keeps the transfer to a single step, and lodges the capability, including the capacity to prepare the next trainers, inside institutions that will keep it running after the funder has gone. That approach costs more at the start and less over time, produces smaller numbers that hold up, and leaves a stronger system rather than a fuller report. The distinction that counts is whether a programme will still matter in five years.
Sources
Kahle, J. B. (1999). Teacher Professional Development: Does It Make a Difference in Student Learning? Testimony to the US House of Representatives Committee on Science. Origin of the “make and take” / “spray and pray” characterisation; distinguishes delivering information from changing content knowledge, practice and pupil learning.
Darling-Hammond, L., Hyler, M. E., & Gardner, M. (2017). Effective Teacher Professional Development. Palo Alto: Learning Policy Institute. Characterises effective development as sustained, collaborative, connected to teachers’ work, and built on observation, modelling, coaching and feedback.
Hibbert, K., & Iannacci, L. (2005). From Dissemination to Discernment: The Commodification of Literacy Instruction and the Fostering of “Good Teacher Consumerism”. The Reading Teacher, 58(8), 716–727. On how packaged, product-led training substitutes consumption for the development of professional judgment.
Kraft, M. A., Blazar, D., & Hogan, D. (2018). The Effect of Teacher Coaching on Instruction and Achievement: A Meta-Analysis of the Causal Evidence. Review of Educational Research, 88(4), 547–588.
Popova, A., Evans, D. K., Breeding, M. E., & Arancibia, V. (2022). Teacher Professional Development around the World: The Gap between Evidence and Practice. World Bank Research Observer / policy research working paper. At-scale evidence from low- and middle-income countries on the PD characteristics linked to improved student learning.
Coburn, C. E. (2003). Rethinking Scale: Moving Beyond Numbers to Deep and Consequential Change. Educational Researcher, 32(6), 3–12. Frames reform in four dimensions: depth, sustainability, spread, and shift in reform ownership.
The cascade model of teachers’ continuing professional development in Kenya: A time for change? (2016). Cogent Education, 3(1). States that multi-tier dilution led to the ineffectiveness of the programme.
Ndongfack, M. N. (2020). Professional Development for Primary School Teachers in Cameroon: Is the Cascade PD Model Effective? (reports the Karalis, 2016, Open University of Greece worked example of cascade reach).
Sims, S., & Fletcher-Wood, H. (2021). Identifying the characteristics of effective teacher professional development: a critical review. School Effectiveness and School Improvement, 32(1), 47–63.
On the “adapted cascade” with prolonged, expert-supported trainer preparation: adapted cascade study for a primary-school digital-education programme (Education and Information Technologies, 2023).
On “insufficient trainer preparation” and dilution through tiers: qualitative studies of the cascade model in teacher professional development (e.g. Dichaba & Mokhele, 2012; Suzuki, 2011; Hayes, 2000).
On teachers trained “in order to train others” transmitting poorly: Rembe (2006), as cited in a 2023 whole-schooling curriculum review.
On trainers lacking “a professional framework or formal set of responsibilities to give legitimacy to their roles”, and the institutional role of teacher educators in implementing change: cited in the same curriculum-implementation literature.
On the “spray and pray” characterisation: Cascade Approach to Training: Theoretical Issues and Practical Applications in Non-Formal Education (2016).