Before you swap the phonics intervention, count the sessions
When only a minority of children clear a structured phonics intervention, check how it's being run before you blame the programme you bought. Delivery is cheaper to check and quicker to fix. Three things to look at before you go shopping: who runs the group and what training they had, how many sessions a week each child actually gets, and whether anyone can say where in the sequence a child stalls. A new programme resets all three to week one, and you lose a term proving what you already knew.
This assumes the catch-up dose is already running and underperforming. If the groups themselves are still being sorted out, that structure comes first. And if you're not yet sure which of your assessments told you these children were behind, that's a separate sorting job.
The question tends to land when screening data comes in and next term's groups get planned. Good timing, because two of the three checks can be fixed in a timetable draft.
Check one: who runs the group, and what training did they get?
This is where the evidence is sharpest, and it isn't a comment on anyone's ability.
The Education Endowment Foundation's toolkit splits teaching assistants into two deployments and gets two different answers. Teaching assistants running targeted, structured interventions with individuals or small groups: between four and six additional months' progress on average, listed at five months for that sub-approach. Teaching assistants deployed as general classroom support: zero months.
Same staff member. What changes is the structure around them.
Two caveats before you quote that in a staff meeting. The EEF rates the evidence moderate and says it rests on a limited number of studies, many of them not independently evaluated, so five months is a direction rather than a promise. And it's UK evidence, though the deployment question travels fine.
The EEF's guidance report on making best use of teaching assistants sets out what the effective version looks like. Sessions are brief, run 3 to 5 times a week, and are sustained over 8 to 20 weeks, with timetabling in place that lets them actually happen. Teaching assistants get 5 to 30 hours of training per intervention from experienced trainers or teachers. The intervention arrives with structured resources and lesson plans with clear objectives, and possibly a delivery script. Assessments pick the right children and track them. And explicit connections are made between the group work and classroom teaching. The report's own health check adds a question worth stealing: is there designated time for the teacher and the teaching assistant to talk?
One warning if you quote it. The report says sessions run 15 to 45 minutes in one place and 20 to 50 in another. Pick one and expect someone to find the other.
So the practical check is: has each person delivering this had initial training in this programme, and does anyone watch a session and give feedback afterwards? Induction alone doesn't hold. Sit in on two sessions run by two different adults and see whether the lesson looks the same.
Watch the error correction in particular, because that's what drifts first. Heubusch and Lloyd reviewed twenty-four studies of correcting errors during oral reading and found corrective feedback reliably improves word-reading accuracy, on word lists and in passages. What they recommended: correct the error as it happens rather than saving it for the end, have the child say the correct word, then reread the whole sentence. And match the technique to what the lesson is teaching, so supply the word when you're after fluent reading, prompt sounding out when you're building letter-sound knowledge. They didn't show immediate correction beats delayed correction, and later comparisons are still mixed on timing. What matters is that correction happens and that it looks the same in both adults' groups.
Check two: is the group getting the dose the programme assumes?
Start with the delivery guidance the programme publishes, because it's usually stricter than the timetable. Schools running MacqLit are told to run it at least 4 times a week in one-hour blocks, in a small group, with 20 minutes of daily one-to-one reading practice on top. MiniLit's published guidance is 4 to 5 sessions a week in groups of up to 4. The trial evidence behind both assumes that much, not less.
The EEF's common elements above land in the same place on frequency: 3 to 5 times a week, sustained over 8 to 20 weeks.
Evidence for Learning's Australian toolkit entry on small group tuition puts it at four months' additional progress over the course of a year, low cost, moderate evidence. It defines small group tuition as one teacher, trained teaching assistant or tutor working with 2 to 5 students. Its design guidance is a rule of thumb rather than a specification: the smaller the group the better, with a noticeable drop in effectiveness once groups go above 6 or 7, and frequent sessions, 3 times a week or so, of up to an hour over about 10 weeks showing the greatest impact.
Two sessions a week sits under all of that. And doubling the minutes on fewer days isn't a swap you can make freely, because the frequency is doing work of its own.
Then count what was delivered rather than what was scheduled. A group timetabled 4 times a week that lost sessions to assembly, swimming and a relief day is a two-a-week group wearing a four-a-week label. Mark a tally in the folder each session for a fortnight and you'll know.
Group size counts here too. Six children in a group built for four is a quieter under-dose: each child gets a smaller share of the turns and the feedback, in the same amount of time.
Check three: can anyone say where the child stalls?
Most schools track whether a child exited. Very few track where the child got stuck on the way.
Curriculum-based measurement, the approach most tiered systems are built on, was designed for frequent repeated administration, weekly or fortnightly, so you can see a trend rather than a verdict. Its oral reading fluency measures are reliable enough to bear that frequency. An end-of-block exit check gives you neither the trend nor the reason.
Once you have skill-level data that often, the diagnosis is usually quick.
Every child in the group stalls at the same skill. That's a delivery signal. The lesson has introduced too much at once, the modelling is thin, or the feedback loop isn't closing. Rosenshine's benchmark is that guided practice should run at around 80 per cent success, so error rates consistently over 20 per cent mean the steps are too big, not that the children can't do it.
Children stall at different points. That's a grouping signal, and regrouping is the normal, expected use of monitoring data. Best-practice guidance is to assess at least monthly, exit the children who've met the benchmark, and reshuffle the rest so groups stay as skill-similar as possible.
One caution on the measure itself. A running record or levelled benchmark lets a child lean on pictures and context, which can hide the exact gaps you're intervening on. A decodable word and non-word check names the missing grapheme-phoneme correspondences directly.
What does justify a switch
A few things genuinely do.
The programme isn't structured or systematic, and levelled texts, three-cueing or picture prompts sit at the centre of it. Or there's no published evidence base behind it at all. Or you've run it at full dose, with trained and observed delivery, with fortnightly skill data, for a complete block, and the trend lines are still flat for most of the group.
What doesn't justify it is a pass rate on its own. Exit data misleads in both directions: Reading Recovery reported strong first-year exit results for decades, and a follow-up of the US federal trial found those children were scoring below comparison peers by Grades 3 and 4 and were more likely to need further help. A number without a mechanism tells you very little about what changing the programme would do.
Head-to-head trials between well-designed structured programmes are rare, so a switch between two of them is a bet rather than an evidence-based move. A switch away from a programme that fails the tests above is a different matter.
When a group genuinely isn't enough
Some children will have had a real dose, faithful delivery and proper monitoring, and still not moved. That's the point at which the tiered model says escalate the intensity, not change the brand. Smaller groups, more frequent sessions, or one-to-one tuition, which the same Australian toolkit rates a little stronger than small group tuition. A lateral move to a different Tier 2 phonics programme buys none of that.
For older children carrying accumulated gaps, keep the explicit decoding work going, but expect it to take longer and show smaller gains, and pair it with vocabulary, fluency and comprehension teaching, which is what the US guidance for older struggling readers recommends. AERO and the Dyslexia-SPELD Foundation have published a multi-tiered system of support resource for reading, which is the Australian reference worth having on the table when you redraw the tiers for next term.
What the three checks need
A session tally, two lesson observations and a fortnightly skill check. None of it needs software.
Where the time goes is building the reteach lesson once you know the stall point, especially when a group has drifted a long way behind the class. Phonicspal builds lessons against the sounds a group has actually got, in whatever sequence your school follows, so a Tier 2 group and the whole-class lesson stay on the same ladder. You can see how that works or start free.
Regrouping for next term is a better lever than a new purchase order. Coach the delivery and put a skill check in the diary, and by the end of the block you'll know whether the programme was ever the problem.