Support Tiers and RTI Decisions
Not every child who is struggling needs the same kind of help, and the help a child needs can change over time. Some children make progress with good classroom teaching; others need a targeted small-group focus on specific skills; a smaller number need more intensive, individualized support. Amal organizes this idea into three levels of support and manages the decisions around moving a child between them through a structured, teacher-controlled process.
The three levels of support
Every child has a current support level. The three levels reflect increasing intensity of focus and delivery.
The first level is core classroom instruction. This is the base: quality teaching in the regular classroom, drawing on the child’s assigned plan as part of normal class activity.
The second level is targeted small-group support. The child receives more focused instruction on specific skills, typically in a group of peers who share similar needs, in addition to classroom instruction.
The third level is intensive individualized support. The child receives a high-intensity, individually tailored focus, with involvement from a specialist, in addition to the other levels.
These levels are internal to the platform and to the adults working with the child. They never appear in any communication to the child or to the parent. A child does not know which level they are on, and the platform never exposes this information in a student-facing or parent-facing context.
How the platform raises a concern
Before a teacher considers changing a child’s support level, the platform raises a signal. It evaluates the full picture for each child: the skill-status results, the reading-area summaries, the profile, the monitoring verdicts, and the fluency trend. From this, it produces one of three alert types.
A skill alert flags one or more skills showing a pattern worth the teacher’s attention. The concern is specific and does not yet suggest a broader problem.
A student alert flags a pattern broad enough to concern the child’s overall reading trajectory. The teacher may consider nominating a formal support-level review.
An immediate-review alert signals a pattern severe enough to require prompt attention. This includes situations where the monitoring layer detected a sharp performance drop across two comparable sessions.
One alert can be open at a time for a student. A more serious alert automatically closes a less serious one when it opens. All alert history is kept permanently; nothing is deleted.
The sentinel: when no alert is raised
Before any alert logic runs, the platform checks a set of sentinel conditions. If any of these conditions is true, the platform raises no alert and blocks any movement in the child’s support level:
- The skill-status engine returned a “not enough to decide yet” result for the child’s evidence
- A reading area shows only partial or not-yet-confirmed status
- The child’s profile is the system guard for incomplete data
- The monitoring layer has a halt verdict
- Fluency checking has fewer than the minimum number of comparable readings
The sentinel cannot be bypassed. No role and no action unlocks it when the underlying condition exists. This protects children from being moved to a higher support level based on incomplete evidence that only appears to show a problem.
The fidelity gate: plan not delivered is not plan not working
The most important distinction in any support-tier decision is between two situations that can look identical in the data but mean completely different things: a plan that was delivered consistently and did not produce the expected response, and a plan that was rarely or inconsistently delivered and therefore never had a real chance to work.
The platform requires confirmation of delivery before escalation to a higher support level is permitted. This confirmation is the fidelity gate. The threshold is approximately 80 percent of planned sessions completed.
A child who has been receiving targeted support in a small group, with the teacher running 85 percent of planned sessions, and who still shows no progress in monitoring, is in a different situation from a child whose teacher ran only 40 percent of planned sessions. In the first case, the data supports the conclusion that the plan was tried and did not produce sufficient growth. In the second case, the data cannot support that conclusion, because the plan was not really tried.
The fidelity gate runs before the escalation criteria are even evaluated. If the delivery rate does not meet the threshold, escalation is blocked regardless of what the monitoring data shows. The gate does not distinguish between a well-intentioned teacher who faced real constraints and any other situation; it simply holds that a plan must be genuinely delivered before its results can be interpreted as evidence of the plan’s effectiveness.
The movement rules
Four rules govern every proposed movement between support levels. All four are evaluated for any proposed change.
The escalation rule defines the criteria for moving a child to a higher support level. Escalation requires that the fidelity gate has been passed first.
The maintenance rule defines the criteria for keeping a child at their current level when the evidence suggests they might be ready to move.
The de-escalation rule defines the criteria for moving a child to a lower, less intensive support level when their growth supports a reduction.
The block rule is the fidelity gate. It runs first and, if it fires, prevents escalation from proceeding regardless of the other criteria.
The movement rules are deterministic: the same evidence and the same rule version always produce the same recommendation. The recommendation is produced by the platform; the decision is made by the teacher.
The specialist gate for the highest level
Moving a child to the most intensive support level requires approval from a specialist, not from a teacher or a standard administrator acting alone. In the current phase, the specialist is a designated educator who has been seated in that role: the platform recognizes them through an explicit specialist designation on their account. A standard administrator without that designation cannot approve the movement, and neither can a teacher who has not been seated as the specialist. The check runs automatically as part of the approval flow, before any decision is applied.
The tier-review process
Moving a child’s support level follows a three-step sequence.
First, a formal support-level review is opened for the child, linking the evidence assembled to support the case. This is the point at which the evidence record is committed to supporting a specific proposed decision.
Second, the platform computes its recommendation. It applies the movement rules to the linked evidence and produces a proposed direction (escalate, maintain, or de-escalate) with its reasoning. This is read-only: the platform is offering an analysis, not making a commitment.
Third, the teacher, or for the highest level a specialist, approves or rejects the proposed movement. Approval triggers the single write path that makes the change. Rejection closes the review with no change.
Every approval writes a permanent record of the decision: who approved it, when, against which evidence, and which movement rules applied. A tier change without a corresponding record is structurally impossible. This is not a log added for compliance; it is the architecture. The change and the record are written in one atomic operation, and if the record cannot be written, the change does not happen.
How the fluency trend reaches a support-level decision
The fluency-check layer reports a trend for each child: whether recent readings are running above, below, or mixed against the expected growth path, or whether there are too few comparable readings to say. That trend is now wired into the support-tier layer, but it never moves a child’s level on its own.
A trend that is running below the expected path is read as a “not progressing” signal. It can raise a student-level alert for the teacher to look at, alongside the rest of the evidence. A trend with too few comparable readings is treated as a reason to wait, not a reason to act: it holds rather than raising a false alarm. A trend that is above the path, or one that is genuinely mixed, raises nothing. The mixed case is handled conservatively on purpose, because a mixed picture is ambiguous rather than a clear sign of trouble.
In every case the trend feeds an alert, and the alert is something the teacher reviews. The fidelity gate and the tier-review process still stand between any signal and an actual change in a child’s support level.
Where this module stands
The support-tier engine and the layers it depends on are built and working. The fidelity tracker that confirms whether a plan was actually delivered, and the evidence layer that assembles the dossier behind a review, are both in place and feeding the engine their real data. The fail-closed default that holds an escalation now applies only where the data is genuinely missing for a particular child, not as a blanket hold standing in for an unbuilt part. For a child with a real delivery record, the gate reflects that record.
This reflects the safety model of the whole platform: when in doubt, hold. The engine never guesses, and it never promotes a child to a higher level of support on incomplete information.
What teachers and parents see
Teachers see alerts and, when a formal review is open, the recommendation and the evidence summary. Teachers make every decision: accepting the recommendation, requesting more information, or rejecting it.
Parents receive updates about their child’s learning progress in plain growth language. They never see support level labels or anything that would allow them to infer a comparative or categorical judgment about their child. Children never see their support level at all, at any point, in any view.