MTSS and Behavior Intervention Software: Why Tier 2 Fidelity and Discipline Data Never Line Up

Why multi tiered support collapses somewhere between the building and the district office

An assistant principal at a middle school enters an office discipline referral at 2:40pm. He has three students in the outer office and a bus duty in twelve minutes. The form asks for the problem behaviour, the location, the possible motivation, the perceived antecedent, others involved and the administrative decision. He picks the first plausible option in each dropdown and writes disrespect in the notes. That record is now evidence in three separate processes he is not thinking about: the school's tier two identification, the district's disproportionality profile, and potentially a federal civil rights data submission.

Meanwhile the reading interventionist is running a tier two group four days a week. She has the group, the curriculum and the progress monitoring probes. Nobody has ever observed her sessions against a fidelity checklist, so when three of her six students fail to respond, there is no way to distinguish a student who needs tier three from an intervention that was never delivered as designed. That distinction is the entire logic of a tiered system, and most districts cannot make it.

The tools are not the problem in the way people assume. SWIS by PBIS Apps does discipline referral data and its triangle reports better than almost anything, at a price that is hard to argue with, and it stays deliberately narrow. Branching Minds is the most complete packaged multi tiered system of support product and does intervention matching genuinely well. Kickboard is strong on behaviour points and school culture practices. Panorama is strong on surveys and student success analytics. Review360 sits on the behaviour intervention side. Each is good at its slice. The district's problem is that a multi tiered system of support is not a slice, it is a set of joins between academic screening, intervention delivery, attendance, discipline, and special education referral, and every one of those lives in a different system with a different identifier and a different refresh schedule.

Problem 1: your intervention catalogue is yours, and the tool ships someone else's

Districts spend years deciding what tier two looks like. Which programmes, at what dosage, in what group size, delivered by whom, with what entry and exit criteria, and what happens after eight weeks of non-response. That work is a district's professional judgement, informed by what it can staff and afford, and it is revised every summer.

Packaged tools handle this with a library plus custom entries. That sounds sufficient and is not, because the important part is not the list of interventions, it is the decision rules attached to them. If a student on a specific intervention has four consecutive data points below the aimline, what does your district say happens, and who is notified, and how long do they have. If your rule is a rule, the system should enforce it. If it is a suggestion, the system should stop pretending otherwise. Most tools store the intervention and leave the decision to a meeting that may or may not happen, which is why tier two groups quietly become permanent placements for the same students year after year.

A custom build encodes the catalogue and the rules together, with entry criteria that can pull directly from your screening data, dosage that generates the schedule, and exit and escalation rules that fire automatically into a named person's queue. When the district revises the catalogue in July, the rules change with it, and last year's decisions remain reproducible against last year's rules.

Problem 2: fidelity is the measurement that makes the rest mean anything

Here is the uncomfortable part. Without a fidelity record, a non-responder is uninterpretable. You do not know whether the child needs more, or whether the intervention happened three times instead of four, or whether it was delivered by a paraprofessional who never received the training. If you refer that child for special education evaluation, your file says you provided intervention and the file cannot show it was provided as designed. That is a weak position in an eligibility meeting and a weaker one in a dispute.

Fidelity fails for a practical reason, not a philosophical one: capturing it costs the interventionist time she does not have, and observation costs an administrator time he does not have. So it is collected in September when the coach is enthusiastic and never again.

Design around that constraint. Session logging should take under fifteen seconds on a phone: who attended, whether the session ran as planned, one tap for the two or three components that matter most in that intervention. Observation should be a short structured form that a coach completes on a tablet during a walkthrough, scheduled automatically so that every interventionist is observed on a rotation rather than whoever the coach remembers. Then the reporting that matters becomes possible: this intervention is delivering at 62 percent of planned sessions in this building, which is why nobody is responding, and the answer is scheduling, not more referrals.

Problem 3: discipline referrals are entered to end an incident, not to inform one

Every field on a referral form is a tax on an administrator who is managing a live situation. The result is systematic: the fastest options get chosen, free text carries the real information, and the structured data that feeds every downstream analysis is unreliable in a way nobody quantifies.

The fix is not more training. It is a form that takes seconds, defaults intelligently from context such as time, location and the entering user, and asks for antecedent and function only where those fields will actually be used. Then a language model can do one narrowly useful job: read the free text narrative and propose the structured codes, which the administrator confirms or corrects in one tap. That keeps the human accountable for the record while capturing the detail that would otherwise be lost. It is a far better use of a model than the summary dashboards vendors like to demonstrate.

The second fix is feedback. Administrators enter referrals into a void. Show a building leader their own data weekly, by location and time of day, and you get the classic PBIS insight that the cafeteria at 11:40 is a supervision decision rather than a student behaviour problem.

Problem 4: disproportionality is arithmetic you should run monthly, not discover annually

Significant disproportionality under IDEA carries a specific and expensive consequence: an identified district must reserve 15 percent of its IDEA Part B funds for comprehensive coordinated early intervening services, and it must review and revise policies, practices and procedures. Discipline data by race and disability also flows to federal civil rights data collection. Both of those arrive as a determination based on data you supplied months earlier.

Risk ratios are not complicated arithmetic. They are a comparison of the rate for a group against a comparison group, subject to your state's threshold and minimum cell size rules. Every district could compute them monthly. Almost none do, because the data is spread across the student information system, the special education system and the discipline module, and the analyst who could join them has a submission deadline every month for something else.

Build the join once and run the ratios continuously, at district level and at building level, with the drill down to the referrals and decisions behind the number. The value is not compliance reporting. It is that a building principal in November can see that suspensions for one subgroup are concentrated in two classrooms and three periods, which is an actionable fact, rather than receiving a district level ratio in July that tells him only that a problem exists somewhere. Pair this with your intervention data and you can ask the harder question, which is whether students in one subgroup are being referred to discipline where others are referred to support.

Problem 5: the intervention record is your Child Find defence in both directions

Multi tiered support sits between two legal risks that pull in opposite directions. Referring too readily contributes to disproportionate identification. Referring too slowly is a Child Find problem, and a delay while a district cycles a struggling student through tiers is exactly the pattern that produces a complaint. The evidence that you got the balance right for a specific child is the intervention record: what was tried, at what dosage, delivered with what fidelity, with what data, and what decision was made and when.

That record has to be assembled from the intervention system, the assessment vendor, attendance, discipline and the referral, and today it is assembled by a psychologist the week before an eligibility meeting. A build that produces that packet in one click, complete and time stamped, is worth the project on its own for a district of any size, and it is the artefact that makes both risks manageable rather than only one.

What this costs and how long it takes

From Digital Heroes delivery experience, a first release covering your intervention catalogue with enforced decision rules, progress monitoring, fast fidelity capture and a twenty second referral form runs $60,000 to $130,000 in 12 to 16 weeks. The full platform adding assessment vendor integrations, attendance and special education joins, behaviour intervention plan workflow, monthly risk ratio reporting and the eligibility packet builder runs $150,000 to $350,000 phased over 6 to 12 months.

What moves the number: the number of assessment vendors you must pull from, because each screening and progress monitoring provider has its own interface and export cadence. Whether your student information system offers a modern interface or a nightly file drop, which is a several week difference. Whether you need building level configuration, since a district that lets each school define its own tier two model is asking for a configuration engine rather than a fixed catalogue. And staffing reality, because the fastest projects are the ones where a coordinator with authority over the catalogue is available weekly.

Build versus buy, and when buying is right

Buy if you are a single school or a small district. SWIS costs very little and does discipline data well, and Branching Minds is a credible packaged multi tiered support product if your model is reasonably conventional. Adding a custom system at that scale is spending on software what should be spent on an interventionist.

Build when two or more of these are true. Your intervention catalogue and decision rules are genuinely district designed and your tool cannot express them. You cannot produce a fidelity figure for any intervention, which means your non-responder data is not interpretable. Your disproportionality answer takes weeks to assemble and arrives after the state's. Your special education referral packets are built by hand from five systems. You have building leaders who would act on weekly data and currently receive none. The threshold is not district size alone. It is the number of systems standing between a question and its answer.

How to choose a developer for MTSS software

Ask them how they would model a decision rule, not an intervention. If their sketch has interventions and students but no rule object with a trigger, a threshold and an owner, they have designed a logging tool and your tiers will stay decorative.

Ask how long their referral form takes to complete for an administrator with three students waiting. If the answer is not measured in seconds, the data quality problem you have now will survive the project intact.

Ask what they have integrated by name: your student information system, your screening vendor, your special education system. Ask about identifier matching across those systems, because that is where the project actually gets hard.

Then ask who owns the code, in writing, before kickoff. At Digital Heroes the district owns the repository and the infrastructure accounts from the first commit, and given that this system holds behaviour and intervention records on identifiable children, you should also require documented access logging, retention rules and an exit plan that returns your data in a usable form.