How to switch school communication platforms without a gap
Migrations in this category rarely fail on the migration. They fail because one item with a fixed external lead time was started three weeks too late, and everything downstream compressed into the first week of school. This page is organised around lead times rather than around a week-by-week calendar, because the lead times are what you cannot negotiate.
| Workstream | Realistic lead time | Compressible? | Owner | What happens if it is late |
|---|---|---|---|---|
| 10DLC brand and campaign registration | 2–6 weeks, externally controlled | No | Vendor, with district data | SMS is silently filtered by carriers. The platform reports success; families receive nothing. This is the item that ruins migrations. |
| Contract notice to the incumbent | 30–90 days before term end | No | Business office | The contract auto-renews and you pay for a year you are not using |
| Export request to the incumbent | 1–4 weeks, vendor-dependent | Sometimes, with escalation | Technology | Everything downstream waits, and the sequence collapses into August |
| Export validation and normalisation | 3–10 working days | Yes, with tooling | Technology | Bad data lands in the new platform and is discovered by families |
| Roster feed and reconciliation | 1–2 weeks | Partially | Technology with SIS owner | Counts do not match the SIS and nobody trusts the audience |
| Consent and validation reconciliation | 1–3 weeks | No — it is judgement work | Technology with counsel | You send to numbers you cannot defend having sent to |
| Staff training | 2 sessions per site, plus materials | Partially | Communications and PD | Front office keeps using the old tool and you run two platforms by accident |
| Family notification sequence | 3 messages over 4 weeks | No — families act on the third | Communications | Support volume spikes in week one and the switch is remembered as chaotic |
| Parallel running | 2–4 weeks | Yes, but do not skip it | Technology and communications | The first real emergency is also the first real test |
| Decommission and archive | 2 weeks after cutover | Yes | Technology and records | Access is lost with records still inside, and the vendor is no longer motivated |
Two items on this list are externally controlled and cannot be accelerated by working harder: carrier registration for A2P text traffic, and the contractual notice window with your incumbent. Start both before anything else, including before the contract with the new vendor is signed if your procurement allows it. Everything else can be compressed under pressure; those two cannot.
Two windows, and which one you should want
The June-to-August clean break. Notice served in the spring, export taken after the last instructional day when the roster is final, configuration and training over the summer, families onboarded on the back-to-school attention spike when they are already opening things from school. This is the right answer in almost every case. The risk is that summer staffing is thin and a single absence on the technology team stops the critical path for a fortnight.
The December-to-January mid-year cutover. Sometimes forced by a renewal date or a vendor failure. It works, but three things get harder: the roster is live so the export is a moving target, family re-onboarding competes with nothing having anyone's attention in January, and staff learn new software in the term they are busiest. If you must do it, cut over immediately after a break rather than before one, and keep the incumbent live for two weeks past cutover regardless of what it costs.
What decides it is usually neither preference nor readiness. It is the notice window in your existing contract. Find that date first; it may make the decision for you, and a short bridging extension with the incumbent to land the cutover in summer is a normal procurement conversation rather than an admission of failure.
Consent carry-over, which is the part with legal weight
An inherited contact point is not automatically a contact point you may use. Work through each number with three questions.
- Where did it come from? A number a parent typed into a district form has a different provenance from one that arrived in a roster import and has never been confirmed by anyone. Both are usable for some purposes; they are not equivalent, and the difference is the first thing counsel will ask about.
- What consent state is recorded, separately from validation? These are two different facts and platforms routinely collapse them into one column. Validation is "does this number reach this household". Consent is "has this household agreed to receive this class of message". A migration that carries the first and loses the second has lost the part that matters.
- Does the traffic class change? Transactional messages about a specific event for a specific student sit differently from a scheduled behavioural campaign. If you are moving platforms and also broadening what you send, those are two decisions and the second needs its own approval.
Practical rule: carry forward everything, but carry forward the state as well as the number, and do not let a migration quietly upgrade unknown consent to assumed consent. Where the incumbent's export does not include consent state — which is common — the honest position is that those numbers arrive unknown, and your first send should be one your counsel is comfortable sending to an unknown-consent number.
The roster disaster, and the guardrail that prevents it
The single most damaging failure in a district migration is not losing data. It is a truncated or partial export being loaded as though it were complete, so every record absent from the file is interpreted as a withdrawal. A district can lose most of its active families in one sync, and the failure is silent: the system did exactly what it was told.
Three controls, in order of effectiveness:
- A withdraw guardrail on the import itself. Kastr aborts a roster sync that would withdraw more than half the active records and reports the abort rather than proceeding. It is a blunt control and that is the point: no legitimate mid-year sync withdraws half a district.
- Reconcile counts per school before the load, not after. Compare the file's per-school counts against the SIS. A school missing entirely is obvious in a table of twenty rows and invisible in a total.
- Load into a non-production tenant first. If the vendor cannot give you one, that is itself a finding worth noting during procurement.
The family message sequence that works. Three messages, four weeks apart from first to last, on every channel you currently have. First: what is changing, when, and what the family needs to do, which should be nothing. Second, two weeks later: the same thing, shorter, plus what to expect in the first week. Third, on the day: how to reach the school if something does not arrive. Families act on the third message. The first two exist so the third is not a surprise, and so that a family who calls in week one has already had two chances to see it — which is the difference between a support spike and a complaint to the board.
Questions people actually ask
When is the best time of year to switch communication platforms?
Between school years, with the export taken after the last instructional day and the cutover landing on the back-to-school attention spike. Summer gives the front office time to learn the tool and lets family onboarding ride a moment when families are already opening things from school. The main risk is thin summer staffing on the technology team.
Can a district switch platforms mid-year?
Yes, and sometimes a renewal date forces it. Cut over immediately after a break rather than before one, keep the incumbent live for two weeks past cutover, and expect roster reconciliation to be harder because the roster is moving. If the choice is available, a short bridging extension with the incumbent to land the cutover in summer is usually cheaper than the disruption.
Do parents have to opt in again after a platform switch?
It depends on what your incumbent's export contains and on the class of message. Carry forward the consent state as well as the number; where the export does not include consent state, those numbers arrive unknown and should be treated that way. Broadening what you send at the same time as switching platforms is a second decision that needs its own approval.
How long does a district-wide migration take?
Eight to twelve weeks of elapsed time for a typical district, dominated by two externally controlled items: carrier registration for text traffic, which runs two to six weeks and cannot be accelerated, and the notice window in your existing contract. The hands-on work is a fraction of that; the calendar is set by what you are waiting for.
What is the most common failure in a platform migration?
Loading a truncated export as though it were complete, so every missing record is treated as a withdrawal. The second most common is starting carrier registration late, so text messages are filtered by carriers while the platform reports them as sent. Both are silent failures, which is why both need a control rather than vigilance.
How do we run two platforms in parallel safely?
Make one authoritative for contact data and consent and the other strictly downstream, and send from only one of them at a time. The failure mode of parallel running is not cost, it is divergence: two contact databases with separate correction workflows drift apart, and the drift is invisible until an emergency send misses the households whose numbers were only fixed in the other system.
One price. Every feature. Locked for three years.
$3.50 per student per year under 5,000 students. No tiers, no add-on modules, no per-message fees. Published on the site because you should not have to book a call to learn a price.