Switching Course Evaluation Software: 2026 Migration Guide for Universities

Migrating course evaluation software is not a data-export problem — it is a continuity problem. The value of an evaluation system lives in longitudinal records, faculty trust, and accreditation evidence that spans years. Switch carelessly and you break the trend line that quality committees and accreditors rely on. Switch well and you inherit every advantage of a modern platform without losing the institutional memory you have built.

This guide is written for QA directors, heads of teaching and learning, institutional-research leads, and procurement teams who have already decided (or are close to deciding) to move off an incumbent such as EvaSys, Explorance Blue, Qualtrics, Watermark, or a legacy system like CoursEval, and now need to run the migration without a governance incident. It is deliberately vendor-neutral: the sequence below applies whichever platform you are moving to.

The short answer

There are three viable migration patterns, and the right one depends on how tightly your accreditation and reporting are bound to fixed numeric instruments.

Rip-and-replace — retire the old system at a term boundary and start fresh. Cleanest operationally, but you lose in-system trend continuity unless you migrate historical data.
Run-both (parallel) for one to two cycles — keep the incumbent for mandated numeric instruments while the new platform runs alongside for depth. Lowest risk, higher short-term cost, and the pattern most European institutions actually use.
Phased by faculty or programme — pilot in one faculty, prove the reporting, then roll outward. Best when internal buy-in is the main constraint rather than the technology.

If your evaluations feed directly into accreditation submissions with fixed quantitative thresholds, favour run-both or phased. If your instruments are due for a redesign anyway, a term-boundary rip-and-replace is often the honest moment to modernise.

Migration pattern	Best when	Main risk	Trend continuity
Rip-and-replace	Instruments are being redesigned; clean break wanted	Loss of in-system history; change-management shock	Only if history is migrated/archived
Run-both (parallel)	Accreditation tied to fixed numeric instruments	Double cost; respondent fatigue if over-surveyed	High — old series continues during overlap
Phased by faculty	Buy-in is the constraint, not the tech	Inconsistent institution-wide reporting during rollout	Mixed during transition

The five things that actually break in a migration

Historical trend lines. Faculty and programme leads expect to see this year against the last five. If the old scores do not come across (or cannot be shown next to new data), you lose the comparison that makes evaluation actionable.
Instrument equivalence. A new platform rarely reproduces your old question set verbatim. Even small wording or scale changes break strict year-on-year comparability — a real methodological issue, not a cosmetic one.
Enrolment and identity plumbing. Course-instructor-student mappings from your SIS/LMS (Banner, SAP SLcM, Moodle, Canvas) must be re-wired. Explorance publicly documents dedicated data-preparation tooling and migration services precisely because this step is where projects stall.
Report access and permissions. Deans see faculty-level data; instructors see only their own. Getting the visibility matrix wrong on day one erodes trust immediately.
The exit itself. Extracting a complete, usable copy of your incumbent data — before the contract lapses — is a task with a deadline you do not control if you leave it late.

A migration plan that respects the academic calendar

Course evaluation is seasonal, and the calendar is unforgiving. Plan backwards from a term boundary — never cut over mid-semester.

T-6 months — scope and export audit. Confirm exactly what you can export from the incumbent and in what format. Most systems export closed-item data to CSV; Explorance Blue, for example, supports CSV export and offers migration services. Establish whether open-text comments, response metadata, and generated reports are exportable, not just the numeric summaries.
T-4 months — data mapping. Decide what migrates live into the new platform versus what is archived for reference. A widely-cited public example: when Georgetown moved off CoursEval to a new system, it archived instructors' historical reports and distributed a ten-year quantitative summary plus five years of individual reports rather than importing everything wholesale. Archiving is a legitimate, often cheaper choice.
T-3 months — instrument design and equivalence decision. Rebuild your instrument in the new platform and decide, explicitly, how you will treat the break in the series (see below). Document the decision so an accreditation panel can see it was deliberate.
T-2 months — integration and permissions testing. Wire SIS/LMS feeds, SSO, and the report-visibility matrix. Test with real (anonymised) enrolment data.
T-1 month — pilot and communications. Run a small live pilot, brief faculty on what is changing and why, and publish the response-visibility and anonymity rules.
Go-live — at the term boundary. First full cycle on the new platform, with the incumbent either retired or running in parallel for one more cycle.

The Likert-to-conversational continuity question

The hardest migration is not tool-to-tool of the same type — it is moving from fixed Likert surveys to a richer model such as AI-moderated conversational interviews. You cannot directly compare a 1-to-5 mean to a themed qualitative dataset. Institutions wedded to Likert-only longitudinal comparison must decide how to blend the two.

The pragmatic pattern is a bridge cycle: keep a small core of your validated numeric items running (in the incumbent or the new platform) so the historical series does not snap, while the conversational layer collects the depth that fixed surveys miss. After one or two cycles you have both the unbroken numeric trend and a new, richer evidence base — and you can retire the legacy instrument on your own timetable rather than under contract pressure. This is the same interview engine that powers general customer and user research on the main Koji platform, applied to students.

GDPR, data retention, and a clean exit

Under GDPR, storage limitation and purpose limitation apply to student feedback as much as to any personal data. A migration is the right moment to reconcile two competing pressures: accreditation bodies want multi-year evidence, while data-protection principles push you to not keep identifiable data longer than necessary. Practical guidance:

Export a complete copy of incumbent data — including open text and metadata, not only dashboards — before the licence ends.
Distinguish identifiable raw responses (short retention) from aggregated/anonymised trend data (longer retention for QA and accreditation).
Confirm the new vendor's hosting location and Data Processing Agreement in writing. For European institutions, EU hosting and a current DPA are table stakes, not nice-to-haves.
Confirm the incumbent's deletion and return obligations at contract end so data is not stranded on a decommissioned server.

When staying with your incumbent is the right call

Honesty serves this audience. Do not migrate if:

You process a significant share of evaluations on paper and depend on ICR scanning — EvaSys handles this natively, and few modern online-first tools replicate it.
Your accreditation submissions are locked to a specific validated instrument mid-cycle, and a break in the series would create more risk than the new tool removes this year.
Your incumbent is contractually locked for another cycle and the switching cost outweighs the near-term benefit — in which case run a small parallel pilot now and plan the full move for the next renewal.

A good migration is one you can defend to a quality committee. If the timing is wrong, the disciplined move is to pilot in parallel and cut over at the next clean boundary.

Where Koji fits

Koji is built for the modern end of this spectrum: AI-moderated conversational interviews with students, automatically themed into accreditation-ready qualitative evidence, with EU/GDPR-aligned data handling. For institutions switching primarily to escape thin free-text and slow manual coding, Koji is designed to slot in as the depth layer during a bridge cycle and then, for many programmes, as the primary instrument once the historical series is secured. It is not a like-for-like replacement for paper-based numeric administration — where that is a hard requirement, keep an incumbent for that slice and use Koji for the qualitative evidence surveys cannot produce.

Related resources

Planning a switch? Map your term-boundary timeline first, secure a complete export of your historical data, and decide your continuity strategy before you sign. If you want to see how conversational evaluation fits alongside your existing numeric instruments during a bridge cycle, talk to the Koji team.