better shortlists for mentor universe
Mentor Universe is a study abroad consultancy where counsellors help students select universities. But as demand surged, the platform struggled to scale. More students began complaining about incorrect university shortlists leading to frequent meetings and declining trust.
the problem
declining trust in counsellors
Mentor Universe, a study abroad consultancy platform, saw rising student complaints about incorrect university shortlists. 5 out of 20 shortlists involved mismatched student preferences - like matching a 8 GPA to a student with an 7 GPA or promising options only to get rejected.
These errors led to more meetings, increased counsellor time per student and reducing trust on the platform.
counsellor research
shadowing counsellors
To move beyond assumptions, I spent hours shadowing counsellors at work. I watched their screens, listened to their reasoning, and followed their process step-by-step. Here's what I found.
Gathering context was messy
Filtering was quick but crudE
Once they had the context, they used a roughly filtered sheet (GPA, test scores, etc.) and then relied on intuition and past experience to make final recommendations.
Both posed their own friction points:
identifying the root cause
the gpa mismatch
The counsellor mistakenly used a 7 GPA filter for student A instead of 8 GPA because he had just created a shortlist for student B with a 7 GPA 30 minutes earlier. Managing many floating student details caused cognitive overload.
the intuition mismatch
Counsellors heavily relied on past experience to recommend unis. This, albeit being crucial, would often create problems due to an absence of a backing factor.
new problem framing
With the root causes clear, I reframed the challenge:
HOW MIGHT WE
How might we prevent counsellors from copying context multiple times and give their intuition a solid backbone to rely on?
Early Ideas & Why They Didn’t Work
Initial ideas revolved around creating a full-fledged AI university shortlist maker based on student preferences.
However, generating shortlists wasn't just a set of rules. It has lots of moving factors (like scholarship amounts) as well as a high degree of counsellor experience and trending news. Trying to incorporate all the rules was extremely difficult.
Huge expenses as well as an unreliable ROI had us shift to something which would increase counsellor efficiency while still being feasible.
final solution
Support, Not Replace
Instead of automating everything, I focused on embedding the right context in the right place.
Auto-embedded student context
GPA, exam scores, and preferences are pulled directly from the student profile.
Data-backed intuition
Counsellors start with a shortlist filtered only by hard data (GPA, test scores). A historical record of admission trends can easily suggest if the counsellor is over-estimating or under-estimating an option.
KEY TAKEAWAY
Context is now built-in. Intuition is now supported by evidence.
tradeoffs
I proposed a student view showing shortlist "health" with similar historical trends. The client rejected it, fearing students would challenge the data and overburden counsellors.
We prioritized our main goal - counsellor efficiency and accuracy and tabled the feature to avoid this risk.
phased rollout
Due to the complexity, the project was decided to be rolled out in stages. Beginning with first migrating the data from Excel to the platform with basic filters. The 2nd phase would include the match percentage scores. The 3rd and final phase would aim at showing similar student profiles for each match.
Final Outcome
2x
Counsellor Confidence
90%
Error reduction in shortlist generation