Running 8 scholarship campaigns from start to finish

B. Context

educations.com is a study abroad search platform owned by Keystone Education Group. It runs annual scholarships for international students, several with partners such as IAESTE and Girls on Campus. During a six month digital marketing internship in Sweden (March to August 2026), I was the operational owner of the scholarship workstream. Each cycle produces indexed web articles, social posts, certificates and partner-facing outcomes, so the programme sits between brand, content and partner management.

C. The problem

Every cycle ran the same chain: finalist selection, notification, interviews, winner decision, acceptance, non-winner notifications, social announcement, articles, certificates. It ran on templates copied between scholarships, and copied templates carried errors: a certificate from a different scholarship, an applicant count phrase that was untrue for the current campaign, a notification drafted from another scholarship’s template. Finalist response rates were low. Any slip lands on an indexed page with a named student on it, so the cost is reputational and personal.

D. My role and the team

I ran the day to day operation: candidate correspondence under my name, finalist and winner articles, interview scripts, evaluation workbooks. My line manager co-conducted finalist interviews with me. My supervisor set the five evaluation dimensions and their priority and made the final winner determination; I produced the ranked assessment that fed it. Partners were credited as co-brands on every published asset.

Drafting was AI-assisted: I set each brief, supplied source material, corrected draft errors and made every publication and notification decision. No LLM sat in the decision path; a person reviewed and decided at every decision point. We chose winners as a team; I decided sequencing, wording and what was held back for verification.

E. Key decisions

1. Decision: Fix the announcement order: written winner acceptance, then non-winner notifications, then social posts. Why: The winner had a short window to confirm. A post or a rejection cannot be recalled, and no finalist should learn the outcome from a feed. Alternative I rejected: Sending all three in parallel to hit the published date.

2. Decision: Treat low finalist response as a funnel volume problem and recommend over-inviting at the finalist stage. Why: Dropout was structural (no reply, withdrawal, eligibility lost after invitation). Re-screening the same pool would not change that. Alternative I rejected: Tightening screening criteria, which shrinks the group further.

3. Decision: Publish only verified facts, and name no third party without consent. Why: Finalist pages are indexed and checkable by anyone, including the students. Inherited copy and assets were wrong for the current scholarship. Alternative I rejected: Reusing inherited templates and cards as they came, for speed.

4. Decision: In the consolidated record, count someone as a finalist only if they appear in a published article; hold unconfirmed fields as explicit placeholders. Why: People who never replied should not sit in a tracked record. Placeholders make gaps visible instead of guessed. Alternative I rejected: Inferring gender, programme or surname from names and page text.

5. Decision: When a public-facing issue arose after a finalist announcement, check the published terms before drafting anything. Why: The published eligibility rules were the reference point for any response, and the legal question belonged to Legal. I mapped the risks (legal, brand, precedent) and gave leadership ranked options with ready-to-send messages. Alternative I rejected: Polishing the requested message as written.

F. What I built or produced

• Finalist notification emails, non-winner notifications with individual closings, and certificate copy, per campaign.
• Finalists articles in house style: third person, three fixed subheadings, no direct quotes, alphabetical by surname.
• Finalist interview scripts (an English speaking layer and a Chinese reading layer) and second interview scripts for winner reveals.
• Winner selection workbooks: questionnaire records, weighted scoring across five dimensions (Globally Minded; Entrepreneurial Drive and Initiative; Motivation to Innovate and Impact Industry; Short and Long term Plans; Funding and Financial Considerations, which weighs the specificity of the applicant’s funding plan and carries the highest weight), an evidence map, review flags and a ranked recommendation.
• Winner articles titled “Winner of the [Scholarship] [Year]”, direct quotes only from interview transcripts, every figure flagged for verification.
• Carousel card copy and TikTok captions for announcements.
• A consolidated 2026 finalist and winner record with Excluded, Verification and README sheets.
• For the public-facing issue: a risk map, ranked options, ready-to-send messages and an internal summary for leadership.

G. Result

• Campaigns: 8, six of them end to end, including Study a Master’s in Europe, Go Global MBA, Sustainable Energy and Environment (with IAESTE), Adult Learners, Study a Bachelor’s in the USA, Social Science Excellence (with Girls on Campus) and the Women’s Scholarship.
• Published articles: 12 finalist and winner articles on educations.com, bylined to the brand, written and published by me.
• Finalists interviewed: 20+ from 15+ countries.
• The announcement sequence held in every campaign I ran.
• Co-branding carried through all partner assets.
• Qualitatively: inherited template errors were caught in certificate, card and article copy before publication. No unverified statistic was published; no third party from a finalist’s materials appeared in public copy.

H. What I would do differently

1. No ranking before interview parity. In one campaign a finalist had no interview record when the assessment was drafted. Next time, every finalist gets the same interview before any ranked view goes to the reviewer.
2. Lock the definition of the Funding dimension before scoring. It was still open (need, planning quality or award leverage) during evaluation, so I scored under two readings.
3. Confirm name, pronouns and programme title at the notification stage with one reply from the finalist. Cards, articles, emails and certificates drifted on exactly these fields.

I. Evidence assets

• Finalists articles: Go Global MBA 2026 (22 May 2026), Study a Master’s in Europe 2026 (8 June 2026), Sustainable Energy and Environment 2026 (29 June 2026), Adult Learners 2026, Social Science Excellence 2026, Study a Bachelor’s in the USA 2026: Public link, Tier A.
• Winner articles for the same scholarships: Public link, Tier A.
• Announcement sequence and cycle workflow: Anonymised diagram, Tier B.
• Interview question structure and five dimension framework: Anonymised description, Tier B (structure only, no scores, no candidate content).
• Published carousel cards and TikTok announcement posts: Public link, Tier A.
• Workbooks, consolidated record, email and certificate templates, interview scripts in full, the issue advisory pack, transcripts and applicant materials: Tier C, not shown.

J. Keywords

Scholarship programme management; Campaign operations; Content marketing; Editorial house style; Candidate communications; Interviewing and assessment; Partner co-branding; Issue management; GDPR-aware data handling; Cross-functional coordination; Social media content; Process design