Referral at Internshala ran as a campaign, not a channel. Every six to eight weeks we bundled internships around a theme — highest-paid internships for engineers, that sort of thing — and attached a leaderboard. ₹10,000 for the most invites, ₹5,000 for second, down the ranks.
It produced referrals. It produced them from about a dozen people.
Everyone else did the arithmetic, concluded they'd never finish top three, and referred nobody. We weren't running a referral programme. We were running a tournament — and a tournament only ever pays a handful of people.
Referral strategy, campaign design, and the operating model behind both — the reward mechanism, the campaign calendar, and the tooling that produced them.
Instead of paying the winners, we gated the campaign itself. For the first three days after launch, the featured internships stayed locked. Invite five friends, get in early, and apply before anyone else.
That changed who the reward was for. A leaderboard rewards the top 1%; early access rewards anyone willing to send five messages. And the thing being unlocked wasn't cash — it was the exact opportunity they'd come to the site for. The incentive attracted the people we wanted rather than prize hunters, and participation stopped being concentrated in a handful of power users.
The dead months weren't a marketing problem. Talking to college students made it obvious: November–December and March–April are exam season. Nobody commits to a two-month internship while they're revising.
The answer wasn't better messaging on the same offer — it was a different offer. One-day internships at companies like Xiaomi, Lenovo and HCL. Real exposure, real brand names, zero commitment. Something you could say yes to mid-exams.
Those months stopped being holes in the calendar.
We templatised everything: a fixed brief format, a shortlist of formats that had already proven themselves, and a locked calendar instead of ad-hoc scheduling. Then we built an in-house tool that automated the product-creation step — landing page generation and internship featuring — which had been the bottleneck making every campaign a bespoke build dependent on product-team time.
Throughput went from one or two campaigns every 45 days to two or three every month.