| Company | HCL Concerts, HCL Technologies’ classical music initiative |
|---|---|
| Audience | Classical music enthusiasts and culturally curious urban Indians |
| Role | Digital advertising strategy. Funnel architecture, audience design, creative direction. |
| Market | India, monthly live-streamed concerts |
The situation
Monthly classical concerts, streamed live, with a mission to platform lesser-known artists and traditional forms. Near-zero digital following, no community, and no recognition outside a small circle of people who were already looking.
The goal was not clicks or registrations. It was watch time — people who turn up to a live stream and stay. That single fact makes almost every default in a media plan wrong.
The constraint
Classical music has no mass appeal and intense niche appeal. The job was never to create demand where none exists. It was to find a small, real audience efficiently and get them to show up while the thing was happening.
Reach-led buying is optimised to find people who will click. In an attention category, those are close to the worst people you can buy: they arrive, register a click, and leave before the performance gets going. The cost of that shows up not as a bad click price but as a watch time number that will not move no matter how much traffic you add.
Broad interest targeting on the actual subject — Hindustani and Carnatic music, sitar, tabla, classical dance — produced better watch time than demographic targeting did. That is the whole thesis in one observation: what someone cares about predicts whether they will stay; who they are does not.
What I did
Built the audience out of watching, not out of demographics
Custom audiences seeded from video-view thresholds — 25%, 50%, 75%, 95% — rather than from all website visitors. Lookalikes built from the 95% viewers produced the highest watch time at the lowest cost per viewer, and comfortably outperformed lookalikes built from undifferentiated traffic. If you want people who watch, model them on people who watched.
Three phases per concert, because a live event is a moment, not a campaign
- Pre-event: artist spotlight content to build curiosity before there was anything to attend
- Live day: urgency messaging against a fixed start time, which is the only moment the offer is real
- Post-event: a highlights reel that re-engaged the audience and seeded the next concert, extending the engagement window by around eleven days per concert
Creative matched to funnel stage rather than one video for everyone
Fifteen seconds to introduce an artist at the cold stage, thirty seconds of them performing at consideration, sixty seconds of performance excerpt with a live CTA for retargeting. Matching the cut to the stage outperformed a generic video by roughly 2.4x on watch time.
Reallocated weekly on watch time per rupee, not on volume
Roughly 40% cold, 35% warm retargeting, 25% lookalike expansion, adjusted each week by which segment was actually producing minutes watched. Dynamic allocation improved watch time per Rs 1,000 spent by about 40% over the run.
What moved, and what it means
Watch time up 1400% in three months. Live event traffic up 10x. Cost per landing page view down 3x. Average view time per session up more than 450%.
The pair to read is watch time up and cost per view down. Attention is usually something you buy more of by spending more. Getting more minutes watched while each view got cheaper means the audience composition changed, not the budget — the people arriving were closer to wanting this before the ad loaded.
The 450% average view time is the number that says it worked rather than that it grew. Total watch time can rise simply because more people showed up. Average view time per session only rises when the individual person stays longer, and that cannot be bought with reach.
For engagement campaigns, audience quality beats audience size by a margin that is difficult to overstate. A small, well-matched audience that watches most of a concert is worth more than a large one that clicks and leaves — and the large one costs more.
What I would tell you before you hire me for this
This was agency-side work at PROHED. I owned the funnel architecture, the audience design, the creative direction and the weekly reallocation. I was not the client.
The larger caveat is that this ran on pre-ATT Facebook. The single most effective lever — lookalikes seeded from 95% video-view custom audiences — depends on pixel and view-threshold signal that Apple’s App Tracking Transparency and the broader signal loss since have materially degraded. Anyone quoting you this playbook unchanged in 2026 is selling you a mechanism that no longer works the way it did.
What transfers is the principle: build your audience out of the behaviour you actually want, not out of who people are. What has to be rebuilt is how you capture that behaviour now — first-party signal, on-platform engagement, retention cohorts. I would expect to spend the first month of an engagement working that out rather than pretending the old route still runs.
Where this is relevant
Anywhere the thing you are optimising for is attention rather than a click: live events and streams, video, podcasts, webinars, community, and any content business whose economics depend on people staying rather than arriving. It applies with particular force to niche cultural categories, where the temptation is to widen the audience to hit a volume target and the effect of doing so is to make every metric worse at once.