Data-Driven Single Selection: Havana vs OMG Case Study

When Camila Cabello released two tracks on the same day in August 2017, her label backed the wrong one. Streaming data corrected the decision within 48 hours. The result reshaped how the music industry approaches single selection, and it remains the clearest example of why real-time analytics, not boardroom instinct, should drive release strategy.

This case study breaks down what happened, what the data revealed, and how labels, managers, and artists can apply the same framework to their own releases.

On August 3, 2017, Cabello released two singles simultaneously: "Havana" featuring Young Thug and "OMG" featuring Quavo. Her label, Epic Records, placed the marketing weight behind "OMG." It had the more commercially proven feature artist, the more radio-ready sound, and the full promotional rollout. "OMG" received stronger initial iTunes positioning and a higher slot on Spotify's New Music Friday playlist.

The label's bet was logical. But by Saturday morning, roughly 48 hours after release, Spotify's internal data told a different story. Listeners were gravitating toward "Havana" by every measurable signal: higher save rates, stronger completion rates, more playlist additions, and faster organic sharing.

Spotify's head of pop programming later confirmed to Billboard that the platform could identify "Havana" as the audience favourite within that first weekend. The label pivoted. "Havana" was elevated from promotional single to official lead single for Cabello's debut album Camila, replacing the earlier release "Crying in the Club" entirely.

The divergence between the two tracks was visible across every major streaming metric, and it widened fast.

Havana's trajectory:

OMG's trajectory:

Same artist. Same release day. Same team. The only variable was the audience's response, and the data captured it in real time.

The save rate, meaning the percentage of listeners who add a track to their personal library, was the earliest and clearest signal separating the two songs.

Save rate matters because it reflects deliberate intent. A stream might be passive. A save is a decision: the listener wants to hear this again. Spotify's algorithm treats saves as one of its strongest engagement signals, directly influencing whether a track enters Discover Weekly, Daily Mix, and eventually algorithmic playlists with broader reach.

Industry benchmarks reinforce this. According to data from Spotify for Artists and independent analysis, strong save rates typically fall within these ranges:

Havana's early save rate signalled that listeners weren't just consuming the track passively. They were claiming it. That behavioural signal, visible within the first 48 hours, predicted the song's long-term trajectory more accurately than any A&R meeting.

The Havana–OMG story crystallised several principles that now inform modern release strategy across the industry.

Every executive at Epic believed "OMG" was the hit. The feature was bigger (Quavo was at peak cultural relevance in 2017), the production was more conventionally radio-ready, and the sound aligned with prevailing market trends. None of that mattered when measured against actual listener behaviour.

This doesn't mean intuition is worthless. It means intuition should be tested. The IFPI Global Music Report 2025 notes that labels invested a record $8.1 billion in A&R and marketing in 2023, yet only one or two out of every ten artist signings become commercially successful. Data-driven single selection doesn't eliminate risk, but it allocates resources toward tracks that have already demonstrated audience traction.

The label's willingness to pivot within 48 hours was as important as the data itself. Streaming momentum is perishable. A song that generates strong early engagement feeds the algorithm, which expands its audience, which generates more engagement. It's a compounding loop. Delayed action breaks the loop.

In 2017, this kind of rapid pivot was still unusual. Today, the infrastructure exists to make it standard. Spotify for Artists, Apple Music for Artists, and cross-platform tools like Chartmetric provide real-time access to the signals that matter: save rates, completion rates, skip patterns, and geographic clusters.

If you cannot determine which of two strong tracks should lead, release both. Let the audience decide. This approach treats a release as a structured experiment rather than a one-shot bet.

The cost of releasing two tracks simultaneously is marginal compared to the cost of backing the wrong single with a full campaign. Cabello's team gained clarity within a weekend. Without that dual release, "Havana" may have remained an album deep cut, never tested and never discovered.

Havana's early Spotify traction didn't stay on Spotify. The save rate and completion rate signals triggered algorithmic playlist inclusion, which expanded the audience, which drove Shazam activity (a leading indicator for Apple Music editorial attention), which increased radio interest. Each platform's data fed the next. The label's pivot enabled this cross-platform flywheel; a delayed decision would have stalled it.

You don't need a major label budget to apply this framework. The underlying principle (test before committing resources) scales down to any level.

Select tracks that you believe have commercial potential but that you can't confidently rank. The uncertainty is the point. It's what makes the test valuable.

Post 15–30 second snippets of each track on TikTok and Instagram Reels. Allocate a small, equal budget ($50–100 per track) to ensure comparable exposure. Run ads simultaneously to control for timing and external variables.

After 7–14 days, compare across these metrics:

The track with the strongest signal profile gets the full rollout: playlist pitching, ad spend, content calendar, and PR. The others can still be released as album tracks, B-sides, or future singles. Nothing is wasted.

Even after committing to a lead single, keep watching the data on all released tracks. Your B-side might be your hit. That's exactly what happened with Havana.

The Havana case study arrived at an inflection point. In 2017, streaming had just overtaken physical and download revenue as the industry's primary revenue source. By 2024, streaming represented 69% of global recorded music revenue, reaching $20.4 billion for the first time (IFPI Global Music Report 2025). The number of paid subscription accounts worldwide grew to 752 million.

This scale matters because it makes data-driven decision-making not just possible but necessary. With over 100,000 tracks uploaded to Spotify daily, the volume of new music makes gut-feel A&R increasingly unreliable. The labels, managers, and artists who build systems for real-time signal detection, and who make fast decisions based on those signals, hold a structural advantage.

Spotify's own transparency report, Loud & Clear, reinforces this: in 2025, the platform reported over $10 billion in total payouts, with more than 1,500 artists earning over $1 million annually. The artists reaching those thresholds are, disproportionately, those whose teams optimise release strategy around data rather than assumption.

In the Havana case, Spotify's internal data identified it as the stronger track within 48 hours of release. For most artists, meaningful signal differentiation between tracks typically emerges within 7–14 days of release, depending on the size of the existing audience and the level of promotional support. The key metrics to watch early are save rate, completion rate, and skip rate in the first 30 seconds.

The save rate measures what percentage of listeners add a song to their personal library after hearing it. It matters because it reflects intentional, deliberate engagement, not passive listening. A save rate above 4% is considered excellent and typically triggers increased algorithmic distribution on Spotify. Save rate was the earliest indicator that Havana would outperform OMG.

Releasing two or more tracks simultaneously can work as a testing strategy, but it requires discipline. The approach is most effective when you allocate equal initial promotion to each track, measure audience response over a defined window (7–14 days), and then concentrate your marketing resources behind the track with the strongest data. Even at small scale, this method reduces the risk of backing the wrong song.

Havana became the best-selling digital single of 2018 globally, achieved RIAA Diamond certification (10 million US units), and surpassed 1.6 billion Spotify streams by 2021. OMG accumulated approximately 87 million streams. The difference, roughly 20x in streaming performance, stemmed entirely from audience behaviour that was visible in the data within the first weekend.

Spotify for Artists provides save rates, completion rates, skip patterns, and listener demographics for any distributed release. Apple Music for Artists offers plays, Shazam data, and library adds. For cross-platform tracking, Chartmetric aggregates signals from Spotify, Apple, YouTube, TikTok, and more. Combined, these tools give artists and managers the same visibility that major labels used to keep proprietary.