Marketing, Startups | Константин Алексеев, Олег Брагинский
Oleg Braginsky, founder of the Troubleshooters School, and student Konstantin Alekseev encountered a classic fintech startup dilemma: Google and Meta had exhausted their scaling potential. A fintech client approached for a traffic acquisition audit – product hit a ceiling attracting users through traditional channels.
What Programmatic Means and Why It Matters for Fintech?
Programmatic represents automated mobile traffic acquisition through machine learning algorithms. Programmatic platforms analyze billions of requests in real-time, determining the optimal audience for displaying ads inside applications.
Key programmatic campaign metrics:
- IPM (Installs Per Mille) – installs per thousand impressions
- ROI – return on investment (revenue / spend × 100%)
- CPI (Cost Per Install) – cost per installation
- CPM – cost per thousand impressions
- CPA – cost per target action/
Fintech products particularly require programmatic sources: Meta and Google's regulatory restrictions often block aggressive scaling of financial products.
Analyzing the Current Situation
Before delving into implementation, we examined the client's existing traffic sources. This stage proved critical – premature scaling into programmatic can result in budget loss if Google or Meta haven't been optimized to their limits yet.
Diagnostics revealed active campaigns in Liftoff – an in-app programmatic platform with advanced ML algorithms. The campaigns demonstrated potential but were managed chaotically. The team discovered three systemic problems:
- How do we synchronize creative updates with Liftoff managers without delays?
- How do we determine the right optimization frequency without destroying machine learning?
- Which metrics should we prioritize when making decisions about pausing creatives?
Liftoff's Platform Characteristics
Liftoff utilizes proprietary ML algorithms for optimizing bids in impression auctions. The platform demonstrates impressive results but requires understanding its nuances:
- Algorithm learning period: N days of active impressions before performance stabilizes. N depends on how fast clients make the target event. This meant the team would need to spend advertising budget on initial model training without guaranteed results.
- Absence of self-serve interface: campaign management operates through personal managers. The human factor creates risks – delays uploading creatives, configuration errors, untimely reactions to metric drops.
- Dependence on creative rotation: the algorithm rapidly "burns out" audiences on a single creative. We needed an established system for updating visuals.
Implementing a Creative Management System
We created a structured process for interacting with Liftoff managers through Google Sheets. The system included status fields:
- To upload – creative sent to manager
- Queue – creative wait for free space in campaign
- Live – displaying to audience
- Paused – temporarily stopped for analysis
- To restart – restart overperformed creatives
- Create native video – to create new video asset.
The spreadsheet integrated with Google Drive – the Liftoff manager received a direct link to the folder with new video files or statics. This eliminated the risk of emailing files with version control loss.
The system worked: time from request to creative launch decreased from 3-4 days to 18-24 hours. Timely creative rotation represents a fundamental component of successful programmatic campaigns. Without it, the algorithm loses efficiency, audiences "burn out," metrics decline.
Optimizing Existing Campaigns
The team analyzed the history of launched campaigns. The picture proved discouraging: the client optimized creatives either too aggressively (every 1-2 days) or by incorrect metrics (stopping creatives with high CPI while ignoring conversion to target action).
We applied cohort analysis with a 3-day window – the minimum period for Liftoff's ML algorithm stabilization. Each creative was evaluated by two factors:
Factor 1 – Entry Efficiency: IPM (installs per 1000 impressions) + CPM (impression cost). A creative with low IPM doesn't "hook" the audience regardless of subsequent metrics.
Factor 2 – Conversion Depth: percentage of users who completed the target action (registration, verification, first deposit). A creative could demonstrate excellent IPM but attract non-target users.
The historically established approach of "looking at CPI" proved erroneous for fintech products. Cheap installations don't guarantee quality users – conversion to monetizable actions is critical.
Monitoring Performance Without Premature Optimization
We implemented discipline: no decisions before forming a complete 3-day cohort. This became the project's most challenging part – pressure from client management demanded "doing something" the very next day after launching a creative.
The team tracked supporting metrics:
- CPA to registration: early indicator of traffic quality before forming deposit cohorts
- CPM: sharp growth signaled increased auction competition or targeting problems
- CPI: secondary metric for quick diagnostics, not the decision foundation
- IPM: drops indicated audience fatigue with the creative.
We established critical metric thresholds with the client. For example, if IPM dropped below 2.5 with an average of 4.2, the creative was marked for replacement regardless of cohort – obvious inefficiency didn't require additional spending for verification.
Scaling Results
After implementing the systematized approach, Liftoff transformed into the second-largest source of new clients with 89% ROI in the first working month. This exceeded expectations – initially the team projected 60-70% ROI for the programmatic channel.
The first phase's success enabled scaling the approach to additional geographies (Southeast Asia) and the iOS platform. As a result, Liftoff secured third place by number of paying users among all company sources – trailing only Google and organic traffic.
A Systems Perspective on Traffic Scaling
This project validated a principle: technology solves only 30% of the challenge, the rest is processes and discipline. The most advanced ML algorithm proves useless without structured creative management and restraint in optimization.
Fintech traffic scaling isn't a story about a "magic button." It's a balance between decision-making speed and patience to wait for statistically significant data. Teams understanding this balance extract maximum value from programmatic channels. Others burn budgets on chaotic optimization.
A pleasant bonus: the client continued working with Liftoff independently, applying the established system to new markets. Sometimes the best consulting outcome is when you become unnecessary.