Using Media Mix Modelling, Realtime Agency Identifies Key Revenue Drivers & Supports Budget Decisions For Glint Pay App

UK-based Glint Pay Ltd. is a fintech company that enables its clients around the world to instantly buy, sell, save, spend and send their physical gold and other currencies through the GlintMastercard® and Glint App. With more than 130,000 registered users, Glint has completed over $350 million worth of transactions to date.

Challenge

As a fairly new company, Glint has ambitious growth targets in terms of downloads, as well as new deposits (first gold buys). In order to meet these goals, the company needed to understand which marketing activities were actually delivering a good ROI.

However, recent shifts, such as Apple’s changes to its tracking policies for third-party apps, had made in-app actions difficult to measure. This led to poor attribution, poor audience quality,and overall inefficiency for app campaigns.

Glint needed a resilient measurement strategy that wasn’t completely reliant on log-level data sources (like SDK, pixels or cookies) so it could truly understand if outcomes like first gold buys were being driven by campaigns rather than simply correlated to channel spend.

Strategy

Glint and Realtime, its performance agency, decided to shift their focus to marketing mix modeling (MMM), which leverages aggregated and privacy-safe data sources to better understand which channels and strategies delivered business outcomes most efficiently. Because it’s transparent, customizable and allows for inclusion of experiments (which helps to minimize analyst bias), Robyn from Meta Open Source was their preferred MMM solution.

To address the challenge at hand, Glint took a two-step approach. First, they built a Robyn model by feeding a variety of data sets into it beyond paid media channel spends, like influencer activity, tweets, interviews, the price of gold and world events. They also had the model factor in other variables such as holidays and competitor activity.

Secondly, Glint conducted further testing and experimentation to validate the results from their Robyn model to ensure they were confident with the findings. For example, Robyn identified Facebook iOS as a significant driver of performance. In order to prove this hypothesis, Glint ran a regional test in the US, splitting the market into two regions: one activating Facebook ads on Android only and one targeting both Facebook iOS and Android ads.

Results

Thanks to the incremental experiment, revenue was shown to be 47% higher for users exposed to Facebook iOS in addition to Facebook Android ads, which proved Robyn’s MMM prediction to be correct. Facebook iOS had been identified by the model as the largest driver of first gold buys among paid media channels, contributing to 9% of the total. In response to this insight, we could confidently increase Glint’s budget on Facebook by 30%.

Gold price was identified as the largest driver of first gold buys overall, accounting for 21% of the total. This key insight helped inform Glint’s media budget pacing strategy—particularly when seeing dips or rises in gold price—so they could capitalize on low-cost periods and pull back spend when gold price was predicted to rise. In addition, knowing the importance of gold price encouraged them to develop a creative strategy that speaks directly to what’s happening in the gold market.

Owned and earned media, such as tweets from @GlintPay and interviews on other channels, were the next largest driver – accounting for 18% of total first gold buys. This finding has been incredibly helpful for Glint in building a more robust organic strategy—they’ve increased posting volume and centered posting topics around market trends.

Key Takeaways
1. Resilient and validated budget allocation

Validating Robyn MMM predictions with incremental experimentation enabled Glint to optimize its paid media spend—for example, by shifting an additional 20% of overall Facebook budget toward iOS. The company is now conducting further testing to see how to scale these changes and push iOS spend without audience fatigue.

2. Media alignment to market factors

Including market factors like gold price in addition to paid media in the Robyn model has helped Glint in two ways. First, it informs their media budget pacing during gold price dips and spikes so they are able to spend the budget allocated to paid media when it is more efficient given the market context. Second, aligning the content of their creative assets to the main performance drivers as predicted by the model has allowed Glint to develop more effective campaigns that speak directly to what’s happening in the market.

3. Organic strategy

The Robyn MMM findings affirmed the importance of organic content, leading Glint to work more closely with Realtime on developing their organic content pillars and messaging.