KOHO is a Canadian financial technology company building a modern financial platform to help users take control of their financial lives. Like many fast-growing fintechs, KOHO faces significant measurement challenges. Customer value is not realized at signup, but rather, users move through a multi-stage journey including registration, funding, deeper engagement, and eventually, lending products, all of which are missed by traditional attribution methods.

It was this challenge that led KOHO to explore media mix modeling, eventually starting to work with Recast in early 2024. Together, they piloted a multi-stage MMM built to reflect KOHO’s unique acquisition funnel. The partnership started with careful trust-building, including validating their model’s accuracy through backtesting, making small budget adjustments, and then monitoring the model’s output. It has grown into an engine for generating marketing hypotheses and placing larger marketing bets.

KOHO’s multi-stage model has allowed them to uncover which channels are driving the most valuable customers, not just efficient signups, while systematic validation has helped them build the trust they need to act on those insights. And by incorporating experiments with GeoLift by Recast, KOHO has turned medium swings into major marketing bets – backed by hard evidence and an incrementality platform they can rely on.

Building Trust Through Systematic Validation

KOHO had previously incorporated media mix modeling into their measurement strategy, and eventually partnered with Recast to build on that foundation. Within a short period, KOHO’s Recast model was deployed and began producing insights.

The first outputs from the model, noted KOHO’s Sr. Director of Performance Marketing Yassine Bakri, were “pushing us to take some very big swings.” But those kinds of swings, he explained, required a structured validation process instead of just trusting the model. Recast encourages this type of disciplined approach and was eager to support KOHO in validating and verifying the model’s recommendations.

KOHO’s trust-building process included multiple layers:

• Build Confidence Gradually: Instead of implementing full budget shifts, they started with 30% of recommended changes. “The model is recommending shifting $1 million of spend into a specific campaign. We started by shifting about 30% of that,” Yassine explains.
• Active Monitoring: The team tracked backend registration volumes and other metrics daily to confirm their Recast model’s forecasts matched reality. As they saw “steady increases” in conversion volumes on the backend that aligned with the model’s forecasts, confidence grew.
• Backtest Review: Working with Recast’s data scientists, the team analyzed their Recast model’s historical forecast accuracy, asking the model to make predictions on data that it had not seen before. This ensured that the model was not overfitting to trends, but was actually picking up on causal signals.

Yassine approached model results the exact way that Recast recommends: be default skeptical in the beginning, and follow a deliberate process to validate the model’s forecasting accuracy. Then, after building trust in the model, begin placing larger marketing bets with its recommendations.

Multi-Stage Insights: Uncovering Paid Social’s True Impact

An added challenge that KOHO faced was in the nature of their customer journey. A customer who downloaded their app but took no further action was much less valuable than a customer who became a power user. However, previous measurement methods didn’t include this crucial nuance.

Specifically, there are three critical stages of the customer journey as defined by Yassine:

• Registration: Users sign up for the KOHO app
• Funded Account: Users add money, signaling intent to use the platform
• Power User: Users engage with lending products or maintain high balances

Recast’s multi-stage model gave KOHO the ability to actually understand which upper funnel channels were driving conversions of their most valuable cohorts, Power Users and funded accounts.

And for Yassine and his team, the results were quite surprising:

“We can see that upper funnel media has some impact on registrations, which is expected,” Yassine noted. “However, it has way more impact on funded accounts.”

The data showed that while multiple channels performed similarly for driving top-of-funnel registrations, certain paid social channels emerged as the clear winner for funded accounts – a metric that matters much more to KOHO’s business. The model revealed that campaigns from these particular channels were about 10% more efficient than average for registrations but about 20% more efficient for funded accounts.

Model insights also exposed how certain lower funnel channels were “over-claiming credit” in their last-touch system. “Paid social can work on the whole awareness and consideration and even pushing people to conversion,” Yassine explains. “But we’d often see those sales attributed to channels further down the funnel.”

These insights fundamentally changed how KOHO thought about channel value. Channels that looked less efficient for registrations were actually driving the customers most likely to fund accounts and become profitable users – an insight whose associated optimizations made an immediate impact on their ROI.

Leaning into Insights from GeoLift by Recast

As Yassine said “We didn’t onboard Recast to get 1% more growth – we’re looking to take bigger swings that will unlock serious growth.”

To have confidence in these bigger swings, Yassine and his team started using Recast’s new experimentation platform, GeoLift by Recast, to have an additional – and statistically powerful – signal to justify some very large changes in their budget mix.

Specifically, they designed a geographic experiment to test a paid social channel’s incrementality through a spend-up test, and the results were incredibly reassuring: the GeoLift by Recast findings came out “almost equivalent to the model.”

This validation milestone transformed how KOHO approached marketing optimization. They now had a proven framework:

• Use MMM to identify high-impact opportunities
• Start with conservative tests (30% of recommendations)
• Monitor performance against model predictions
• Validate big swings with lift testing
• Scale successful tests with confidence

“Right now with everything that we are seeing, there is no reason why we should not trust the model and make even larger swings,” Yassine notes. Lift testing with Recast will be a key part of KOHO’s measurement framework going forward, and will be used both for model validation and to identify new areas of growth. 

Results and the Path Forward

KOHO’s approach to marketing measurement has delivered tangible results, and because of steps they’ve taken to build trust in the model, they’re able to act on the insights in ways that most marketing teams are not.

The company now incorporates MMM insights into monthly planning sessions, combining model forecasts with Google Performance Planner and internal models to create unified growth projections.

For KOHO, Recast has become more than a measurement tool – it’s a strategic partner in building sustainable growth. As KOHO continues to scale in the competitive Canadian fintech market, their multi-stage measurement approach provides a crucial advantage. They can optimize not just for immediate conversions, but for the customer behaviors that drive long-term value. And with each test, each validation, and each successful bet, their confidence in making data-driven decisions grows stronger.

Do you have a complex media mix like KOHO and think your business is ready for Recast’s incrementality platform? We’d love to talk.