Project period: January-July 2026. At the beginning of the year, digital was one of AVO bank's key customer acquisition sources. The bank was growing its credit card business and launching a new microloan product, so marketing had to solve two tasks at once: increase acquisition volume while maintaining economic efficiency. Historically, the performance model focused mainly on the upper funnel - app installs, registrations and the cost of acquiring a new user. At one stage, this was enough, but as the business scaled, the team needed to look deeper. A registration does not mean that a customer will complete KYC and scoring, receive approval, complete the application and actually receive a banking product. A business can acquire a large volume of low-cost users and still generate relatively few final customers; alternatively, it can acquire fewer upper-funnel users, improve traffic quality and generate more customers with an issued product. For the next stage of growth, AVO bank chose the second scenario and spent seven months systematically rebuilding performance around final business value.
Problem
Higher advertising investment brought more installs and registrations, but did not create proportional growth in customers who reached product issuance. Part of the audience dropped out at KYC, scoring, approval or application, so a low-cost upper-funnel user did not necessarily become a low-cost customer for the bank. The more performance scaled, the more visible this gap became. At the same time, a second problem emerged: saturation. Additional budget continued to generate conversions, but every next conversion became more expensive while the final business outcome grew more slowly than the upper funnel. It was easier to scale installs than to scale customers with a real banking product. The task was therefore not to abandon upper-funnel metrics, but to change the principle behind performance management. AVO bank needed to evaluate acquisition quality by how many users actually created business value, scale investment so that it grew the final outcome, and understand in advance where the limit of efficient growth was. The main goal was to increase the number of customers reaching product issuance while keeping acquisition cost under control.
Solution
1. Built an end-to-end funnel down to the real banking product. The team connected advertising data with the bank's internal DWH and sent deep events to AppsFlyer, Firebase and advertising systems. Instead of a few upper-funnel metrics, the team could now see one customer journey: install - registration - KYC - scoring and offer approval - application - product issuance - first transaction. Two events became especially important. Offer approval meant that the customer had completed identity verification and scoring and the bank was ready to offer a credit product. Product issuance meant that the customer had completed the process and actually received the banking product. Product issuance became digital's key business outcome: the team could now answer not only how much a registration cost, but how much a customer who truly reached the product cost. 2. Shifted advertising algorithms from traffic volume to customer quality. The transition happened step by step: campaigns first moved from installs and registrations to KYC, and then to offer approval. Moving immediately to the deepest event was not possible because deeper events occur less often and provide advertising systems with less statistical signal. New scenarios were therefore tested in parallel and scaled after proving their effectiveness. As a result, algorithms stopped looking only for people likely to install the app and started learning from users similar to those who successfully passed scoring and received approval. Offer approval became the optimal learning event - close enough to the business outcome, but frequent enough for stable algorithm learning. Advertising learns on offer approval; product issuance is the outcome the entire system is built to deliver. 3. Turned 360° from a reach activity into a performance scaling tool. After moving to deep-funnel optimisation, it became clear that simply increasing the performance budget eventually led to audience saturation. The team tested a different hypothesis: instead of only buying existing demand, could the bank first expand and prepare that demand? From April 1 to May 31, AVO bank ran a two-month 360° campaign, with the launch of the microloan as one of the main communication moments. The media mix included TV, OOH/DOOH, indoor, transport and other mass-reach touchpoints. The logic was simple: 360° creates and warms demand, performance receives a better-prepared audience, deep-funnel quality improves, the cost of the final result falls and digital gains more room to scale. The team did not attribute the whole uplift to 360°. To separate the effect of reach communication from the change in performance logic itself, it also compared the 360° period with the following Always On period under the same deep-funnel optimisation logic. 4. Built a model that answers where the next marketing dollar should go. After end-to-end analytics and capacity experiments, the team saw that a good CAC today did not show whether the same campaign would remain efficient after another budget increase. Together with Data Science, AVO bank developed a proprietary predictive ML model for marketing investment optimisation. It uses historical campaign data and the deep banking funnel, estimates campaign elasticity and builds a saturation curve for each campaign: where there is still an efficient growth zone, where the optimal range lies and where additional budget becomes economically inefficient. The model compares campaigns and recommends the most efficient split, answering a practical question: how many additional approvals will the next budget increment generate and what will the incremental conversion cost be? The model is regularly retrained; during testing, the team not only followed its recommendations but also deliberately moved budgets against them to check whether cost and conversion volume would change as predicted.
Results
The main outcome of the full transformation was a 40% decrease in the weighted calculated cost per customer who reached product issuance across the bank's overall funnel after the full transition to the new performance model, compared with the period before full implementation. The final customer with a product became the main business metric. KYC-to-approved-offer conversion increased by 50%, and the same pattern in paid performance traffic confirmed that the change was systematic. At comparable digital investment levels, after the move to deep-funnel optimisation, the cost of an approved offer across key performance channels was cut in half versus comparable pre-transformation periods. Upper-funnel actions could become more expensive because the improvement came from higher traffic quality, not from buying even more cheap users. During the period when deep-funnel optimisation and 360° worked together, performance investment was 18% lower than the average of comparable periods at the beginning of the year, while the number of customers with an issued product was 56% higher. Because two changes were active at the same time, the team ran an additional check to isolate the 360° effect: under the same deep-funnel optimisation logic, average approval cost across paid performance traffic during 360° was 24% lower than in the post-360° Always On period. This showed that reach communication had become part of the overall performance model and increased its effective capacity. The ML model forecast matched actual performance in 98% of tested scenarios - across expected approval volume, cost changes and response to budget reallocation. In seven months, AVO bank changed four layers of digital marketing: measurement moved from the upper funnel to the final product; optimisation moved from traffic volume to customer quality; scaling moved from buying existing demand to expanding it; and forecasting moved from reacting to past results to managing future return. AVO bank's performance marketing stopped being only a traffic acquisition system and became a managed growth system - from the first advertising contact to a customer with a real banking product.
