Marketing Mix Modelling for Financial Services

Financial services brands spend heavily on media but often struggle to measure what is working. MMM handles the long sales cycles, regulatory constraints, and blended product portfolios that make other measurement approaches unreliable.

Insurance companies, banks, lenders, and investment platforms share a common measurement challenge. The customer journey from first awareness to product purchase can span weeks or months. A prospect sees a TV ad in January, searches for quotes in March, and converts in April. Digital attribution tools attribute the sale to the March search and ignore everything that came before. MMM captures the cumulative effect of all media over time, which makes it far better suited to long purchase cycles.

What makes financial services measurement different

Financial services marketing operates under FCA rules (in the UK) and equivalent regulations in other markets. You cannot use certain types of personalised targeting, some data cannot be passed to third parties, and your measurement approach may need to be explainable to a compliance team. MMM is entirely privacy-safe because it works with aggregated data, not individual customer records. That alone makes it more suitable for regulated categories than identity-based attribution.

Financial products also have very different margins and customer lifetime values. A current account customer might be worth very little in year one but thousands over a ten-year relationship. A personal loans customer might generate significant short-term revenue but churn quickly. This means the output metric you model against matters. Modelling against applications is different from modelling against completed accounts, which is different again from modelling against twelve-month revenue. Choose the right dependent variable before you start.

The typical financial services media mix

• TV: still dominant for insurance and banking brand-building, particularly for price-comparison-driven categories like motor and home insurance
• Paid search: high spend, high CPC (cost per click), but often over-attributed because it captures intent created by other channels
• Comparison sites: major volume driver for insurance, though often treated as a distribution channel rather than a media channel
• Direct mail: remains effective for mortgage and investment products targeting older demographics
• Radio: cost-efficient reach builder, particularly for local brands or regional building societies
• Digital display and video: growing share, particularly for challenger banks targeting younger customers

Data challenges and what to pull together

The biggest data challenge in financial services MMM is connecting media spend to business outcomes. Most brands have media spend data sitting in agency dashboards and product sales data sitting in completely separate CRM or core banking systems. Pulling them into a weekly time series often requires coordination between the marketing, analytics, and IT teams. Start that conversation early, because data extraction from core banking systems can take longer than expected.

You will also need to capture macro-economic factors. Interest rate changes, house price movements, and unemployment levels all affect demand for mortgages, loans, and savings products independently of media. If March 2023 saw a sharp rise in savings applications because the Bank of England raised rates, the model needs to know that. Otherwise it will incorrectly attribute that volume to whatever media was running at the time.

Comparison site spend behaves differently from direct media. Include it as a variable, but treat it as a performance channel with a very short adstock decay rather than a brand-building channel. The model will tell you whether that assumption is correct.

Seasonality in financial services

Financial services has clear seasonal patterns, though they vary by product. Insurance renewals peak at the anniversary of the original purchase date, creating a rolling seasonal pattern that differs by product line. Mortgages spike in spring and summer when the housing market is most active. ISA (individual savings account) applications spike in February and March ahead of the tax year end. Personal loans have a smaller but consistent January spike linked to post-Christmas debt consolidation.

Common measurement mistakes

The most common mistake is using cost-per-application or cost-per-quote as the primary efficiency metric, without accounting for the quality of those applications. Media that drives a high volume of declined applications or low-lifetime-value customers looks cheap in the funnel but is expensive in practice. MMM can be built to model against quality-adjusted outcomes if your data allows for it.

A second common mistake is failing to account for the adstock of TV. Financial services brands often run heavy TV campaigns for a month and then go dark. The TV exposure continues to drive applications for several weeks after the campaign ends. If you only look at sales during the campaign period, you underestimate TV's contribution and overestimate the channels that were still active when the delayed TV effect converted.

How to use MMM outputs in financial services

Once the model is built, it gives you a contribution breakdown by channel, a set of response curves (which show what happens to sales if you increase or decrease spend in each channel), and a budget optimiser. For financial services brands, the response curves are particularly useful for TV planning decisions, because they show the point of diminishing returns clearly. Most insurance brands are surprised to find that their TV investment is already above the optimal level on a cost-per-sale basis, while search and display have room to grow.