A borrower who takes a home loan from an NBFC in November was not influenced by a single marketing touch. They may have first seen a Google Display ad in August, clicked a non-branded search result in September, opened a lifecycle email in October, called the DSA they saw on WhatsApp in October, visited the branch website directly on the day they decided to apply, and submitted a form via the branded search result they typed when they finally searched for the institution's name. Six touches. One funded loan. Which channel gets the credit? The Marketing Analytics Agent AI runs five attribution models simultaneously and presents the answer differently to each stakeholder who needs it.
Why attribution matters — and why no single model is correct
Attribution is not a mathematical question — it is a strategic question about which channels the institution wants to incentivise and maintain. A last-touch attribution model that gives 100% of the credit to branded search (the channel the borrower used to submit the application) tells the institution that branded search is its most valuable channel. It also tells the institution that all the awareness-building channels that created the consideration and intent that led to that branded search are valueless — which is incorrect and will lead to defunding the top-of-funnel channels that feed branded search. A first-touch model makes the opposite error. A linear model treats all six touches as equally valuable, which is also incorrect because a Google Display impression that the borrower glanced at in August is not equally contributory to a conversion as the DSA call that answered their specific product questions.
The Marketing Analytics Agent AI runs five models simultaneously — first touch, last touch, linear, position-based (U-shaped), and data-driven (algorithmic, based on historical patterns of which touch combinations actually produce funded loans) — and presents the one most relevant to each business question being asked.
A live 6-touch attribution: Ananya Krishnamurthy's home loan journey
Which model to use for which decision
| Attribution model | What it credits | Use for this decision | Risk if misapplied |
|---|---|---|---|
| First-touch | The channel that first created awareness — Display, non-brand search, social prospecting | Upper funnel budget decisions: "How much to spend on awareness vs conversion channels?" | Over-investing in awareness channels; defunding conversion channels that close the deals awareness creates |
| Last-touch | The channel at the moment of application — typically branded search or direct | Ad platform optimisation (Google Smart Bidding uses last-touch by default) · Short-term conversion campaigns | Defunding all channels that build intent but don't receive the final click · Branded search cannibalises organic |
| Linear (equal weight) | All channels equally | Baseline comparison · Identifying which channels are always in the journey regardless of length | Over-credits early passive touches (Display impression) equally with conversion actions (DSA call) |
| U-shaped (position-based) | 40% first touch, 40% last touch, 20% split across middle touches | Budget allocation when both awareness and conversion are intentional investments · Most balanced for lending | May over-credit first and last touches vs the middle touches (email, DSA call) that often determine whether the journey completes |
| Data-driven (algorithmic) | Touches that historically have the highest correlation with funded loan completion, based on LOS data | Actual budget decisions for maximum ROMI · Channel investment aligned to what actually causes conversions | Requires 90+ days of funded loan data to be statistically reliable · Early in campaign, may have insufficient data |
For Ananya's journey, the data-driven model assigns the highest credit to the DSA call (28%) and the lifecycle email (24%). This reflects the historical pattern in the institution's data: borrowers who engaged with a DSA personally and received a triggered email within the same 30-day window completed to funded loan at 2.8 times the rate of borrowers who did not have both of these touches. The Display ad and non-brand search get lower data-driven credit despite appearing earlier in the journey — historically, journeys that begin with Display but never reach a DSA or email touch rarely fund. The final branded search click gets only 16% in the data-driven model because branded search is predominantly navigational for buyers who have already decided — it adds convenience, not conviction.
The institution that uses last-touch attribution will systematically defund the channels that build the relationships that make the last touch possible
If the Marketing Analytics Agent AI reported only last-touch attribution, the dashboard would say: Google Branded is responsible for 100% of Ananya's conversion. The DSA call that turned a curious browser into a committed applicant would receive zero credit. The lifecycle email that introduced the pre-approval offer and made Ananya aware she was already pre-qualified would receive zero credit. The institution would cut DSA commissions and email investment as "unattributed spend" and pour more budget into branded search — only to find that branded search volume falls because the awareness and intent-building channels that feed it have been defunded. The Marketing Analytics Agent AI presents all five models not to confuse the decision-maker but to prevent the attribution model choice from being made by default — because the default, last-touch, is almost always wrong for understanding how lending relationships are built.