Here’s a scenario that plays out constantly in marketing reviews: Google Ads reports 200 conversions last month. Meta reports 180. Total across platforms: 380. Your CRM says you had 190 new customers.
Something isn’t adding up. Both platforms are right within their own model. Both platforms are taking credit for every conversion they touched. The reality is somewhere in between.
This is the attribution problem. Multi-touch attribution (MTA) is the attempt to solve it — to understand what actually caused each conversion, across all the touchpoints that contributed to it, so you can allocate credit (and budget) more accurately.
AI has changed how MTA works and how good it can be. Here’s the honest picture.
Why Last-Click Attribution Is Misleading
For years, most digital marketing ran on last-click attribution. The channel that got the last click before conversion got 100% of the credit.
The problem is obvious: a customer might have seen your brand video on YouTube, clicked a Meta prospecting ad, read an organic blog post, and then clicked a branded Google search ad to buy. Last-click gives all credit to Google branded search — the easiest, lowest-cost conversion that happened to be last.
This systematically over-rewards bottom-funnel channels and under-rewards top and mid-funnel activities. Marketing teams optimising on last-click attribution consistently underinvest in brand, awareness, and demand generation — because those channels appear to ‘not convert’ even when they’re doing the heavy lifting of bringing customers into the funnel.
The result over time: acquisition costs rise as bottom-funnel capacity saturates, brand health weakens because nobody’s investing in building it, and the channels doing real demand creation are starved of budget because the attribution model can’t see what they’re doing.
Data-Driven Attribution: The AI Upgrade
Google’s Data-Driven Attribution (DDA) — now the default in Google Ads and GA4 — uses machine learning to distribute conversion credit across touchpoints based on their actual contribution to conversion probability.
Rather than applying a fixed rule (last click, linear, time decay), DDA compares two populations:
- Users who converted after being exposed to a specific set of touchpoints
- Similar users who were exposed to the same touchpoints minus one specific ad
The difference in conversion rates tells the algorithm how much that specific touchpoint contributed. It repeats this analysis across millions of touchpoint combinations to build a probabilistic credit model that’s specific to your account.
In independent studies comparing DDA to last-click:
- Upper-funnel channels (display, video, social awareness) typically receive 35-55% more credit in DDA than in last-click
- Branded search receives 25-40% less credit (it gets much of its conversions from demand that other channels created)
- Overall budget reallocation based on DDA signals tends to improve new customer acquisition by 12-22%
For Google Ads specifically: DDA is now the default and requires only that you have sufficient conversion volume (typically 3,000+ conversions in 30 days for full model quality). Most accounts running Smart Bidding are already using it.
The Walled Garden Problem
Here’s the fundamental limitation of any platform-native attribution, including DDA: Google can only see what happens on Google. Meta can only see Meta.
A customer journey that goes YouTube > Meta > Google Search > Purchase is invisible to any single platform. Google sees the search and the purchase. Meta sees the Meta touchpoint and the purchase (if it has the pixel). Neither can see the full sequence.
Third-party multi-touch attribution attempts to solve this by:
- Using a first-party pixel on your site that captures all channel touchpoints
- Building path-to-conversion analysis using your own data
- Applying attribution models to distribute credit across the full journey
The tools in this space: Northbeam, Triple Whale (e-commerce), Rockerbox, Wicked Reports (B2B). All use variations of data-driven or rules-based attribution applied to first-party touchpoint data.
What they can see:
- All on-site sessions and their source/channel
- The sequence of touchpoints before conversion
- Time between touchpoints and conversion
What they can’t see (the walled garden problem):
- Ad impressions that didn’t result in clicks (view-through effects)
- Platform-level audience exposure data
- What happened inside Meta’s or Google’s platform before the user visited your site
This is a genuine limitation. No third-party MTA tool fully solves it because it requires data that platforms don’t share.
Marketing Mix Modelling: The Complementary Approach
Marketing Mix Modelling (MMM) takes a different approach to the same problem. Instead of tracking individual user journeys, it analyses the statistical relationship between marketing spend across channels and business outcomes (revenue, new customers, leads) at an aggregate level.
Traditionally, MMM was expensive, slow (results in 3-6 months), and required specialist consultants. AI has changed this significantly.
Google’s Meridian (open-source, launched 2025), Meta’s Robyn, and commercial tools like Measured and Analytic Partners now provide AI-powered MMM with:
- Results in days rather than months
- Ongoing model updating as new data arrives
- More accessible implementation (Meridian is free)
- Better handling of digital channel fragmentation
What MMM gets right that MTA gets wrong: it can attribute value to channels with no direct response signal — TV, out-of-home, brand awareness campaigns, PR. These influence conversion behaviour but leave no trackable digital footprint. MMM’s statistical approach can detect their contribution.
What MTA gets right that MMM gets wrong: individual-level insight. MMM tells you ‘Meta drives 23% of revenue.’ MTA tells you ‘Users who see a Meta ad and then search branded on Google convert 3.2x better than users who search branded without Meta exposure.’ Different insights, different uses.
The Attribution Stack We Use
For clients with sufficient scale (typically Rs. 75L+/month in spend), we operate a three-layer attribution stack:
Layer 1: Platform-native DDA (Google) + last-click with view-through window (Meta). Day-to-day campaign management.
Layer 2: Third-party first-party data MTA (Northbeam for performance, Triple Whale for e-commerce). Weekly cross-channel budget allocation decisions.
Layer 3: Quarterly MMM update using Meridian. Strategic channel mix decisions and offline/upper-funnel investment evaluation.
Each layer answers different questions. Platform DDA optimises daily campaign decisions. Third-party MTA guides weekly budget allocation. MMM informs quarterly strategic investment.
Using only one is a shortcut that will eventually produce wrong decisions.
Benchmarks: What Changes When You Fix Attribution
In our client portfolio, the transition from last-click to proper multi-touch attribution has produced consistent patterns:
- Top-funnel investment increases: Average 18% more budget allocated to upper-funnel after MTA implementation reveals its contribution
- Branded search budget decreases: Average 12% reduction as brands realise branded search is capturing, not creating, demand
- New customer acquisition improves: Median 21.4% improvement in new customer rate at equivalent blended ROAS, because investment has shifted toward demand creation
- Blended ROAS appears to decline initially: This is normal and expected. Removing inflated last-click credit from easy bottom-funnel conversions will make overall ROAS look lower. The business gets healthier.
The last point is the hardest to manage with clients. When you fix attribution, performance looks worse before it looks better. The discipline is to trust the more accurate model even when it shows a less flattering picture.
The Starting Point
If you’re currently on last-click and want to move toward better attribution:
- Enable DDA in Google Ads and GA4 — free, significant improvement, no implementation work required if you’re already running Smart Bidding
- Set Meta’s attribution window to 7-day click, 1-day view — the most balanced view for most businesses
- Audit your CRM to GA4 conversion match rate — if it’s below 75%, fix first-party tracking before investing in MTA tools
- Deploy a third-party MTA tool at Rs. 75L+/month spend
- Run a quarterly MMM if you’re at 2 crore+/month and investing in brand channels
Attribution won’t tell you what to do. It’ll tell you what’s actually happening. That’s the value: making decisions based on reality rather than platform mythology.
