Marketing attribution models explained for non-data founders
Last-click is lying to you. First-click is also lying. Here’s a plain-English tour of attribution models and which one to actually use when you don’t have an analyst on the team.
Every founder we work with has, at some point, killed a channel because “the dashboard” said it wasn’t working, only to watch overall sales drop the following month. The dashboard was not lying on purpose. It was just using an attribution model that quietly favoured one channel over another. Here is the short, opinionated tour.
The five models you will hear about
- Last-click. The default in most ad platforms. Gives all credit to the last touchpoint before purchase. Wildly overstates Google Search and brand campaigns.
- First-click. Gives all credit to the first interaction. Overstates top-of-funnel channels like Meta prospecting, Reels, and SEO.
- Linear. Splits credit evenly across every touchpoint. Sounds fair, hides reality.
- Time-decay. More credit to touchpoints closer to conversion. A reasonable middle ground for short sales cycles.
- Data-driven. Algorithmic credit assignment based on which paths actually converted. Powerful, but only if you have enough conversions and clean data.
Why last-click lies the loudest
A customer sees your Reel on Tuesday, Googles your brand on Saturday, clicks the Google Ad, buys. Last-click hands the win to Google Ads. You then increase Google budget, decrease Meta budget, and watch your funnel shrink because the Reel was the actual demand-creator. This pattern is so common it is the single biggest cause of bad budget decisions we see.
What model to use when
- Short sales cycle, transactional product.Time-decay or data-driven in GA4. Last-click is fine as a sanity-check, not a decision-maker.
- Long sales cycle, B2B or high-ticket.Position-based (40-20-40) so first-touch and last-touch both get credit. Layer in self-reported attribution — literally ask buyers how they found you.
- Heavy organic and content. Use a custom data-driven setup or accept that you will need to triangulate across GA4, ad platforms, and incrementality tests.
The cheat code: triangulation
No single model is right. The strongest founders we work with cross- reference three signals weekly: GA4 data-driven attribution, platform-reported conversions, and a one-question post-purchase survey asking “where did you first hear about us”. When the three agree, act. When they disagree, that is your signal to investigate, not to panic.
What to stop doing immediately
Stop adding up Meta-reported conversions and Google-reported conversions and treating the sum as truth. They double-count. Stop killing a channel based on one model. Stop trusting an agency that reports only on their own platform’s attribution.
How we help at The Nerdish Mic
We build attribution stacks for founder-led brands that combine GA4, server-side data, and post-purchase surveys into a single weekly view you can actually trust. If your channel-mix decisions currently feel like rolling dice, we’d love to fix that for you.