The attribution model you’re using right now is making budget decisions for you. That’s not a metaphor. When your analytics platform assigns 100% of a conversion to one touchpoint and 0% to the five touchpoints that came before it, it’s telling you — implicitly but forcefully — where your money should go. Most of the marketers I talk to have never actively chosen their attribution model. They’re on last-click because it’s the default. That default is a budget opinion, and it’s almost certainly the wrong one.
We published a short piece on this at the agency recently — “Attributionsmodeller — är det något för mig?” — aimed at clients who’d never heard the term. This is the longer version. The practitioner’s version. If you already know what an attribution model is but haven’t committed to which one fits your business, this is the framework for deciding.
The five models and what they actually do
Every attribution model answers one question: when a user converts after multiple touchpoints, who gets credit? The answer changes everything downstream.
Last-click attribution gives 100% of the credit to the final interaction before conversion. This is the default in Google Analytics and the most widely used model in the market. It’s also the most misleading for any business where the buying journey involves more than one session. Last-click systematically overvalues branded search, retargeting, and direct traffic — channels that catch people who were already going to convert. It systematically undervalues the channels that created the demand in the first place.
Last non-direct click is Google Analytics’ reporting default (as distinct from the MCF default). It ignores direct traffic as a final touchpoint and gives credit to whatever came before it. Slightly better than pure last-click because it doesn’t credit people typing your URL into their browser. Still single-touch. Still hiding the full journey.
First-click attribution gives 100% of the credit to the first interaction. The mirror image of last-click. It overvalues awareness channels — display, social, organic search for broad terms — and ignores everything that happened between first visit and conversion. Useful exactly once: when you want to understand which channels bring new users into your ecosystem. Not useful for budget allocation because it can’t see the middle or the end of the journey.
Linear attribution splits credit equally across every touchpoint. If there were five touchpoints, each gets 20%. This is the simplest multi-touch model and a dramatic improvement over any single-touch approach. Its weakness is that it treats a random display impression the same as a high-intent branded search click. But for organizations moving from last-click to anything better, linear is a defensible starting point.
Time-decay attribution gives more credit to touchpoints closer to the conversion. The logic is that interactions nearer to the purchase decision were more influential. This tends to be a reasonable approximation for transactional businesses with short consideration windows. It still undervalues the top of the funnel, but less aggressively than last-click.
Position-based (U-shaped) attribution gives 40% credit to the first interaction, 40% to the last, and distributes the remaining 20% across everything in between. This is the model I recommend most often as a starting point for considered-purchase businesses. It acknowledges that both demand creation and demand capture matter, while still giving some credit to the nurture stages.
There’s also data-driven attribution, which uses your actual conversion path data to build a custom model. In Google Analytics, this requires a minimum volume of conversions and paths to generate statistically meaningful weights. Most mid-market accounts don’t hit the threshold. If you do, it’s worth testing — but treat it as a hypothesis, not an oracle. The model is only as good as the data going in, and cookie-based tracking has real gaps.
Matching model to business
The right attribution model depends on your conversion cycle, not your industry label.
Short-cycle transactional businesses — e-commerce with low average order value, impulse purchases, commoditized products. The buying journey is short. Two or three touchpoints, maybe one session. Last non-direct click or time-decay are reasonable here because the journey is compressed enough that single-touch approximations do less damage. Linear works too. The difference between models matters less when the path is short.
Considered-purchase businesses — B2B services, high-value e-commerce, SaaS, anything where the buyer researches across multiple sessions over days or weeks. This is where attribution model choice matters most. Last-click will tell you to cut your awareness spend and double down on branded search, which is exactly backwards. Position-based gives you a more honest picture: it credits the channel that introduced the buyer AND the channel that closed them, while acknowledging the middle. If you’re in this category and still on last-click, you’re misallocating budget. Full stop.
Subscription and LTV businesses — where the initial conversion is just the beginning and lifetime value is what matters. Attribution gets harder here because the model needs to account for not just “who brought them in” but “who brought in the customers that stay.” Most analytics setups can’t answer this without offline data integration. The pragmatic move is to start with position-based for acquisition attribution and build a separate LTV analysis that segments by acquisition channel. Trying to solve both in one attribution model usually means solving neither well.
The implementation reality
In Google Analytics right now, you can compare attribution models in the Model Comparison Tool under the Multi-Channel Funnels section. This is the single most useful report for attribution decisions and almost nobody uses it. Pull up your conversion data, switch between models, and watch how channel valuations shift. The channel that looks like your best performer under last-click might drop to third under position-based. That gap IS your misallocation risk.
The practical ceiling is data sampling. Once your reporting window exceeds a certain data volume, Google Analytics samples the Multi-Channel Funnels reports. You get directionally correct data, not exact numbers. For high-volume accounts, this means keeping your analysis windows shorter or accepting approximation. It’s still more accurate than using last-click without question.
The other constraint is cross-device. Analytics tracks cookies, not people. A user who researches on mobile and buys on desktop looks like two separate users with two separate journeys. There are workarounds — User ID tracking if you have authenticated sessions — but most organizations have significant cross-device blind spots. Attribution models can only attribute what they can see.
When to revisit
Set a calendar reminder to re-evaluate your attribution model quarterly. Not because the right model changes often, but because your channel mix, campaign strategy, and conversion paths evolve. A model that was right six months ago might be overcrediting a channel you’ve since scaled back, or undercrediting a new channel you added.
Also revisit after any significant campaign change — launching a new paid channel, shifting budget between awareness and conversion campaigns, changing your landing page strategy. Each of these changes the shape of the customer journey, and the attribution model should still match that shape.
The question I keep asking clients: if you switched from last-click to position-based today, which channel would gain the most credit and which would lose the most? If you don’t know the answer, run the comparison. The gap between those two numbers is the size of the budget conversation you’ve been avoiding.
Has anyone here gone through this exercise and been surprised by the result? I’m particularly curious about the B2B side — whether the branded-search-overvaluation pattern holds as strongly in longer sales cycles as it does in the e-commerce cases I see most often.