Last-click attribution is the default reporting model for most ecommerce businesses. It is also one of the fastest ways to destroy a paid media strategy while your dashboards tell you everything is fine.
If you have ever found yourself questioning why a channel "isn't working" based on your attribution data, then switched budget away from it, and then watched overall revenue gradually soften – there's a decent chance last-click attribution was steering you wrong the whole time.
What last-click attribution actually does
Last-click attribution gives 100% of the credit for a sale to whatever touchpoint the customer clicked immediately before converting. Every other interaction they had with your brand across the entire journey gets nothing.
So if a customer finds you through a Meta ad, comes back a week later through a Google Shopping result, and then converts after receiving a promotional email – the email gets all the credit. Meta gets zero. Google gets zero. According to your reporting, email is your revenue engine and paid social is burning money.
That's not what happened. That's just what your attribution model recorded.
The sale almost certainly required all three. The Meta ad introduced the brand. The Google search confirmed it. The email closed it. But last-click only sees the final step, and that distortion shapes every budget decision that follows.
The slow-burn problem nobody notices until it's too late
The insidious thing about last-click attribution isn't what it does immediately. It's what it pushes you to do over time.
Because acquisition channels (paid social, SEO, organic content) are almost never the last click, they consistently look underperforming in last-click reports. So you pull budget. You reallocate to email and branded search, which do show up as the last click because they sit at the bottom of the funnel. Your cost-per-acquisition appears to improve. Your reporting looks better.
And then, three to six months down the line, new customer volumes start declining. The email database stops growing at the same rate, and branded search volume drops. The "efficient" channels you doubled down on are producing less, because there are fewer new customers entering the top of the funnel to feed them.
This is one of the most common growth plateaus in ecommerce. Brands harvest the demand their acquisition channels built, stop refilling it, and mistake the short-term efficiency gain for a sustainable strategy.
Why your platform numbers never add up
There's a related problem that makes this even messier: every platform claims all your revenue.
Meta reports it drove £80,000 last month. Google Ads says £65,000. Klaviyo says £55,000. Add those together and you're looking at £200,000 in attributed revenue… but your Shopify dashboard shows £95,000 in actual sales.
This happens because each platform runs its own attribution window and claims credit for any conversion where a customer touched that platform. When someone sees a Meta ad, clicks a Google Shopping result, and converts from an email, all three platforms log a conversion. All three claim it.
Your channel-level ROAS (return on ad spend) numbers look extraordinary. Your actual margin tells a different story. And if you're making budget decisions based on those platform-reported figures, you're essentially letting each channel make the case for its own importance, which is not a neutral exercise.
First-click vs last-click vs data-driven attribution: what each model is actually doing
There are six main attribution models in common use. They're not all equally useful, and a couple of them have fairly specific situations where they make sense.
Last-click you already understand. Good for understanding what closes a sale. Poor for everything else. Despite being the default across most platforms historically, it's a genuinely bad fit for any business running activity across more than one channel.
First-click is the opposite. All credit goes to whatever introduced the customer. Useful if you're specifically trying to understand how new customers discover you, but it ignores everything that happened between discovery and purchase, which is often a lot.
Linear splits credit equally across every touchpoint in the journey. More honest than either single-touch model, though it does treat a brief display impression the same as a cart abandonment email, which isn't quite right either. Serviceable rather than ideal.
Time decay weights credit more heavily towards the touchpoints that happened closest to the conversion. Works reasonably well for ecommerce with shorter, faster purchase cycles. Still tends to undervalue the channels that do awareness and consideration work earlier in the journey.
Position-based (U-shaped) gives 40% of the credit to the first touchpoint, 40% to the last, and distributes the remaining 20% across everything in the middle. It's a practical, opinionated model that acknowledges both acquisition and conversion matter, which for most ecommerce businesses is true.
Data-led attribution is a different category entirely. Rather than applying fixed rules, it uses statistical modelling to assess which touchpoints in a conversion path actually influenced the outcome, comparing paths that converted against paths that didn't. No fixed percentages. The model learns from your actual data. GA4 now uses this as its default model, replacing last-click in 2022, which is worth knowing if you haven't updated your GA4 settings recently.
Which attribution model is most accurate?
Data-led attribution, by a significant margin – but with a catch. It needs volume to work. GA4's model generally becomes reliable once you're generating a few hundred conversions per month. Below that threshold, there isn't enough data for the statistical comparison to be meaningful, and you're better off using position-based or time decay as a working model.
The more useful reframe is this: no single attribution model tells the complete truth, and chasing the perfect one is a distraction. What you're actually trying to do is stop using a model you know is wrong (last-click) and replace it with one that at least reflects how your customers actually behave, which is to say, across multiple sessions and channels before they buy.
One additional metric worth adding to your regular reporting: Marketing Efficiency Ratio, or MER. Total revenue divided by total ad spend, across all channels combined. No platform overlap. No attribution window arguments. Just a clear read on whether your overall marketing investment is working. If your MER is holding or improving, your system is broadly healthy. If it's declining while your channel-level ROAS numbers look strong, something in the full picture is broken – and that gap is often where the attribution distortion is hiding.
Why this matters more when you're running acquisition, conversion and retention together
Besides being a reporting issue, attribution changes how you invest in each stage of the customer journey, which has compounding consequences.
At Gravytrain, we work on the basis that acquisition, conversion and retention are three stages of the same strategy, not three separate disciplines. What you learn about customers at the retention stage should inform who you're acquiring at the top. What happens in conversion tells you whether acquisition is bringing in the right people. They feed each other.
Last-click attribution breaks that connection. It assigns all value to whichever stage happens to produce the final click, so you end up making investment decisions for each stage in isolation – often defunding acquisition because it doesn't show up as the last click, over-attributing to CRM because it frequently does, and then missing why conversion is declining because you haven't considered what changed upstream.
The question worth asking isn't "which channel is driving revenue?" It's "what is each channel supposed to do in the journey, and am I measuring it against that standard?" Paid social's job is to bring the right people in. It should be measured on reach, new visitor volume, and the quality of that audience downstream, not on last-click conversions, which it will almost never win.
What to do about it
Switch your GA4 attribution model to data-led if you haven't already. It's in Admin > Attribution Settings. This changes how GA4 reports on conversions going forward and gives you a more honest read on channel contribution than last-click has been providing.
Stop adding up revenue numbers across platforms. Your Shopify total is the ground truth. Use platform-level data to understand contribution and efficiency within each channel, not to total up a revenue figure that will always be larger than what actually happened.
Build MER into your monthly review. Revenue divided by total ad spend. Track it month on month. It cuts through the platform noise and gives you a stable view of whether the marketing system as a whole is working.
Before reducing a channel based on weak attribution data, pause it and measure the effect on everything else. This is called an incrementality test, and the results regularly surprise brands who've been running on last-click. Paid social campaigns that appeared to generate almost nothing often produce a clear decline in overall revenue when they go dark, because they were doing acquisition work that last-click never credited them for.
And map your actual customer journey before deciding which attribution model fits your business. If most of your customers convert within 24 hours of discovering you, time decay probably makes sense. If your cycle is two to four weeks, you need a model that gives meaningful credit to the touchpoints at the beginning of that journey, because those are often the most important ones.
If your current reporting is raising more questions than it answers, we can help you build a clearer picture across acquisition, conversion and retention. Get in touch.
