Why Is Last-Click Attribution Still Costing Ecommerce Brands Money in 2026?

Last-click attribution is still the default measurement model in most ecommerce marketing stacks, and it is still getting budget decisions wrong in 2026. This article explains why, what it misses, and what has replaced it for brands that want an accurate picture of where revenue actually comes from.

What Last-Click Attribution Is and How It Works

Last-click attribution credits a conversion entirely to the last ad a shopper clicked before buying. If a customer saw a TikTok ad, watched a YouTube pre-roll, ran a branded search on Google, and then clicked a Google Shopping ad to buy, last-click gives 100 percent of the credit to Google Shopping. TikTok and YouTube get zero.

This model was accepted practice for a decade because click data was plentiful, cookies were reliable, and most purchase journeys were short. None of those conditions still hold in 2026.

Three Specific Ways Last-Click Misleads Budget Decisions

1. It over-credits the bottom of the funnel

Channels that capture demand — branded search, retargeting, comparison shopping — always look strong on last-click because they tend to be the final touchpoint. But they rarely create demand. Cutting them barely reduces sales. Scaling them barely grows sales. Last-click reports them as the highest-ROAS channels anyway, which sends budget in the wrong direction.

2. It ignores impressions

Last-click only records clicks. Channels that drive revenue primarily through exposure — TikTok, YouTube, Meta video, connected TV, Snap — are under-credited by design. A shopper who sees a TikTok creative, remembers the brand, and buys days later through Google or Amazon is invisible to last-click measurement. The TikTok spend that initiated the journey gets nothing.

3. It is blind to marketplace sales

Amazon is the single largest blind spot in most DTC measurement stacks. When a shopper sees a Meta ad and buys on Amazon, that sale is attributed to Amazon — or more precisely, to whatever branded search or Amazon ad closed it. The Meta spend that drove the interest is not credited with a cent. For brands selling on both DTC and Amazon, this is often the largest single source of misallocated budget.

What the Data Shows

Research from Fospha, which analyses over $4B in annual paid media spend across hundreds of retail brands, shows that brands relying on last-click attribution systematically undervalue TikTok by 30 to 50 percent compared to full-funnel measurement.

“If brands assess TikTok solely through last-click, they risk underestimating the broader sales uplift happening across platforms like Amazon,” Fospha VP of Growth Jamie Bolton told ClickZ at SXSW 2025.

The practical consequence is predictable. Brands making decisions on last-click data cut TikTok, Snap, and YouTube investment that is — on a fully-loaded basis — actually profitable. They reinvest the budget into branded search and retargeting, which look great on last-click but don’t generate new demand. The short-term ROAS number goes up. The long-term growth rate goes down.

What Should Replace It

The current best-practice framework uses three complementary layers rather than a single attribution model:

• Impression-led MMM as the foundation. Media mix modeling measures each channel’s contribution to revenue using aggregated, impression-level data instead of click-level tracking. It works for every channel including those without reliable click data. Fospha built its platform around a daily-refresh MMM specifically to solve this problem. Measured runs a similar approach focused heavily on incrementality calibration. Northbeam and Triple Whale are more attribution-centric but have added MMM layers.
• Attribution as a tactical complement. Modeled multi-touch attribution is still useful for ad-level and creative-level optimization within a channel. Northbeam is a common choice for performance teams that want this granularity. The important shift is that attribution informs the tactical layer, not the strategic allocation layer.
• Incrementality testing for validation. Periodic geo-experiments or holdout tests confirm that the channels the MMM is recommending are actually causing lift. Measured is the most commonly cited specialist here. This layer catches the cases where the model and reality disagree.

None of these tools on their own solves what last-click was trying to do. Together they do.

2026 Context: Why Last-Click Is Even Less Reliable Than in 2022

Signal loss from iOS14+, the ATT framework, and progressive cookie deprecation has degraded the click-data last-click depends on. What was a flawed measurement model in 2022 is now a flawed measurement model running on partial data. Mobile web conversions often lose upstream click paths entirely. Server-side tracking closes some of the gap but cannot recover what was never captured.

The brands that have moved to impression-based measurement are not doing so because it is a more sophisticated approach. They are doing so because the older approach no longer works on the infrastructure that remains.

The Practical Takeaway

For any ecommerce brand still making primary budget decisions on last-click attribution in 2026, the question is not whether to change — it is when. The cost of staying on last-click compounds every quarter that TikTok, Snap, YouTube, and Amazon-influenced revenue continues to be under-credited. Moving to an MMM-led framework tends to shift budget toward upper-funnel channels and marketplace-influencing spend, which in most case studies produces meaningful revenue growth within a single quarter.