Cart abandonment has averaged 70.22% across twenty years of studies — the model stops trying to fix it

Every store that has ever looked at its funnel has had the same conversation about the same number. Seven in ten carts are abandoned, somebody proposes a checkout redesign, and the number comes back at seven in ten. The published record is unusually clear on this: fifty separate studies spanning 2006 to 2025 average 70.22%, which is a span covering the arrival of the smartphone, one-page checkout, guest checkout, digital wallets and buy-now-pay-later. A metric that does not move across two decades of structural change to the thing it measures is describing how people shop, not how badly a particular checkout is built.

Seventy per cent of carts have been abandoned for twenty years, through every checkout redesign, every digital wallet and every trust badge anyone has ever added to a payment page.

1. 01

   A number that has not moved in twenty years is not a leak, it is a behaviour

   The cart is being used as a shortlist, a price check, a shipping calculator and a way of saving something for later. Most of those sessions were never a purchase intent that got interrupted. Budgeting to recover all of them is budgeting against a category error, and it usually crowds out the work that does move.

2. 02

   The device split is the inversion nobody budgets for

   Desktop converts 74% higher than mobile, and mobile carries 69.9% of the traffic. In retail specifically that is 3.7% against 2.0%. The majority of sessions arrive on the surface that converts worst, which makes mobile checkout the one part of the funnel where a redesign is arguing with the data rather than against it.

3. 03

   Returning visitors convert at 2.9% against 1.7% for new

   That is a 71% difference on the same site with the same checkout, which is roughly the size of the whole device gap and considerably easier to influence. It is also the argument for the lifecycle flows, stated as a conversion-rate fact rather than as an email-channel preference.

4. 04

   The best-earning automation in the published set is a stock notification, not a discount

   Back-in-stock earns $9.14 per email at a 6.72% conversion rate, welcome $6.16 at 2.11%, and abandoned cart $3.59 at 1.72%. Recovery is real and it is third on that list. The flow that tells somebody the thing they already wanted is available again outperforms the one that chases somebody who left, by roughly two and a half times per email.

5. 05

   AI-referred traffic converts at 1.3% and is growing 55% year on year

   That sits below paid search at 2.8% and above organic social at 0.7%. It is a channel worth instrumenting now and not yet worth reallocating budget into, and the model separates those two decisions deliberately because most plans conflate them.

**70.22%** average across **50 studies (2006–2025)** — unmoved by two decades of checkout redesigns, wallets and trust badges.

The model instruments before it optimises. Phase one splits the funnel by device and by visitor type, and separates cart abandonment from checkout abandonment, because the two are usually reported as one figure and have completely different causes. Phase two builds the capture and recovery layer the published data rewards: back-in-stock on every sold-out variant first, then welcome, then cart and checkout recovery. Phase three works on mobile checkout specifically, on the basis that it carries most of the traffic and converts worst. Media allocation comes last and is sized on the returning-visitor conversion rate rather than on the blended one, because the blended figure hides which half of the audience the spend is actually buying.

1. 1

   Weeks 1–2

   Instrument

   Funnel split by device and visitor type, cart abandonment separated from checkout abandonment

   Owner: OmniFlow

2. 2

   Weeks 2–4

   Capture layer

   Back-in-stock on sold-out variants, email capture, consent and suppression rules

   Owner: OmniFlow

3. 3

   Weeks 4–6

   Recovery flows

   Welcome, cart and checkout recovery built with exclusion sets and per-flow reporting

   Owner: OmniFlow

4. 4

   Weeks 6–10

   Mobile checkout

   Mobile-specific checkout work, measured against the desktop rate rather than against last month

   Owner: OmniFlow + client dev

5. 5

   Week 12

   Review

   Conversion by device and visitor type, revenue per recipient by flow, blended figure reported last

   Owner: OmniFlow

The model reports conversion by device and by visitor type rather than as a single site figure, and revenue per recipient by flow rather than as one email number. At the published benchmarks a store with 69.9% of its sessions on mobile is running most of its traffic through the surface that converts at 2.0% against 3.7%, and that split is the largest single structural fact in the funnel. Abandonment itself is reported as context, not as a target, because no published evidence suggests it is movable. Every benchmark input is replaced with the store's own figure from week two.

Modelled. Inputs: Baymard Institute, cart abandonment rate (50 studies, 2006–2025); Contentsquare, 2026 Digital Experience Benchmark (99 billion sessions, 6,000+ sites); Omnisend, 2026 Ecommerce Marketing Report (20B+ campaign emails, 27,000+ brands); Triple Whale, Ecommerce Benchmarks 2026 (53,000+ brands, August 2025 – July 2026). Modelled outputs are not a forecast or a guarantee of results. Cited: Baymard Institute *Cart Abandonment Rate Statistics* (2026).

Every figure below is published research, not a client result. They are the inputs to the arithmetic above, listed so it can be checked rather than taken on trust. The bracketed number points to the full citation at the end of this page.

These are the surfaces this engagement is run and measured from, shown with representative figures built around the benchmarks cited on this page. Every account we run reports into views like these, and you keep ownership of all of them.

Each figure on this page, the system it is read from, and the definition and window it is measured over.

Not applicable — this is a modelled engagement, not a delivered one. The model's weakest input is the abandonment figure itself: 70.22% is a mean across fifty studies run by different people over twenty years, with different definitions of what counts as a cart and what counts as an abandonment. It is strong evidence that the rate is stable and weak evidence about any individual store's rate. The instrumentation phase exists to establish the store's own definition and its own number before any decision is made on either.

A modelled capability, not a client account. Figures illustrate what the model produces and are labelled as modelled wherever they appear. See all our work

Full citations for everything cited on this page, with the sample and period each study covers, so you can go and read the original.