Leaks at the Add-to-Cart Step: Clicks That Go Nowhere

The add-to-cart step fails mechanically more often than persuasively: the visitor decided, acted, and the interface lost them. The recurrent modes are an unselected variant blocking the click with an unclear message (or none), a button pushed below a wall of options on mobile, feedback so quiet the visitor cannot tell the add happened (and clicks again, or leaves), and page states - a zoom overlay, a sticky bar - swallowing the tap. Audit the step as machinery: every product type, on a phone, one-handed, including the error paths. Mechanical leaks are the funnel's best news: they repair completely and stay repaired.

Requiring a variant choice is legitimate; failing to communicate it is the leak. The patterns: the button appears to do nothing because the unselected size only announces itself in a line of small text far from the tap; the default variant is the sold-out one; the picker resets after an image swipe. The repair standard: the blocked click produces a visible, proximal prompt (the picker highlights, the message sits beside the button, the page scrolls the picker into view), and a sensible default is preselected wherever one honestly exists. Recordings at this step are unambiguous - watch for the repeated button tap, the funnel's clearest distress signal.

After the click, the visitor needs one clear confirmation within a moment: the cart count changes visibly, a drawer or message confirms the item and its variant, or the page moves to the cart - one pattern, consistently, across the store. Ambiguous feedback produces double-adds, hesitant returns to the product page to check, and a measurable share of quiet abandonments from visitors unsure what state they are in. Choose the pattern by your store's shape (the till bench weighs drawer against cart page) and never run two patterns on different templates; consistency is what makes feedback readable.

On phones the button competes with variant pickers, quantity, delivery snippets and sticky bars for the same small screen. The leak pattern: options push the button below two scrolls, and the visitor who has chosen everything cannot find the way to commit. Sticky add bars solve reach and create their own risks (screen theft, accidental taps over content) - the speed and mobile bench covers the trade-off. Whatever the pattern, the test is physical: a new visitor, one hand, a bus ride's worth of distraction, adds successfully without hunting. If the test needs two hands or full attention, the step leaks by design.

Some step drops are caused one inch away: a quantity selector defaulting oddly, a subscription toggle preselected that changes the price at the click (read as a trick, abandoned as a principle), a personalisation field required before adding. Each attaches a surprise or a tax to the commitment moment. The audit rule: at the click, nothing about the deal may change and nothing new may be demanded. Options that alter price or terms belong visibly before the button, in the same visual block, priced in the open.

Instrument the step honestly: adds per product view by device, plus a qualitative pass of recordings after the change. Mechanical repairs show fast, clean movement - one of the few funnel repairs that does - because the intent was always there. If the step rate refuses to move after the machinery is provably sound, the diagnosis was wrong: the leak is consideration (the doubt list, previous guide) or the traffic arriving at the product has changed its intent. The stages are neighbours; a clean mechanical audit is what earns you the right to look next door.

A nastier mechanical class deserves its own line: the add succeeds and the state then betrays it - the cart empties on the next page because of a cookie or session fault, the item appears in the drawer but not at checkout, the quantity doubles back at the cart. These read in aggregate data as consideration or cart leaks and are found only by doing the thing: full test purchases per template, per device class, after every theme or app change, including the unglamorous paths (back button, second tab, returning an hour later). Keep the test log with the change log; state bugs correlate with changes, and the correlation is the diagnosis when a step rate falls off a cliff on a release date.