The Invisible Phase of the Purchasing Decision

Most purchasing decisions have already begun long before they show up in anyone's analytics. The previous articles in this series made that case in structural terms. What they did not do, however, is spell out what that invisible phase actually looks like when you follow it through real scenarios, step by step, and hold it against the data a retailer would typically see on their dashboard.

That is what this article sets out to do. Three scenarios, each increasingly common, each structurally quite different, and each producing the same curious result: a customer journey that looks perfectly legible in the data whilst concealing the most consequential part of the decision.

Scenario 1: The AI-assisted pre-selection

A woman is considering a high-end espresso machine. She does not begin with a Google search. Instead, she opens ChatGPT and types something along the lines of: Which espresso machine would you recommend for home use if I care about temperature stability and quiet grinding?

Within seconds, she receives a tidy comparison of three or four models, contextualised and ranked. No advertisements, no sponsored placements, no discernible commercial agenda. It reads, for all intents and purposes, like disinterested advice, which is precisely what makes it so influential.

She makes a mental note of two models, opens a browser, types one of the names into Google and lands on a retailer's product page. A brief comparison, a glance at the reviews, and the order is placed.

The retailer's analytics tell a reassuringly familiar story: a user arrived via brand search, visited two product pages and converted. Short journey, clear intent, healthy conversion rate. Nothing to investigate.

What the analytics do not tell is that the decision was effectively settled before the first click occurred. The shortlist was assembled by a system the retailer has no visibility into and no means of influencing. Every product that failed to appear in that AI-generated response was eliminated from contention before the journey even began, irrespective of how polished its product page might have been or how sharp its pricing.

Scenario 2: The social imprint

A man follows several interior design accounts on Instagram. In one of them, a particular kitchen appliance appears in the background of a story, not as an advertised product but simply as part of the scenery. A few weeks later, a TikTok video surfaces in his feed: someone cooking with the same appliance, mentioning in passing that it has served them well for two years.

At no point has this man actively searched for a kitchen appliance. There is no conscious purchase intent. Yet something has quietly taken root: a familiarity with a specific brand and a specific product, one he would struggle to articulate if asked about it directly.

Months later, when his existing appliance finally expires, he does not search for "best kitchen appliance 2026." He searches for the brand name. He visits the shop, confirms the price, and orders.

The retailer's dashboard paints a flattering picture: direct traffic or brand search, rapid conversion, low bounce rate, modest acquisition cost. By every available metric, this is a textbook loyal customer with strong brand affinity.

In truth, that affinity was not cultivated by the retailer's marketing department. It was built, gradually and incidentally, across platforms the retailer neither monitors nor controls. The retailer can see the outcome but has no reliable account of how it came about, which means there is no way to engineer it again deliberately.

Scenario 3: The recommendation behind closed doors

An entrepreneur needs a new e-commerce platform for her shop's relaunch. She posts the question in a private Slack community where roughly two hundred fellow D2C founders exchange notes.

Within two hours, seven replies arrive. Four of them recommend the same platform, each with personal experience and specific assessments attached. Another platform, one that would objectively be a strong contender, goes entirely unmentioned.

The entrepreneur researches the recommended platform, reads the documentation and books a consultation. In the provider's analytics, she shows up as organic traffic via branded search. The journey looks like a standard research-to-conversion sequence.

What the data does not capture is that the shortlist was decided in a private Slack channel, a space no tracking pixel will ever reach. The weight of four peer recommendations outstripped anything a product page, a case study or an advertising campaign could have achieved. And the platform nobody mentioned? It had no opportunity to intervene in a process it did not know was happening.

What connects these scenarios

In all three cases, the visible customer journey appears complete. The data is clean, the attribution plausible, the conversion rates stable. There is nothing in the numbers to suggest that anything is missing.

That is precisely the difficulty. The invisible phase does not announce itself through anomalies. It produces no gaps, no outliers, no red flags. The visible data remains internally consistent, which makes its incompleteness all the harder to detect.

The upstream influence differs in each case: algorithmic pre-selection in the first, social imprinting in the second, peer endorsement behind closed doors in the third. But the structural effect is the same: by the time the user appears in a system the retailer controls, the decision space has already narrowed substantially.

What this means for retailers

The three scenarios suggest that understanding how users behave within the shop, whilst still operationally necessary, no longer provides a sufficient account of why certain products enter consideration at all whilst others never get a look in.

The real selection happens at a point the retailer can neither govern nor fully reconstruct. This forces a shift in strategic attention towards a question that was long treated as settled: under what conditions does a product enter a user's decision space in the first place? Those conditions are not shaped inside the shop. They are shaped upstream, in a phase where visibility within AI systems, presence in social contexts and standing within closed networks determine whether a product is even perceived as an option.

The invisible phase of the purchasing decision is not, therefore, an extra step to bolt onto the existing model. It is a reminder that the model itself starts at a point where the most consequential choices have already been made.

The next article, The Loss of Attribution, examines the limits classical attribution models are reaching in contemporary e-commerce, and why more data will not resolve the underlying problem.

This article is part of the Decision Journey Analytics series, examining how e-commerce decisions evolve beyond the classical funnel.