Case Studies

How to understand exactly who you are selling to.

Each case documented here examines a different failure point in how e-commerce businesses define their audience — and what changed when they looked more carefully at actual behaviour rather than assumed demographics.

Three documented cases

These are condensed accounts of real analytical work conducted through Lorsked's webinar methodology. Each case was selected because it illustrates a distinct dimension of customer engagement — from segmentation logic to on-site behaviour mapping to post-purchase communication gaps.

Audience segmentation

Mapping buyer segments for a mid-size apparel retailer

The retailer had built four customer personas over three years — none of which matched the cohorts that actually drove repeat purchases. A structured webinar series with their marketing team used session recordings, return-rate data, and live polling to identify that their most loyal buyers shared a motivation the personas had never captured: they were purchasing for specific social occasions, not general wardrobe building.

Rebuilding segments around occasion-based intent rather than demographic proxies shifted how product descriptions were written and which items were surfaced in email sequences.

6 weeks analysis duration

4 → 9 segments identified

3 markets covered

Cart abandonment intent analysis workshop

Behavioural intent

Reducing cart abandonment through audience intent analysis

A consumer electronics platform was treating all cart abandonment as a price sensitivity problem. Participant analysis from three live Lorsked sessions revealed a more fragmented picture: a significant portion of abandoning users were not price-sensitive at all — they were uncertain about compatibility and lacked the information architecture to resolve that uncertainty before leaving.

The intervention focused on product page restructuring and a contextual FAQ layer, not discounting. Abandonment at the specification review stage declined measurably within the first quarter after implementation.

3 sessions live analysis

Q1 result first measurable shift

The analytical sequence behind each case

Establishing the actual question

Most engagements begin with a stated problem that shifts once the data is examined. The first step is separating the symptom from the structural issue — which is rarely the one the client arrived with.

Behavioural data before survey data

What users do on a site is more diagnostic than what they say in a survey. Session data, scroll depth, exit points, and return patterns are examined before any qualitative layer is introduced.

Live webinar as a structured analytical tool

Lorsked's sessions are not presentations — they are working sessions where participants examine real data together. The interactive format accelerates the identification of patterns that would take weeks to surface in asynchronous analysis.

Intervention scoped to the finding

Recommendations are constrained to what the data actually supports. No case study here includes a recommendation that outran its evidence base.

Since 2016 Lorsked has structured its methodology around one consistent premise: understanding who you are selling to is an empirical problem, not a creative one.

Transnational Cases documented across markets in Central Europe, the Nordics, and Southeast Asia — each with distinct consumer behaviour patterns requiring separate analytical treatment.

Patterns that recur across cases

Personas age faster than behaviour does

In every case examined, the formal persona documentation was at least 18 months behind observable user behaviour. Personas are snapshots — they require active maintenance to remain useful.

Price sensitivity is frequently misattributed

Abandonment and churn are regularly attributed to price when the underlying driver is information deficit, trust gap, or friction in the decision path. Discounting in these cases produces short-term lift and long-term margin erosion.

Segment count is not segment quality

More segments do not produce better targeting. The apparel case moved from four segments to nine — but the value came from the criteria used to define them, not the number itself.