Organic Search Growth for a Fragrance Retailer

The catalog was strong. The storefront could barely be read by a search engine.

Audit Findings

Almost nothing on this site was wrong in an unusual way, which is the normal case. The problems were the ordinary ones that accumulate when a storefront is configured by people whose responsibility is retail rather than crawl budget.

Faceted navigation was generating URLs combinatorially. Every combination of brand, size, concentration, and price band existed as a distinct, indexable, crawlable address, and the crawler was spending its visits on the several thousand near-identical listing pages that resulted rather than on the product pages that generated revenue. Variants compounded the effect. A 50ml and a 100ml presentation of the same fragrance occupied separate URLs with near-identical copy and no canonical relationship between them, so they competed with one another and neither prevailed.

Collection pages rendered their product grids client-side, which meant the category pages, the pages carrying the commercial intent, delivered an empty container to a crawler and populated it afterwards. That is sometimes rendered and indexed. Depending on it is a strategy that works until the day it does not.

Server logs settled the remainder of the argument. They established where crawler attention was actually directed, which was not where it had been assumed to be, and provided a baseline that was a measurement rather than a tool’s estimate of one.

That baseline was agreed and recorded in the first week, which determines whether a search program can be assessed at all. Most disagreement about search performance is disagreement about the position before any work began, and that is not resolvable retrospectively.

Crawl Before Content

Crawlability, speed, and product schema were corrected before a single word of content was written. The remainder was then built out against real query data.

The order is the entire method, and it is the part clients most frequently propose to skip because content is the visible half. The reasoning is arithmetic. A crawler allocates a site of this size a finite budget of requests. Where most of that budget is consumed by parameterized facet URLs, new content is discovered late, recrawled rarely, and ranked against competitors whose pages are visited daily. Forty well-written articles published into that condition can sit unindexed for a quarter.

Facets were therefore closed to crawling except for the small number of combinations with genuine search demand behind them, which were promoted to real, linkable, statically rendered landing pages instead. Variants were consolidated onto one canonical product URL with the sizes presented as options. Pagination was made crawlable and coherent. Sitemaps were split by content type so that indexation could be observed per type rather than as a single figure that never moves.

Performance Measured in Real Conditions

Storefront performance work fails when it is conducted against a laboratory score. The measurement that matters is the experience on a mid-range Android handset over a mobile network, because that describes most of this traffic, and a synthetic run on a desktop connection reports that everything is satisfactory.

Largest contentful paint on the product page was dominated by a hero image delivered at desktop dimensions to every device, behind a carousel script that had to initialize before the first frame could paint. That is a five-line fix with a disproportionate effect. Beneath it sat a heavier problem: a stack of third-party tags comprising two analytics vendors, a reviews widget, a chat widget, and a personalization script trialed and never removed, all loading in the head, all blocking, and each defended by someone. Removing those for which no owner could be named was the single largest improvement of the engagement.

Structured Data and What It Delivers

Structured data is oversold and remains worth implementing properly. Marked-up price, availability, and currency are what allow a listing to present as more than a blue link, and in a category where the same bottle is sold by twenty retailers, being the listing that shows a price and an in-stock state is a real advantage over being the one that does not.

What was true was marked up, and only what was true. Ratings were emitted only for products that carried reviews, because inventing aggregate ratings in schema is both a policy violation and trivially detectable. Breadcrumb markup followed the corrected information architecture, which had the secondary effect of making the taxonomy legible internally as well as to the crawler.

Content Written Against Queries

The content plan was derived from Search Console query data, the site’s internal search logs, and the questions the retailer’s own staff answered by email: three sources of language people genuinely use, in place of a tool’s list of terms sorted by volume.

What that surfaced was comparison and substitution intent. Buyers do not search for the category in general terms. They search for a discontinued scent and a close equivalent, for how one house’s formulation differs from another’s, or for what performs in Florida heat without dissipating within the hour. Those are answerable questions carrying commercial intent, and almost no one was answering them, because answering them requires a person with category knowledge to sit down and write.

Every piece was linked into the collection pages it discussed, in both directions. Editorial content that does not link to the commercial pages it supports is a blog. Linked properly, it is an argument for why those pages deserve to rank.

Results

Traffic continued to climb for months after the engagement concluded.

That is the accurate statement of the result and also the strongest available one, because it is the property that distinguishes the channel. A paid campaign stops on the day the card stops. Technical work of this kind compounds: pages that are crawlable, fast, and internally linked continue to accrue value, and content published in the second month was still earning in the twentieth without further expenditure.