E-Commerce Competitive Marketing Analysis: What We Found and What We Recommended

The site ranked on page one for 30 keywords. Only 15 of those were standard organic listings. Thirteen appeared inside an informational search panel, and two inside an AI overview. Every one of the thirteen panel placements came from the same single blog post.

Those thirteen keywords accounted for roughly 12,300 of the approximately 15,500 monthly searches covered by the page-one set.

• 15 — standard organic listings a searcher can click through to the site
• 13 — placements inside an informational panel, all from one blog post
• 2 — placements inside an AI overview

A placement inside an informational panel answers the searcher’s question on Google’s own page. It builds familiarity. It rarely produces a visit, and almost never produces an order. Counting it as search performance makes a program look considerably healthier than it is — and it concentrated most of the site’s apparent visibility on one page that could lose its placement at any time.

Treat the informational rankings as brand reach rather than revenue, and build a separate layer of commercial rankings on product and category pages — the exact route Competitor A had already proven works in this category. Write descriptions that establish what a product is, in the words people search for, rather than asserting that it is good.

The nearest available wins were already visible in the data: 31 keywords sat in positions 11 to 30, representing roughly 8,700 monthly searches. Those are close enough to move without a large content program.

Fifteen live ads, most with fewer than 100 impressions. They had been created across a three-and-a-half-month window and all left running. Every ad was running on every available placement. The copy was near-identical across the set — only the images changed. Daily budget was roughly $20.

All of this was read from the platform’s public ad library, without access to the account. So was the vendor’s team structure: the page transparency panel listed the countries of the people managing it, which matched the outsourcing arrangement the owner had suspected but never confirmed.

With this many active ads competing for a relatively small daily budget, the campaign had too little spend relative to the number of creative variations to generate useful comparative data quickly. Too many simultaneous tests leave too little signal on any one of them, which makes it difficult to determine what is actually working and difficult to improve the account intelligently. The ads keep running because there is never enough evidence to justify switching them off. An account in this state looks busy and produces very little usable information — which is a different problem from an account that is losing money for an understandable reason.

Running identical copy across fifteen ads also means the test that was apparently being run — images — was the only variable, while the far more consequential variable was held constant.

Reduce the ad count to what the budget can genuinely support, and cut the placements that do not suit the product rather than defaulting to all of them. Confirm that the tracking pixel and remarketing audiences are actually configured — a question worth asking any vendor directly, because the answer is diagnostic on its own.

Then use the asset the business already had: build a lookalike audience from the existing customer email list instead of assembling an interest-based audience from scratch. Budget discipline of this kind is covered further in our piece on setting an e-commerce marketing budget.

Several internal links — a bundles item in the menu, a button on the homepage, a link in the footer — all landed on a duplicated variant of the About page instead of their intended destinations. This is a known consequence of how the theme handles duplicated page templates, and it had accumulated quietly over time. Social links pointing out from the site were broken as well.

A free-shipping message was displaying on an order well below the threshold that actually qualifies for it. The cart’s upsell prompt presented a dollar amount alongside a close button, with no framing of what the shopper gained by accepting or gave up by declining.

Paid traffic amplifies whatever a site already does, including the parts that do not work. Every one of these issues was already costing the business unpaid visitors; increasing spend would simply have meant paying for the same failures at a higher rate.

The shipping mismatch deserves particular attention. It surfaces at the precise moment a shopper is deciding whether to trust an unfamiliar brand with a credit card. And the upsell, though clearly accepted by a good share of customers, had never been tested against a version without it — so nobody knew whether the added cart value was worth the added step.

Fix the storefront first, and treat it as a prerequisite rather than a parallel workstream. The owner had built the site himself and was capable of doing most of the repairs; where that is not the case, it is work we can take on through our website design and optimization service. Either way, the sequence is the point.

On the upsell specifically: the platform supports split testing. Test it rather than defending it. A prompt that raises average order value while suppressing completed checkouts can look like a win in one report and a loss in the one that matters.

Three variants of the business name were live across the website, the Google Business Profile and the Facebook page — the residue of a partial rebrand that was never carried through to every property. Searching the brand name alone did not reliably surface the business; it took the name plus the city.

The business profile carried a single review, dated photography and no product listings. Separately, the products were appearing on a third-party marketplace the owner had signed up for and largely forgotten, at prices meaningfully above his own.

Around 80% of this business’s revenue came from in-person events. Those customers go home and search the name. The business profile is the first thing they find and frequently the only thing they read before deciding whether to order again. A name split three ways divides that search; a thin profile ends it.

The marketplace listings mattered for a different reason. A customer who compares the brand’s own price against a reseller’s higher price does not conclude the reseller is expensive. They conclude the brand’s pricing is arbitrary.

Standardize on one name across every property. Add the product catalogue to the business profile with current photography, replace the dated images, ask for reviews consistently at events, and reply to every review that arrives — those replies are read by prospective customers far more often than by the reviewer. Audit the third-party listings and decide deliberately which ones to keep.

One recommendation here was to not buy something: this profile needed a proper setup, not an ongoing optimization retainer. At this budget, recurring money belongs in media. An analysis that only ever concludes “buy more services” is not an analysis.

Repeat purchase was running somewhere between 5% and 10% by the owner’s own estimates. The subscription option existed but the incentive was shallow — a modest discount, nothing else. It had attracted only a small subscription base, and that base had already declined by the time of the review. Reported return on ad spend was between 1x and 1.5x, and roughly 0.75x in the most recent measured month.

When almost all revenue arrives on a first order, first-order profitability is the whole business. A return of 0.75x is not a slow payback period. It is a loss, and the second purchase that would eventually cover it does not arrive for nine customers in ten.

Buying the first customer at a loss is a legitimate and often correct strategy — but only when the repeat rate is known and high enough to justify it. Here it was neither measured precisely nor high. That single fact ruled out an entire category of aggressive acquisition tactics that would otherwise have been on the table.

Make the subscription worth joining before spending more to acquire the people who might join it. A sample of something new is frequently more persuasive than a deeper discount, and it spends inventory rather than margin — particularly useful for a producer with stock approaching its date.

Then measure the repeat rate deliberately rather than estimating it, because it is the number that decides whether acquisition can ever run at a planned loss. Until it is known, every acquisition plan has to pay for itself on the first order.