The challenge
Williams Food Equipment wanted the commercial side of the business to grow without letting go of the home cooks who built it, and wanted the whole thing to clear a ROAS of 5.5. Results were holding below that, and there was room to grow in 3 places.
- One set of campaigns served both audiences, so the higher-margin commercial range shared its budget with everything else
- Sales data wasn’t fully reaching Google Ads yet, which held back its smartest bidding options
- 1,162 products weren’t yet showing in Google Shopping
Organic search had its own opportunity. The site had grown over the years into thousands of similar category and brand pages competing for the same searches, and product pages weren’t yet giving Google the full product information it uses to show prices and ratings in results. Organic revenue had softened since early 2024.
On more than 40,000 products, every small inconsistency gets repeated a few thousand times, so the catalog needed one consistent way of working across every channel.
The channels also ran separately, so a win in one couldn’t yet reach the others. The client put it plainly:
“A series of agency relationships that simply couldn’t handle the complexity of our business – large, diverse product catalog, consumer and B2B.”
The solution
We started in May 2022 with Pay Per Click, Google Ads and Microsoft Ads together. Paid Social followed, then CRO and SEO. All 4 run off the same catalog, so the feed and structure work underneath shows up in every one of them.
Pay Per ClickRebuilding the catalog around who is buying
ServicesGoogle Ads · Microsoft Ads · Feed Management
DataFeedWatch was already installed when we took over, with plenty of room to do more. We reworked color mapping, shipping, GTINs and descriptions until the disapprovals fell from 1,162 to 23, which put roughly 1,100 products back in front of buyers who had not been able to see them. A set of Google Ads scripts now watches the account around the clock, among them a disapproval alert that flags a product the moment Google drops it.
Then we cut the campaigns apart by end user, so home cooks and professional chefs stopped bidding against each other out of the same budget. We rebuilt Shopping around buying intent, separating low, medium and high, and pulled the commercial products clear of the rest.
On top of that sits product bucketing: custom labels that sort the catalog by audience and by how each product actually performs, so budget follows the winners instead of spreading evenly across 40,000-plus SKUs.
Feed rules in DataFeedWatch, where the attribute work was done.
Fixing the tracking, then copying it across
Enhanced conversions weren’t set up yet, so we wrote the code in-house, briefed the client’s developers and shipped it on the thank you page through Google Tag Manager. That opened up the stronger bidding strategies for the account.
The same structure went onto Microsoft Ads, with Dynamic Search Ads added on top. Microsoft skews toward business users, which makes it quietly strong for the trade side of this catalog. Performance Max had launched a few weeks before we picked the account up, and we started testing segmentation on it straight away.
Paid SocialServicesMeta Ads · Creative · Feed Management
Meta was live, with room for a clearer structure. We audited what was running, paused the campaigns that weren’t pulling their weight, rebalanced the catalog campaign that was taking most of the budget, then rebuilt the structure from scratch on the back of research into the market, the audience and the brands that already sold hardest.
A layer for every job
The latest overhaul drew a hard line between acquisition and remarketing. We switched on Meta’s new-customer acquisition setting, so the acquisition campaigns now reach genuine prospects only and stop spending on people who already buy.
- Top: a broad, always-on Advantage+ campaign running the full consumer catalog, with no interest layering, so the algorithm has room to find buyers
- Middle: category campaigns for the ranges a broad campaign drowns out. Electrics carry a high average order value. Knives come in so many types, plus factory seconds (knives with small cosmetic flaws) and personalized engraving, that they need a campaign of their own
- Bottom: one retargeting pool covering product viewers, add-to-cart and checkout starters in that order, with catalog ads showing people exactly what they left behind
Retargeting also got something it never had before: short videos on why to buy from Williams, built on the store’s selling points. They are among the best converters in the account.
Meta ads in rotation: the Factory Seconds promotion, the electrics category ad, and the Le Creuset Bleu Riviera unboxing.
Brands get moments
Around that we run brand sets built on events: Le Creuset for the Bleu Riviera launch and the limited-edition Petal collection at Mother’s Day, All-Clad for bakeware promotions, Wusthof around the Classic Ikon chef’s knife. Each set excludes the others, so we never pay twice to reach the same person.
A brand video runs alongside them, built on the things that genuinely earn trust here: Canadian, family-owned since 1968, free delivery over $99, stocked by home cooks and restaurants alike, gift registry on the shelf.
Creative follows the brand’s own editorial plan. We mix the formats deliberately, Reels-first video against dynamic catalog carousels and collection ads, so Meta has enough variety to put the right creative in front of the right person.
Filling the shop floor
Local campaigns carry their own budget and point at the store itself. Williams runs a calendar of recurring events, big sales such as the Warehouse Sale and the Open House Sale, and chef demonstrations in between. The campaigns run on engagement, and the goal is people through the door: we promote the store’s own organic posts, and where the event exists on Facebook, the ad links straight to it so people can sign up.
The in-store team at Williams films a steady supply of local footage, sales and demos included. We plan the angles from our own analysis of each audience and keep the campaigns stocked with it. Video consistently beats static images here.
SEORestructuring 7,000 categories
ServicesTechnical SEO · On-Page Optimization · Content & Copywriting · Off-Page & Link Building
We opened on the technical layer. Product pages were missing JSON-LD thanks to a parent-child data flow issue and a theme bug, which kept the whole catalog out of Google’s SERP features. We stripped the old Microdata, standardized the JSON-LD, validated it, and had the pages re-indexed inside 24 hours. The sandbox images went the same week, with the client’s development team.
Alongside that we rewrote meta titles and descriptions and built content around the categories and brands that convert hardest. Screaming Frog and Semrush crawl the site on a cycle so nothing creeps back in.
Turning a recipe blog into a buying guide
The blog we inherited leaned toward recipes. Many posts were short recipes from a database, also published elsewhere on the web and not yet linked to the products Williams sells. We retired the ones that had run their course and rewrote the ones whose topic still held up or still drew searches, and every month we revise old posts and publish new ones.
New posts go after informational searches. Product-led terms belong to the category pages, so the blog takes the questions instead: how to choose, how to care for, which of 2 brands suits whom. Comparisons stay objective, every recommendation gets justified, and every post links into the catalog and into the rest of the blog.
The keyword research runs every quarter on Canadian data, part AI and part human, so new topics surface and fading ones drop down the list. Drafts come out of a workflow we built in Make: topic, main keyword and secondary keywords go into a Google Sheet, OpenAI’s latest models write the article, and it comes back as code that pastes straight into BigCommerce. Every draft then gets our edit, plus custom infographics generated in Williams’ own style and palette in place of stock photos.
Turning 7,000 categories into 3 levels
The bigger project is collapsing years of over-categorization into a structure no deeper than 3 levels, clearing out duplicate category and brand pages and focusing crawl budget on the pages people actually visit.
Doing that by hand was never realistic, so we built tooling instead. We export the products, reorganize them with AI so none end up orphaned, and bring in keyword data to decide what to keep, merge or remove. Every proposed structure gets checked against both search demand and whether a human could navigate it.
The workflow runs on custom Claude skills we built in-house: one writes on-brand content with internal linking, one reads the live category tree and proposes its replacement, one runs SEO audits, and one takes brand pages end to end, writing the copy and setting the redirects from old slugs to new so nothing 404s. BigCommerce offers no bulk editing, so we wrote our own script: it pulls every category into a Google Sheet, we edit titles, descriptions and meta data there, and it pushes the lot back.
CRORebuilding the pages that do the converting
ServicesConversion Audit · UX & Design · Testing & Implementation
Most of the work sits on the product page, where we redesigned the above-the-fold with the client’s development team. We mocked up every change in Figma, grounded it in established CRO research, and walked through the reasoning before anyone built anything. Conversion rate improved.
The home page has had a full redesign too, built together with a dev team that ships quickly. The commercial category pages get the same treatment, pulling together product information from across the site. Those are the pages the paid campaigns land on, so the work feeds straight back into the other 3 channels.
The changes come from Clarity recordings, heatmaps, click and scroll maps, on-site surveys and heuristic UX analysis. We watch the data itself as closely as the pages, checking GA4 for tracking gaps and filtering bot traffic, so the numbers everything else gets judged on hold up.
The results
Reporting changed first. A monthly summary became a bi-weekly working session with live Looker Studio dashboards and a quarterly strategy layer, which let the client plan around the account instead of reacting to it.
On paid, revenue climbed 390% year on year and ROAS reached 5.85x, a 244% improvement, against a target of 5.5. Spend rose 42% over the same stretch, so the growth arrived with better efficiency rather than instead of it. On organic, between October 2024 and February 2025, revenue ran 111% ahead period on period and 70% ahead year on year.
- Transactions up 398%
- Impressions up 211% and clicks up 67% across the top-converting categories and brands
- Feed disapprovals down from 1,162 to 23
The same people have held the account since 2022, so the work compounds year on year instead of restarting each time someone new picks it up.
”They always bring potential improvements to our attention, but they do so with an analytical, evidence-based approach.”
+390%
Revenue · year on year, from paid
+70%
Organic revenue · year on year, Oct 2024 to Feb 2025