HomeGadgets - Four Months In

Where the project stands, and where it's heading: towards Canadian retail data an agent can actually act on.

It's been about four months since I started building HomeGadgets, so it feels like a good moment to share where it stands, what I have learned, and (more importantly) where it's going.

The platform currently tracks over 20,000 unique SKUs from 130 top brands, including Samsung, LG, Nikon, Sony, Whirlpool, Apple, HP, Lenovo, Google, Dyson and more, across 120+ major Canadian retailers, and it keeps growing weekly. On existing product categories it already maps anywhere between 60% and 95% of Canadian retail volume, so it is getting to authoritative levels: a single site mapping the entire retail ecosystem for major consumer electronics.

The retailer catalog includes flagship names such as Best Buy, Walmart and Costco, and others such as Leon's, Newegg, Canada Computers, RONA, Visions, London Drugs, Memory Express, Tasco, Henry's, Canadian Appliance Source, Germain Larivière, etc. (Amazon is not in the catalog: we do not collect Amazon prices. Some products carry a plain affiliate search link to Amazon.ca so you can check it yourself.)

It's built in Canada, for Canadians, and focused on specific high-value durable goods that rank 3rd in one-off major purchases from consumers, after homes and vehicles. It currently focuses on 20 product categories: kitchen and laundry appliances, TVs, soundbars, monitors, laptops, cameras (mirrorless, DSLR, compact and action/360), lenses, vacuums, and headphones/earphones. Additional consumer-electronics categories will be added shortly, making it a one-stop shop for over $20B of yearly consumer spending by Canadians. (By Labour Day that number will include an additional $5B of consumer spending from a handful of adjacent consumer-electronics categories, so stay tuned!)

On other milestones, just a few days ago the site crossed $1M in referred gross merchandise value. Traffic is growing organically (zero advertising spend), but interestingly (and not surprisingly) close to half of traffic is already referred by AI chatbots (ChatGPT, Gemini, Claude, Perplexity, etc.).

Given these trends, I started to merge in another project of mine (ShopButler.ai) and started building tools, utilities and infrastructure that will welcome emerging agentic shopping experiences (an AI being instructed to find a specific item, under given conditions, and actually buy it on your behalf and arrange for shipping/delivery as per your instructions).

Doing this surfaced some major cracks in existing e-commerce and in the findability of best offers, some of which are guiding the next stage of deployments.

The number one pain when shopping online

Let's assume you already narrowed down what exactly you want to buy. Your next step is figuring out where can I get the best deal? It turns out today's tools do not necessarily deliver on that simple requirement: find the cheapest place where I can buy "this", now. That means it is truly cheapest, it is truly available for you (in inventory, shippable to you, or available for pickup near you), and, equally important, it is exactly what you want (not a mismatch).

A search query of what you want (e.g. using Google, Bing, etc.) yields two types of results:

1. Sponsored: in your face will be the retailer that paid the most to match your query, not necessarily the cheapest.
2. SEO-ranked: the list will gladly present the retailers who have won the search-engine-optimization arms race, yet again, not necessarily the cheapest.

The cheapest may be on page 1, or page 17. So clearly that does not necessarily work for you.

What about Google Shopping? Well, the prices you see there are not prices "now". They are actually entered by a retailer, and they can also be stale. Clicking on a retailer may very well result in seeing a price that is different from the one Google displayed.

But you are smart, and immediately ask your favourite AI chatbot (Gemini, ChatGPT, Claude, etc.). The problem with that is that AI chatbots ALSO rely on SEO. They have read the pages, and the source of "truth" is based on what they find first (the top page results), not what may be on page 2, or 10, which could in fact be the one offering the best price.

I think by now we have all acknowledged that AI chatbots are notoriously lazy. You may be successful probing and poking a few times until you are satisfied (or tired), but the problem is you will likely leave money on the table settling for what was presented to you, which, at this point I have proven on countless occasions, can very well be far from the best price or terms as you wanted them.

So, to get around that, I've been doing the unglamorous work of progressively mapping the Canadian retail universe for consumer electronics: what's available, who sells it, where, and at what price. Some limitations are definitely in place. Some retailers, still in 2026, want you to call them or send them an email to get a price. Some require you to register as a shopper (you are giving them your PII so they can render prices!). They are common among appliances and some office supplies. By design, those entities are not covered. Fortunately, based on my own metrics, they are less than 5% of retail volume, so I doubt they will in fact be price-competitive to begin with.

Understanding what exactly is for sale, everywhere

At first glance this looks like a straightforward data-collection problem. It isn't. The real challenge is data normalization. You'd expect OEM manufacturer part numbers (MPNs), GTINs and UPC codes to be consistent and/or displayed across retailer sites. In practice they're anything but. A product code like ABC/DE from an OEM might appear as ABCDE, ABC-DE, or with characters quietly altered, added or deleted. Product names rarely match the manufacturer's canonical format. Data that starts clean and structured at the OEM level becomes inconsistent and fragmented by the time it reaches every retailer, resulting in one giant puzzle of tens of thousands of irreconcilable table fields. So, data normalization is hard work.

Additionally, every retailer renders information (e.g. builds its website) using different tools, structures, mechanisms, etc., compounding the data-normalization challenges even further.

That said, price comparison isn't a new idea. There have been plenty of decent-ish tools over the years, so the challenges are well known. To overcome that, players have traditionally compared prices within a curated list of retailers who play along, by definition resulting in a limited sample of places to shop, and even though they can positively display "cheapest", it is only true within the much smaller sample of retailers.

Tools like Shopbot, PriceGrabber, etc. in Canada used to be quite good for their time, but they have not been updated in years. Others that still exist are great at spotting one-off "deals" (sort of Groupon), but not at mapping entire catalogs.

Finally, a few newer Canadian sites have appeared in the last year trying to solve similar problems, so it'll be interesting to watch how the space evolves.

Other challenges and acknowledgements

The site is not perfect, not by a long shot. Some categories (laptops especially) are inherently hard to compare side by side in retail: a single product line/family (e.g. a Lenovo P14s) can have over 4,000 different builds, with variations across processor, RAM, storage, integrated graphics, colour, keyboard layout, display type, ports and more, and no retailer carries every combination, just a subsegment. That said, at HomeGadgets you can filter by the specs and price you want and still save real money.

Other categories such as appliances, monitors, cameras and smartphones are far more standardized and much more likely to have a larger number of matches across retailers.

In terms of price "freshness", the front end currently renders prices that are refreshed up to twice a week, and that cadence is moving to daily.

But the bigger shift I'm spending some cycles on is real-time prices: surfacing the best price now, verified the moment you ask, instead of "as of the last refresh." I've been testing methodologies for exactly that, and I'm genuinely excited about it. More news soon. Whichever tool you use, results are always ranked by price, then alphabetically to break ties, for every SKU. No retailer or manufacturer can pay to influence the ranking.

Real-time pricing, and other tools shipping

That real-time direction is the point of the whole thing. The endgame isn't just a nicer price-comparison website. It's nearly universal Canadian retail data organized cleanly and freshly enough that software can act on it: an assistant that finds you the genuinely best price, and where to buy it, on demand. That's where shopping is heading, and reliable, normalized data is the piece almost nobody has.

A few things already live:

• AI search inside HomeGadgets (homegadgets.ca/ai). Ask real questions like "Where can I buy the cheapest 24 cu ft French-door fridge from Samsung or LG that ships to Vancouver?" Soon it'll fold in delivery details too: time windows, shipping and haul-away costs.
• An MCP server (mcp.homegadgets.ca/mcp) you can connect to Claude or ChatGPT to query the catalog directly. Today it reads the published front-end data; a real-time query capability is under testing: that's the one I'm most excited about.
• An enterprise API for real-time data integration, available on request.
• No ads. (I suppose that counts as a feature these days.)
• Price history, deep specs, consumer ratings (aggregated from third parties), and side-by-side spec/price comparisons across categories like TVs and more.
• Browser extensions and iOS/Android apps are planned for later this fall.

If you're shopping for a monitor, TV, appliance or other electronics, give it a try. And if something looks off or plainly broken, please tell me here, or via homegadgets.ca/report-bug.

Two final notes. The site uses some affiliate links, including Amazon, but they don't influence how prices are ranked. And I'm building this independently, in my spare time: no team, no company behind it. The goal from day one has been to understand retail data normalization at scale across verticals (I actually started with food and groceries), as we move toward agent-driven shopping, where systems need reliable data to act on people's behalf. In the meantime, it might just help you land a deal on your next gadget.

Thanks for checking it out.

Rodrigo Madriz