Not the GP, but I figure their point is as follows:
If I'm running an e-commerce website, I don't mind pay-per-search since those searches may turn into sales, so the cost is justified. My income scales with search count, and the Algolia price is part of user acquisition costs.
If I'm running a SaaS business, the search is a feature for customers who have already paid, so I don't see any further returns from the search being used. The more a client uses search, the less I'm profiting from having them as a client. They could potentially even cost me money to service them!
Interesting. This reminds me of Canada Post's address complete API. You can integrate it with your website or app to ensure the addresses the users enter are valid within Canada and entered correctly. The pricing is around 5–10¢ per search [0]. At that price point, it only makes sense to use it for e-commerce if you are going to deliver physical goods to the customer and want to minimize the number of returned "address not found" packages. But if the pricing were lower, e.g. close to Google Map's 0.3¢ per search [1], I imagine the list of potential uses would have increased substantially.
Many local governments publish shapefiles (geojson) that have address information. You could parse out addresses and have a fairly simple free option. Since this data is based on property taxes, these data sets aren't going to be missing parcels (or the tax collector would be missing potential revenue).
That still leaves out a lot: unincorporated lands, First Nation reserves, small settlement where all the mail goes to a post office instead of individual buildings, etc.
Also, one of the main benefits from Canada Post API is that it gives you the canonical address Canada Post uses to deliver packages. I don't think this always overlaps with addresses municipalities use to assess taxes. For example, I can imagine for an apartment building belonging to a real estate management company, there would be a single entry in the municipality database (because there is only one owner who needs to pay taxes), but one entry per unit in Canada Post database.
Yeah, my initial suggestion likely misses some corner cases. One case I am fairly confident it covers is (not sure if correct term) mutual interest land ownernship (aka, condos). Each layer has the same shape, but they're stacked one top another; each layer is a billable portion of the interest (e.g., a single condo on the shared land).
A poor man's approach could be to query the publicly available data sets, and use that data when there's match. Then, only pay for queries that fall into the edge cases.
Does an e-commerce site search really need anything more than fuzzy matching on item names? It pay not be perfect, but if it provides relevant results 90% of the time, you're paying a big premium for SaaS search that may only net you a small tail end of additional sales.
In particular for ecommerce sites it pays to invest in search. A large percentage of transactions is initiated by search queries. You are correct that fuzzy matching will likely get you decent results. However, finding the right product 90% of the time or 98% of the time makes a huge difference if you have a high volume of transactions. Synonyms and a better spelling model can massively improve your results and the resulting conversion.
But search relevancy is only one part of the equation. Think about a physical store and how the milk is usually in the back and a few high margin items are close to the counter or strategically placed. The same can be true for an ecommerce store. If the search engine has the ability to take business metrics like revenue and margins, or customer data like loyalty programs or brand affinity into account, you can much better optimise for your desired business outcome.
We just recently switched one of the largest ecommerce retailers in Australia over from Algolia to Sajari by doing the above and increasing their conversion rate by 10%.
