Do people use your product to search for products or information?

It is hard for me to fathom an online product that doesn't. I help guide my clients to decide where to invest their hard-earned dollars into their infrastructure to make sure they get the biggest return on investment.

For this example, we are sticking with e-commerce, as SaaS and subscription models are a bit tougher to directly calculate. The short answer for subscription models is: if it decreases churn or increases CAC, then that is a good thing.

For something like e-commerce, you can quantify things much more directly, as you can clearly track whether or not a user puts the search result in their shopping cart and buys it.

Here are a few ways you could quantify the efficacy of your investment into search, as it applies to e-commerce.

Let’s start with the obvious but not optimal gross sales revenue per search. Gross is easier to measure than net, but not necessarily better for the business long run. Bigger net = more profit margins for the business. When trying to move products, your search ranking algorithm should take that into account.

If the user hasn’t explicitly stated they are looking for the product with 1% margins, then it is silly to show those products over the ones with 50% margins.

That gets us to net revenue per search, but “per search” made doesn’t really equate back to dollars and cents either. It doesn’t matter how many searches are made, but more how much it costs us to service those searches. Each search may be a fraction of a fraction, but after a couple billion searches, it adds up quickly.

So that brings us to net revenue per dollar spent on search. This includes the cost of the underlying server infrastructure to serve up the searches, ongoing developer hours to maintain the service, and possibly even the initial developer hours to build it, amortized over the lifetime of the search service.

Let's say you find out your primary search tool leads to $10,000 a day in gross sales, which translates into $4,000 net revenue. This looks great, sure, but it is only half of the equation.

What if you're paying $5,000 a day for infrastructure (that sounds insane, but you would be amazed at the costs you incur at scale with a poorly designed system)?

That is why, when I design these massive search engines, I really take the time to focus on keeping those searches fast and cost-effective. In the above example, a good target to shoot for would be closer to $100/mo, effectively giving you a 100x ROI on your investment into search. Keep in mind that search is only a small part of your overhead—shipping, logistics, and all that other stuff still applies.

The same math can be applied to the AI tools you are adding to your website, like a chatbot. If the net revenue is being chipped away by the cost of the LLM models, is that feature really worth it?

If you are not measuring these things, there will be a large gap in the visibility of these key investments you are making in the business.

If you need help figuring out how to measure these things and/or design massively scalable, lightning-fast, cost-effective search engines, that is what I do for a living, so please feel free to reach out anytime.