Tracking Computer Hardware Deals in Avogadro One

How to never miss a good deal

In today's post I wanted to show how Avogadro One can be used for rather mundane daily tasks, such as looking for deals on computer hardware and other tech stuff.

Whenever I feel that it's time to upgrade my computer, I try not to rush into buying the new part right away. Often, a great deal is just around the corner and regretting paying a higher price is never fun. But it's also not much fun checking prices on multiple shopping sites daily. Thankfully, there are people who already do this and publish the deals on their RSS news feeds, like Tom's Hardware or Wired. The only problem is, these news feeds aren't just about shopping deals, but the technology sector in general. Thus, we're back to combing through tons of content hoping not to miss the one post about the deal of a lifetime.

This is where Avogadro One shines. In fact, I've been using it for exactly this for a few years, and it paid off, especially during hot shopping seasons around Black Friday and Christmas.

If you click this link, you will be taken to a copy of my Avogadro One project that tracks posts about good hardware deals posted by Tom's Hardware, Ars Technica and Wired, which should look like this (showing only a few posts here):

Avogadro One Shared Project shcreenshot

It's a read-only copy, which you can easily copy into your own account and modify as you please.

Under the Hood

While sharing this project is a nice opportunity to show off demonstrate the sharing feature, the main goal of this post is to illustrate how our app works so that you could set up your own projects with minimum hassle.

There are three main ingredients to create a useful project in Avogadro One:

1. A list of relevant sources. For tech deals, it makes sense to use RSS news feeds from websites that focus on technology, computers, hardware etc. As already mentioned, for this project I used three sources: Tom's Hardware, Ars Technica and Wired. It is actually possible not to define sources at all, but this makes the project rely more heavily on the next two components, which may worsen the quality of the results.

2. Keywords. For this project I crafted a query that might look intimidating at first glance:

title:(deal OR deals) OR sale OR "lowest price"

Let's take that apart:

  • title:(...) means that we want to find the words inside the parentheses in the title of posted articles. Here we want to see those articles whose titles mention the words deal or deals.
  • alternatively we want to see all articles that mention the word sale or the exact phrase "lowest price" (without the quotes) anywhere in the article (title, abstract, main body of the article - whatever is available).

After this, the project will only show articles from Tom's Hardware, Ars Technica and Wired that either mention deal or deals in the title or sale or "lowest price" anywhere. But this is often not enough to filter out irrelevant content. For example, Wired often posts deals for other product categories, while all sources periodically write about major corporate transactions described as a "deal" or a "sale". This is where the next step comes in.

3. Machine learning. By assigning manual relevant/irrelevant labels to the articles that made it past step 2, I've trained the machine learning  algorithm to identify irrelevant articles and hide them. Sometimes the algorithm makes a mistake, but it's all very transparent in Avogadro One: you can always correct an automatic relevance label to help the app do better. Here's what it looks like on the inside (i.e. when you are inside your own Avogadro One project):

Avogadro One project relevance screenshot

The articles with grey titles are all irrelevant. The ones with our logo icon were labeled automatically, the others - manually.

Final Notes

This is a rather simple use case and the project for it is quite simple too: just one stream that marries several sources with a single search query, refined by manual relevance labeling. This can be easily done with the free Avogadro One account. More complex projects might require multiple streams to be set up (e.g. applying different queries to different sources).

We encourage you to sign up and give it a spin and then share your own projects with us or your friends.

Leave a Reply

Your email address will not be published.