I published an essay about different methods for organizing a collection of RSS/ATOM/JSON feed sources on February 1, 2024. Yukinu, who runs the excellent and cozy Yukinu Blog, published an essay inspired by my article titled RSS Feed Organization Strategies And New Feed Cost. I thought this was neat for two reasons. Firstly, Yukinu and I follow each other’s blogs via feed, so this exchange offers an example of how feeds can facilitate site-to-site discussions. Secondly, my own essay was inspired by other blogs and articles around the web, so it was fun to see it inspire a new article on the subject. In the spirit of demonstrating the power of feeds to connect independent websites to one another in addition to connecting sites with readers, I decided to think of an interesting (I hope, at least) follow-up to Yukinu’s response. Below, I will summarize Yukinu’s essay, offer some proposed addenda to Yukinu’s take on feed processing, and then examine the most important shared point of emphasis of our essays – determining purpose first and solutions second.
(It goes without saying that you should read my original essay and Yukinu’s reply before continuing with this article. If you are unfamiliar with feeds, see my introduction to RSS and other feed formats, my argument in favor of using feeds and read-it-later tools, and our feeds here at The New Leaf Journal and related projects.)
Yukinu’s Take
Below, I will respond to different parts of Yukinu’s essay before spinning off from it into a new topic.
Priority-Based Feed Organizing
Yukinu previously organized feeds by priority:
[I]n the past my strategy for organizing my RSS feed used a combination of the categorical and ranking methods; I organized my feed into 3 ranks (high, medium, and low priority), and categorized them based on topic.
Yukinu
However, Yukinu explains that this approach frayed with a large number of feeds, noting that reasons included how most feed readers structure feeds, needing additional categories, and uncertainty about how to process low priority feed items.
I never went for a full priority approach, but I do have some quasi-prioritization in my own feed set. I explained in my essay that my current system for organizing feeds is based on update frequency (daily, weekly, and sporadic), but I also have categories for local news and links (links are for feeds that primarily link elsewhere). My decision to separate local feeds (local for me being New York City news) is a sort of prioritization since I enjoy reading local news in the morning.
I agree with Yukinu that a full priority approach would show cracks with a large feed count. I would also question having a “low priority” category, at least how it is used here, for the reasons articulated by Yukinu.
A different way of thinking about priorities
Yukinu’s discussion of priorities gave me an idea about priority-based systems. Yukinu’s former system used priorities based on reader interest. For example, I always read Yukinu’s new posts when they pop up in my feed – so I suppose that would be high priority for me under the priority system presented. But what if we thought of priority in a different way? I suggest time sensitivity.
I have some feeds that have time sensitive articles and others that do not. For example, my news feeds often have time sensitive articles, pieces that may be more relevant now than five days in the future. Conversely, while I always read Yukinu’s blog posts when I see them, they are usually not time-sensitive, as in they are not about an ongoing and evolving event. The same is generally true of my New Leaf Journal articles with rare exceptions (even then I try to write current events articles in a way that will be interesting even after the event in question is no longer current). My original feed article and Yukinu’s response should have the same value in 2027 as they do today.
(Note: There is a fair case that many articles about current events are actually unimportant and the fact that their value may decrease in 48-72 hours is evidence that they were entirely skippable in the first place. I largely agree and I choose feeds to avoid those types of articles. When I discuss time sensitivity, I am referring more to the idea that articles about an event may build off one another such that having read articles A and B will make it possible to get the most out of article C.)
It may be worthwhile for curators with large feed collections to categorize or otherwise tag or code feeds that regularly post time sensitive articles and blogs that are best read close to when they are published. At the moment, I have a good intuitive sense of what is time sensitive and I am able to make quick decisions on what to do with new feed items, but I will think about whether my feed usage may be enhanced by time sensitivity-based labels, which my current feed reader happens to support.
Yukinu’s New Method
Yukinu formulated (pun intended for reasons you will see) a new strategy by considering three questions: What is a feed, why do we curate feed collections, and how do we process feeds? In addition to reading Yukinu’s explanation, see my introduction to feeds and feed readers.
Yukinu notes that feed items – that is articles or other media (e.g., posts, podcasts, videos) are either signals or noise. Signals are feed items the curator is interested in. Noise refers to feed items the curator is not interested in. Yukinu succinctly explains that “[t]he reason we curate our own feeds is to accumulate the former piece of information, Signals, at a more efficient rate than passively acquiring them (for example, through local news channels broadcasting in the background).” Yukinu then suggests a purpose for organizing feeds: “[A]ccumulate as many signals in as little time as possible.”
Yukinu then proposes thinking of organizing feeds as thinking about how to maximize the signal ratio, that is to maximize the number of signal items and minimize the number of noise items from a given feed collection. For example, different feed curators may devote different amounts of free time to working through feeds each day. Thus, Yukinu recommends considering how much time one will devote to feeds and how long it takes to process signal items and noise items in order to figure out how to keep a feed collection manageable.
My summary is simplified and I recommend reading Yukinu’s article – which includes illustrative formulae and charts – to fully understand the method and, perhaps, apply it in full or in part to your own feed collection (the general idea is applicable to other forms of reading/watching/listening).
My Thoughts On Yukinu’s Method
I touched on the signal to noise point indirectly in my essay – for example where I linked to an earlier article that I had written about subscribing to category-specific feeds, where available. To use a simple example, I subscribe to the New York Post’s Metro feed for local news instead of the paper’s main RSS feed (see essay on the method). This is to increase my signal to noise ratio. I am interested in the local headlines but I do not read the Post for national or international news (on the rare occasion the Post breaks a non-New York story, I will usually discover its reporting through my other sources), much less clickbait and celebrity nonsense. There are some sources that I avoid because (A) they do not offer granular feeds and (B) the main feed has too much noise and too little signal.
I agree fully with Yukinu’s emphasis on signal to noise as a principle, but I apply the ideas with a different points of emphasis respect to my own feed collection. While I bear in mind the idea of signal-to-noise when I choose which feeds to add to my collection, I do not consciously consider how much available time I have to devote to feed processing each day or how long it takes to process one feed or another. Moreover, I am more concerned (to the extent I am concerned) with how long it takes to read signal items than how long it takes to distinguish them from noise items.
(Note for clarity: For the purpose of the instant discussion, I am using processing to refer to going through new feed items and identifying individual feed items as signal or noise. This is distinct from actually reading (or watching or listening to) feed items processed as signals.)
I devote a decent amount of time to reading feed items each day. I usually go through my feeds in the morning while having breakfast and again in the evening over dinner. I do not set aside a specific amount of time for feed reading, that is just my ordinary practice. When I open my feeds, I usually swipe away items that I am not interested in reading in each of my feed sections before reading. I am usually able to discern from the title and description of a feed item whether it is signal or noise to me and make a quick decision. While I never timed myself (and note that my feed reader is phone-based, unlike Yukinu’s computer-based example) I doubt that I ever spend 10 minutes on processing despite the fact that my feeds typically have more noise than signal.
Despite some differences in how I think about processing feed items, I agree with Yukinu’s definitions of signals and noise and the general principle that one should want to maximize the number of signal items relative to noise items in a feed set. I also agree that there are additional relevant factors to consider such as the tools a particular feed reader may offer to reduce the amount of noise in feeds or that some high signal-to-noise feeds may nevertheless be processing-intensive because of the number of new feed items.
I submit a few additional points to consider in trying to craft a relatively high signal-to-noise feed collection, regardless of processing methodology.
Focusing solely on efficiency may overlook the value of some signals vs other signals. For example, imagine that we have two feeds that produce three articles per work day, so 15 articles from Monday through Friday. Let us imagine that Feed A produces 8 signal items on average and Feed B produces 4 signal items on average. If our sole objective is increasing signal to noise ratio, Feed A is the more valuable feed. However, that may not tell the whole story. To use an example, I subscribe to the news feed from Anime News Network. I will venture that its signal rate is in the 30-40% range, which is certainly better than my Hacker News page one feed. But is it more valuable? Most of the signals from Anime News Network are short news posts about upcoming anime or similar news. My Hacker News feed sometimes leads me to new tech essays, small web sites, or apps that I may want to try. While the Hacker News feed has more noise (for me at least) than the Anime News Network feed, its signals are often more fulfilling. While I would not recommend creating a scoring system for item value for no other reason than such a project runs the risk of spending more time ranking than reading signals, feed curators should consider what they glean from signals from a particular feed.
Along similar lines, if one has time constraints, he or she should not overlook the types of articles that are produced by certain signals. For example, I publish some unusually long articles. I doubt that most of our feed subscribers (I have no idea how many feed subscribers we have) are only following me for short photo posts or children’s poetry re-prints. Bearing in mind that the ultimate purpose of maintaining a feed collection is usually (but not always, there are special cases – notably in the research sphere) for reading interesting articles, one should consider the weight – in terms of word count or media length, of signal items.
If one is applying a strict signal-to-noise equation, it is also important to consider that what may be a signal in isolation can be noise in the context of a particular feed collection. For example, I subscribe to feeds that sometimes cover the same signal story. I have had cases where The Washington Times, Smithsonian Magazine, and Atlas Obscura have all covered the same story within the span of a few days. I have experienced similar occurrences with some of my video game and tech feeds. There are cases where a subsequent article on the same subject is better than the first article, but there are others cases in which one account is sufficient. In the latter case, a signal-quality story can become noise in a particular feed collection. This is inevitable for people who subscribe to multiple feeds covering the same or similar topics. In some cases two feeds may truly be redundant (e.g., I dropped Crunchyroll’s anime news feed because it is not meaningfully distinguishable from the broader Anime News Network feed), but there are others (see my Smithsonian Magazine/Atlas Obscura case) where the overlap does not create undue redundancy.
Purpose in Feeds and Tech
Let us return to Yukinu’s description of the purpose of maintaining a feed collection:
The reason we curate our own feeds is to accumulate the former piece of information, Signals, at a more efficient rate than passively acquiring them…
Yukinu
This is the most important line in Yukinu’s article – even more than the particular method for curating feed collections. I articulated a purpose for choosing a method to organize feeds in the conclusion to my article:
The only general rule to organizing feeds is to always distinguish means from ends. The end of maintaining a feed collection is to follow and actually engage with good writing and media on one’s own terms in his or her own digital home. Organizing feeds is simply a means to that end. Confusing means and ends leads one into focusing on organizing instead of engaging with good writing and media – the sort of disordered productivity I have warned against in these pages.
N.A. Ferrell
These two statements of purpose are not talking about precisely the same issue. Yukinu is describing why one would curate feeds at all while I am referring to why would would organize a set of feeds into categories or similar. These passages do hint at some differing points of emphasis – Yukinu’s quote focuses on curating feeds to accumulate useful information efficiently, while I focus more on controlling one’s digital information environment. However, what is more important than is that we both considered the purpose of curating a feed collection and organizing it before discussing how best to fulfill our objectives.
Purpose is a guardrail for organizing feeds or choosing different methods of doing things in the digital sphere. If one dives right into curating a feed collection and messing around with a feed reader without considering why he or she is making a feed collection or messing around with a feed reader, he or she runs the risk of spending more time trying to optimize their optimizations. I quote myself:
People who fall victim to the productivity porn siren song aspire to achieve some abstract concept of being productive, but without actually considering what they want to produce.
N.A. Ferrell
A carefully organized list of feeds provides little value to a feed curator if he or she does not actually use the feeds to achieve some meaningful end. The only reason to organize feeds into categories or to improve a feed collection’s signal to noise ratio is to further some end beyond the acts of organizing or optimizing. The end may be ensuring that only the best writing reaches your digital doorstep, finding signals in the most efficient way possible, or keeping tabs on important events in your professional field. One reason I did not advocate for one particular method of organizing feeds over another (despite sharing my own) is because people may have very different reasons for curating a feed collection and thus use different tools for parsing it.
Purpose is valuable outside of the feed context. For example, people may keep tabs on the latest computer productivity tools without considering whether using any of those productivity tools would facilitate more actual production. One may create a big wish list of games on Steam to keep tabs on the next Steam sale while already having 150 unplayed games in his or her Steam library (see my thoughts on unproductive window shopping). Someone may read my very positive review of the Ghostwriter markdown editor and decide to download despite not having any particular use-case for a markdown editor. These are just a small number of examples. Considering purpose and properly distinguishing means from ends allows people to tailor tools as a means for actually accomplishing something instead of tailoring for its own sake. Regardless of whether a particular feed curator agrees with my principles for organizing feeds or Yukinu’s approach, he or she will do well to follow us in considering ends first and means second, thereby spending time crafting solutions for real problems instead of solutions in search of problems.