Earth Notes: On Website Technicals (2018-03)

Updated 2024-05-31.
Tech updates: Auto Ad imbalance, incremental build, readability, tags, ad borders, TechArticle and Report, SoftwareSourceCode.
Readability was an important goal for this month. The tools have continued to help me curb my tendency to ramble on and on without a full stop! In passing this also prompted me to be a good deal more generous with images in my text: more and larger.

2018-03-30: SoftwareSourceCode

Today I discovered the schema.org SoftwareSourceCode markup.

I have marked up a couple of links to code repos, and many (~90) code snippets, as a test. Google's Structured Data Testing Tool is able to parse all of them. I'm slightly concerned that a search engine will have difficulty making sense of code text fragments that have had HTML metacharacters such as '&' escaped to allow them to be rendered correctly. Google's tool seems to be doing the right thing.

2018-03-26: Ads on Lite

As an experiment I shall try restoring ads for the 'lite' site. It gets 30% of traffic, but I notice that the majority of ad revenue is from mobile devices on my full site, which is odd.

I turned off most ads for mobile/lite a few months ago given the page weight, and their apparent ineffectiveness.

Note that some pages always carry ads (both full and lite site versions), and some pages never carry ads. I never run ads on the home page, for example.

2018-03-25: TechArticle and Report

For all my main pages I have been using the schema.org Article type. There are subtypes such as TechArticle which I am now using for the more techie pieces (tagged 'tech'), and Report for any (other) 'research' tagged pieces. I don't think my 'research' warrants ScholarlyArticle. I'm not sure whether review-tagged pages should move to a subtype too.

I should maybe try to make use of the proficiencyLevel field of TechArticle. 'Expert' seems right level, 'Beginner' being the only alternative given everywhere.

2018-03-20: Moar Readability

I continue to chug away, improving the readability scores of pages for which I am now getting warnings.

I am feeling a little inspired today (after a Governors' Visit Day to our local primary school) by the simplicity of some of the independent creative wiring of year 2! Short can be good.

Expecting instant changes in visitors would be foolish, however: Google: Can Take 6+ Months to See Rankings Improve from Quality Improvements.

It's slow and hard work bringing some pages up to target readability. Especially the consumer-facing pages that need the higher scores. The text can start to take on a "Peter and Jane" feel.

Given experience of the slog involved, the default threshold currently remains at 55, 'tech-heavy' is now set at 40, and consumer-friendly at 72. (Wikipedia claims 80--90 is conversational English for consumers, but I think that it would make me cry to go there, at least until I have a very reliable scoring system that does not choke on random parts of my page headers and footers.)

Since the common borderless 'native' Auto Ad format seems to be doing well for revenue, I have stripped the red border and 'Advertisements' tag from the generic ads that I still insert manually and algorithmically. I have even reset the actual ad style from my long-standing yellow background and red border that I use across multiple sites, back to AdSense defaults.

2018-03-18: Article Tags

I have generalised the tagging mechanism to allow any article zero or more tags. As of today six tag 'words' are defined to indicate things from tech-heavy to social-media targeted. This is partly to drive a variable readability threshold, and partly to annotate the main site-guide links.

In passing: when I used Google to find a "CSS colour picker", G brought up its own interactive one directly above the search results!

2018-03-11: Readability

I'm now computing (with a view to improving) the Flesch Reading Ease of my core pages, at the moment using the reado npm package. Currently I have set the default warning threshold at 55 fairly difficult to read, but for core non-technical pieces will aim for ~85 "conversational English for consumers," while maybe letting heavy tech pieces get as low as 40 to 45.

I may look for a replacement for reado in due course, as for quite a number of pages it reports 'NaN' rather than a useful score!

I have come across the suggestion that there should be an image for every 100 words for maximum shareability. That would be a tall order!

2018-03-10: Site Incremental Build

I have been tweaking the pages-incr make target to ensure that if there is nothing that it can build then it terminates quickly, and conversely, if run enough times it will build everything that it can (the core pages of full and lite site versions).

2018-03-08: Google AdSense Auto Ad Imbalance

I have an experiment with AdSense Ad Balance testing for a fill rate of 100% or 50% and for now 50% is not looking good. Possibly connected, possibly not, Auto Ads and Ad Balance between them seem to be inserting lots of unused (white) slots on my pages which look messy, and I should think could be avoided, even without causing scroll 'jank'.

I think that with Auto Ads and having tweaked some other settings to allow a slightly wider range of ad types, I am seeing a tiny increase in my still minuscule revenue!

2018-03-19 update: I cancelled that experiment and I have started a 100%/90% replacement experiment, which at first blush seems better.

2018-05-27: I cancelled that second experiment as it was showing ~25% RPM drop after an initially promising start. Also, given that in time for GDPR on the 25th I stopped most advertising other than EOU, so maybe half the traffic, I'm not clear that continuing the last 20 days out of 90 would be meaningful. Over 20,000 impressions had been racked up.