What I Got Wrong About WordPress in 2020

Looking back at WordPress writing from five years ago is instructive. Some predictions held up; many didn't. The patterns of what tends to be wrong are useful for evaluating current predictions about platform direction.

The honest review of what I (and many in the WordPress community) got wrong in 2020-2021 isn't an exercise in self-flagellation; it's calibration. If the prediction style that produced wrong calls is still being used today, the current predictions in that style deserve skepticism.

Wrong: Gutenberg would fail to replace classic editor for serious sites

The argument in 2020: Gutenberg was clunky for serious editorial work, the Classic Editor plugin had millions of installs as a defensive measure, and WordPress had a history of failed re-architecture attempts (the customizer, the REST API as a frontend layer).

The argument turned out wrong on the substance. Gutenberg matured over five years. The block editor is now the default for most new WordPress sites. Full Site Editing extended the model beyond posts. Theme developers committed to blocks. The Classic Editor plugin still has installations but the trend is clearly toward the block editor.

The lesson: I underweighted the platform commitment. Automattic and the broader WordPress project invested heavily in Gutenberg over five years. Sustained investment over time beats the initial rough launch.

Wrong: AMP would become essential for mobile SEO

The argument in 2019-2020: Google was promoting AMP, mobile search results favored AMP pages, and sites that didn't adopt AMP would lose mobile traffic.

The argument turned out wrong on what Google would do. Google reduced AMP prioritization in 2021-2022. The performance benefits AMP provided became achievable through standard responsive web techniques. AMP usage declined; most sites that adopted it have removed it.

The lesson: I overweighted Google's stated direction. Google promoted AMP heavily, then quietly de-emphasized it as the web platform caught up. Following Google's stated direction without monitoring whether the direction held was a mistake.

Wrong: WordPress would lose ground to Webflow and Wix

The argument in 2020-2021: visual builders were getting better, the WordPress experience for non-technical users was still rough, and the trend toward "WordPress is too complex" would accelerate.

The argument turned out partially wrong. WordPress's market share remained roughly stable at 40-43% of all websites. Wix and Squarespace grew but mostly grew the market rather than taking WordPress share. The "WordPress is dying" prediction never materialized.

The lesson: I underestimated the installed base advantage. WordPress's network effects (themes, plugins, developer expertise, hosting optimization, content portability) create switching costs that visual builders can't overcome with feature improvements alone.

Right but for partial reasons: PHP would remain dominant

The argument in 2020: PHP would remain WordPress's foundation despite developer interest in newer languages. This held up but the reasons were different than I expected.

I thought PHP would remain dominant because alternatives would fail to gain WordPress integration. The actual reason it remained dominant is that PHP itself improved substantially. PHP 8 added significant performance improvements, type safety, and modern features. The argument that PHP was outdated became less true over time.

The lesson: technologies sometimes improve enough to defuse "they're falling behind" predictions. Don't assume incumbency means stagnation.

Wrong: headless WordPress would become the default

The argument in 2020-2021: developers preferred JavaScript, frontend frameworks were maturing, and headless WordPress would become the default architecture within 3-5 years.

The argument turned out wrong on adoption rate. Headless WordPress grew but remained a minority architecture. The operational complexity (editor preview, plugin compatibility, structured data handling) prevented broad adoption. Most sites that started headless either continued with the complexity or migrated back to traditional WordPress.

The lesson: developer preferences don't always determine architecture choices. The operational tax of headless is real and falls on people who aren't the developers who chose the architecture. The decisions get made (or remade) by people who experience the operational costs.

Partially right: managed WordPress hosting would consolidate

The argument in 2020: WP Engine, Kinsta, Pressable would either consolidate or be acquired by larger entities, and the market would have 3-4 dominant managed WordPress hosts.

The consolidation happened (WP Engine acquired Flywheel, became part of larger entities; Kinsta remained independent but grew; Pressable was part of Automattic). The dominant-host count is roughly right. But the prediction about specific market structure was too confident; the actual structure has more nuance with multiple tiers and specialized providers.

The lesson: directional predictions hold better than structural ones. "Consolidation will happen" was right; "the market will look exactly like X" was overconfident.

The patterns in being wrong

Looking at the predictions that didn't hold up, the common failures:

1. Underweighting incumbency. WordPress in 2020 had advantages I didn't fully credit. Same for PHP. Predictions of major shifts often underestimate how much existing structure resists change.

2. Overweighting stated direction from gatekeepers. Google's promotion of AMP wasn't a reliable signal of where Google would stay. Similarly, Automattic's stated direction sometimes shifts after a couple of years.

3. Underweighting operational costs. Headless WordPress has real operational costs that aren't visible from the developer side. Predictions about adoption need to account for who actually bears the costs.

What to do with this calibration

For current predictions about WordPress direction, apply skepticism in the same patterns. Predictions that bet heavily on platform shifts, on Google's stated direction, or on developer preferences overcoming operational costs deserve more scrutiny than predictions that bet on incumbency, on organic improvement of mature technologies, and on operational considerations defeating architectural aspirations.

The honest framing: I was confident in 2020 about predictions that turned out wrong. I'm probably similarly confident now about predictions that will turn out wrong in 2030. The discipline is to flag the bets that share patterns with previously-wrong predictions, and to weight them accordingly.