Why your PKM tool migration keeps failing (and the cycle behind it)

PKM migration fails because the export-import step loses formatting and metadata, the new tool starts empty (so it feels alive), and within months the same write-only pattern reappears under a new logo. The fix is not picking a better tool. It is picking a tool that does not require migration: capture from anywhere, ask in your own words, no folder tree to recreate, no taxonomy to maintain.

In November 2022, TechCrunch reported that Bending Spoons had agreed to acquire Evernote, marking the end of an era for the 20-year-old note-taking app (TechCrunch). The deal closed in early 2023. Within twelve months, the free tier was restricted to 50 notes and the company had laid off effectively the entire pre-acquisition staff. The migration wave that followed was the largest in PKM history, and most of it failed.

Not failed as in did-not-finish. Failed as in finished and then collapsed again. People exported from Evernote, imported into Notion, hit Notion's friction, exported to Obsidian, ran out of patience with linking, and ended up with three half-populated vaults and a folder of orphan markdown files. This is the migration cycle. It is not a tooling problem. It is a structural problem with how PKM tools are built.

Why PKM Migration Keeps Failing

The failure has four layers, and they stack.

The trigger is rarely the tool. It is a price hike, a feature change, or an acquisition. Evernote restricted free accounts to 50 notes in November 2023. Notion raised AI add-on pricing. Obsidian Sync increased its annual rate. The user reaction is not slow consideration; it is sudden flight. The new tool is chosen on vibe, not fit, because the choice is being made under irritation. Bending Spoons subsequently raised $155 million in equity financing in February 2024 (TechCrunch), which signaled to many users that further monetization was coming and accelerated the exodus.

The export step loses fidelity. ENEX preserves text content but degrades web clips, nested lists, table structure, code highlighting, and some attachment metadata. Markdown is more portable but cannot represent rich content like Evernote web clips at all. Community importers like enex2notion (GitHub) exist precisely because the official paths leave gaps. Every migration ships with a tax in lost formatting, and the tax compounds across multiple migrations.

The import step has hard limits. Notion caps individual pages at 1,000 blocks. Large notes get truncated or split. Tables become unstructured text. Code blocks lose syntax highlighting. Nested lists flatten. The post-import vault looks like the pre-export vault from a distance and looks broken when you actually open notes you care about.

The behavior pattern survives the tool change. The thing that produced the graveyard in Evernote is going to produce a graveyard in Notion too. Same user, same capture habit, same absence of recall ritual. The tool change creates a clean slate UI, which feels like progress for three months. Then the slate fills, and the user is back where they started, with one extra failed migration in the rear-view.

A user on the Drafts forum captured the trigger in one line: "I refuse to be chained to this company any longer; it'll take forever" (forums.getdrafts.com). Another, in the same thread: "25k notes since 2007; I was so pissed, I clicked cancel." These are not isolated. This is the modal Evernote-to-anything experience between 2023 and 2026.

The Export-Import Loss: Where Data Actually Dies

The migration loss is concrete. It is not vague unease. It is specific format families that do not survive the round trip.

Web clips. Evernote's web clipper captured full HTML with images, styling, and original layout. ENEX exports preserve this as HTML, but no destination tool renders that HTML faithfully. Notion strips it. Obsidian converts to plain Markdown and loses the visual context. The web clips were one of Evernote's strongest features and are the most lossy element in any migration.

Nested lists. Evernote allowed deeply nested bullets and checklists with arbitrary indentation. Notion flattens nesting past a few levels. Markdown preserves indentation but loses styling. Long outlines that were the spine of a user's vault arrive in the new tool as semi-structured text that needs manual repair.

Tables. Evernote tables export as HTML tables. Notion attempts to convert to database rows but often produces unstructured text or blocks that are no longer tables. Obsidian users typically rebuild tables by hand in Markdown.

Attachments and inline images. ENEX bundles attachments inside the archive. Most importers extract them, but the link between an image and its position in a note can break. Inline images become attached files. Annotations on PDFs are usually lost.

Reminders and metadata. Custom reminders, modification timestamps in some cases, and tag hierarchies often do not survive. The reminder you set in 2019 to revisit a note in 2024 is gone. The chronology that gave the vault meaning is partially scrambled.

A second forum comment, from the same Drafts thread: "25k notes since 2007; I was so pissed, I clicked cancel" (forums.getdrafts.com). 25,000 notes over 16 years is the high end of the user base. For someone in that position, no migration path preserves the vault. The choice is partial loss now or paying indefinitely later.

The Empty-Tool Illusion: Why New Feels Better

The new tool always feels better. This is not because the new tool is better. It is because the new tool is empty.

An empty graph is legible. An empty folder tree is navigable. A daily-note template you set up yesterday feels alive. There is no backlog, no taxonomy drift, no graveyard. The relief is real, and it is a function of starting over, not a function of choosing correctly.

The relief lasts about three months. That is the typical interval between the migration completing and the new tool accumulating enough debris to reproduce the original problem. Three months is also how long it takes to rebuild a capture habit that fits the new tool's UI. By month four, you are saving things to the new vault. By month nine, the new vault has the same write-only problem as the old vault.

The pattern is consistent enough that experienced PKM users describe their relationship with the category in cycles: Evernote-era, Notion-era, Obsidian-era, post-PKM. Each era ends the same way: a buildup of unread saves, a feature change or pricing event, an export, a fresh start. The cycle is what fails, not any single tool.

Breaking the Cycle: A Tool That Does Not Need Migrating

The migration cycle continues because every tool in the category requires the user to recreate their taxonomy in the new tool. Folder trees do not port. Tag hierarchies do not port. Custom database views do not port. The migration cost is the cost of rebuilding the structure, and that cost is what makes the next migration feel like a fresh start.

A tool that does not need migrating has different properties.

No taxonomy to rebuild. No folders. No tags. No databases. There is no schema to recreate in the new tool because there is no schema in the current tool. The capture surface is the same on day one and day one thousand.

Source-of-truth lives at the URL or content level. When you save an article, the original is the canonical version. The tool stores enough to recall it. The tool does not try to be the rendering engine for the original. When you migrate or switch, the URLs are still URLs. The content is still on the open web (or in your own video, voice note, or screenshot).

Recall is by meaning, not by structure. You ask in your own words. The system does the work of finding what you saved. You do not need to remember the folder, the tag, the page hierarchy, or the database view. There is no structure for migration to break.

dEssence is built this way. You save articles, videos, voice notes, and screenshots from Chrome, Telegram, or the web app at dessence.ai. There are no folders, no tags, no organizing. When you have a question, you ask in your own words. If a better tool comes along in three years, your saves are URLs and content blobs that you can export to wherever. There is no taxonomy to lose because there was no taxonomy in the first place.

This is the structural answer to the migration cycle: pick a tool whose value does not live in the structure you have built inside it. Then there is nothing to migrate when the next price hike comes.

Honest About dEssence

The no-migration model fixes the cycle. It also has trade-offs. Four honest caveats.

• dEssence is in public beta. The recall layer works and capture surfaces are stable, but the product is still maturing. Edges are labeled honestly in-product.
• The paid tier is not finalized. The free plan caps at 500 saves. Pricing will be published before that cap matters for early users.
• There is no native iOS or Android app. Mobile capture works via the Telegram bot and a mobile-web fallback.
• If you have a 25,000-note Evernote archive and need every note ported with formatting intact, dEssence is not a migration target for the old vault. It is a place to start fresh for what you save next. The honest recommendation for a vault that size is to leave it as a searchable archive and stop adding to it.

The migration cycle is solvable. It is not solvable by picking the next tool more carefully. It is solvable by picking a tool that does not accumulate the kind of structure migration breaks.