Jellyfish vs DX (2026): Allocation Model or Survey Standard

TL;DR: Jellyfish and DX are both enterprise, both quote-only, both Gartner-era leaders, and they answer different questions. Jellyfish tells your CFO where engineering effort went: a patented allocation model splits each engineer's week from Jira and Git exhaust, feeds capitalization reports, and supports R&D tax-credit claims. DX tells your leadership how the engineering system feels and performs: the Core 4 framework and the DXI survey, benchmarked against 800+ organizations, now with Atlassian's distribution behind it. Pick by primary buyer, finance versus productivity program. Then notice what neither flagship number lets you do: recompute it. That's where a third option enters.

Most "Jellyfish vs DX" pages are written by one of the two vendors. This one is written by a third, Busfactor, which means we have no reason to flatter either and every reason to be accurate, because the comparison only helps us if you trust it. Everything below is a snapshot as of July 2026, cited to each vendor's own material; both products move fast, and quote-only pricing means numbers from third parties are estimates, flagged as such.

Who each tool is for

Jellyfish is engineering intelligence with a finance co-buyer. Its customer logos are enterprise (Box, Priceline, DraftKings), its research corpus spans 20M PRs, and its signature promise is effort accounting without timesheets: connect Jira and Git, and the platform tells you - and your FP&A team - how engineering investment was allocated, what can be capitalized, and what may qualify for R&D tax credits.

DX is the measurement-science company. Founded on the research lineage behind DORA and SPACE, it owns the two frameworks the category now speaks in, the DX Core 4 and the Developer Experience Index (DXI), and was acquired by Atlassian for roughly $1B, which puts its instruments on a path into every Jira-running org on earth.

The honest first observation: these are not close substitutes. Plenty of orgs run both a finance-grade allocation tool and a DevEx survey program. The comparison gets real when budget forces a choice of which system of record to buy first.

How Jellyfish actually works: the Work Model

At the heart of Jellyfish is what it calls the Work Model: it "ingests, cleans, normalizes, and scores signals" from Jira and Git data exhaust, then expresses effort in FTEs: fractions of each engineer's week "spread proportionally across all the tasks the engineer worked on based on timeline evidence," rolled up to initiatives and investment categories. No timesheets, no instrumentation of developers. The current pitch adds AI-powered categorization "to get accurate allocation calculations regardless of data quality."

That model feeds DevFinOps: reports that "automatically determine capitalizable status," on-demand audit-ready outputs, and R&D tax-credit support. It also underpins AI Impact, vendor-agnostic AI measurement across Copilot, Cursor, Claude Code, Amazon Q, Gemini, Windsurf and more, derived "directly from Git, planning systems, and workflow data," plus token-spend views by tool, team, or initiative.

Two facts a buyer should hold at once. First: this genuinely works on messy data. That is why finance teams buy it, and why zero-setup allocation at enterprise scale is Jellyfish's moat. Second: the scoring and weighting function is the patent. Neither you nor your auditor can recompute how a given engineer-week was split, the AI categorization method is not disclosed, and no reproducibility guarantee appears anywhere in their public material. The defense is "patented" and "audit-ready" - authority claims, not recomputability.

How DX actually works: DXI and the Core 4

DX's flagship number, the DXI, is "a composite score derived from 14 standardized Likert-scale survey items," computed as a mean of driver sentiment scores and benchmarked against "over four million data samples from more than 800 organizations" and 40,000+ developers. DX states a one-point DXI increase "translates to saving 13 minutes per week per developer" - a coefficient from internal regression work whose derivation is not published.

The Core 4 wraps that into a four-dimension framework of Speed (diffs per engineer), Effectiveness (the DXI itself), Quality (change failure rate), and Impact (share of time on new capabilities), with collection explicitly layering system metrics, self-reported surveys, and experience sampling. Notably, half of the framework's key numbers are perceptions by design.

And credit where due: DX's own research is unusually honest. Their AI Measurement Hub puts actual AI productivity gains at "5-15%, rather than 50-100%," and warns that acceptance rate is a flawed measure because accepted code is often heavily modified before commit. When a vendor's researchers undercut the hype cycle with their own data, that's the good kind of vendor research.

Jellyfish vs DX: where each one wins

Jellyfish wins on the finance motion. Zero-instrumentation allocation that survives messy Jira, capitalization reports finance actually files, and the R&D tax-credit framing that can make the product self-funding in a CFO's eyes. DX offers R&D capitalization too, but Jellyfish has made the FP&A workflow its home turf. It also ships delivery-side surfaces DX doesn't center: its Life Cycle Explorer buckets issue flow into phases, and incident integration is part of the platform story.

DX wins on measurement science and distribution. It owns the category's standard (even rival vendors ship Core 4 rollout support), and its benchmark cohort (800+ orgs, 40k+ devs) is the deepest survey base in the field. Its methodological triangulation is real research craft, its published AI numbers are the most honest in the market, and post-acquisition it rides Atlassian's reach. If your goal is a credible, benchmarked DevEx program, DX is the category default.

Where both are exposed. Neither flagship number can be re-derived by the customer. Jellyfish's is an opaque scored model splitting person-weeks; DX's is an averaged survey sentiment with an unpublished money coefficient. Both now put LLM layers near load-bearing places: Jellyfish with AI categorization inside allocation and AI agents that "interpret results," DX with an AI chat over org data. And both are sales-led, annual-contract, quote-only motions: no self-serve trial, no published price to sanity-check.

Pricing: two quotes, no list prices

Neither vendor publishes pricing, so here is what's checkable as of July 2026. Third-party estimates put Jellyfish around $20-40 per developer per month with annual minimums reported near $30K, and community reports describe it as impractical under roughly 30-50 engineers. Estimates, not list prices. DX's pricing page describes modular pricing by licensed tools, developer-seat licensing, usage tiers for MCP access, contracts starting at a one-year term, and a free proof-of-concept for a subset of the org. Budget for a sales cycle either way, and re-verify anything above before it goes into a business case, because vendor pages drift.

Where the third option enters

Suppose what pushed you into this comparison is trust in the number - the board asked where the money went, or the CFO asked whether the capitalization would survive an audit. Then note that both of these excellent products fail the same specific test: re-run last quarter and match it byte-for-byte. A patented model can't offer that check by design; a survey can't even in principle.

That test is the entire premise of Busfactor: zero models in the metric path, numbers quoted from your own rows rather than generated, and exports that print their provenance (run id, engine version, ruleset version, content hash) so an auditor can re-derive the finance numbers instead of trusting an authority claim. On top of the checkable numbers sits what neither Jellyfish nor DX sells: a verdict. Metrics judged against published bands with the receipts linked, drains priced in your currency, a payback estimate on every prescription (the money surfaces, the assessment). And the people posture is structural: no per-person composite exists on any surface, and scorecards frame strengths and the cost of losing someone, never a rank. Flat, published, self-serve pricing; no quote required.

We've written the direct head-to-heads too: Busfactor vs Jellyfish and Busfactor vs DX concede where each beats us, with citations.

Jellyfish vs DX: the decision table

You are…	Pick
Finance is the co-buyer; capitalization + R&D tax credits are the job	Jellyfish
Enterprise Jira estate with messy data and no appetite for taxonomy setup	Jellyfish
Building a benchmarked DevEx program on the industry-standard framework	DX
Deep in the Atlassian ecosystem and heading deeper	DX
Need numbers a CFO or auditor can recompute, not take on authority	Busfactor
Want a graded diagnosis with priced fixes, self-serve, this week	Busfactor

If you're mapping the wider field first, start with the metrics that actually predict delivery, then the adjacent matchups: DX vs Swarmia for the survey-versus-telemetry fork at self-serve prices, and LinearB vs Faros AI for the mid-market versus data-platform fork.