FILE №.07Engineering · Field Report · Vol. 1
Filed May 2026 Author M. Fuhlmann Pages 1–14
← Knowledge Library
Field Report · From the Later.com talk · March 11, 2026

How to measure developer impact without punishing the people you measure.

SubjectEngineering effectiveness
Org size71 engineers
Tools observedDXI · GitHub · LaterBot
AI exposureCursor · Claude Code
Talk length~50 min · 15 slides

Contents

  1. The four pitfalls of developer impact measurementp. 01
  2. The baseline — what DXI shows, and where it stopsp. 03
  3. LaterBot — the homegrown layer that closes the gapp. 05
  4. The five pillars, one ladder, no black boxesp. 08
  5. Stacking it together — measuring AI's precise impactp. 10
  6. Eight things that actually workedp. 13
§01The Problem

The four pitfalls of developer impact measurement.

Every engineering org wants to know if it's effective. Most reach for the wrong instruments. Four traps explain why classical attempts fail — and why measurement still has a bad reputation among the people being measured.

In the opening of their talk, O'Brien and Foley lay out the catalogue of failure modes that have made the field of developer productivity measurement quietly notorious. Each of the four is recognisable to anyone who has sat through a quarterly review built on the wrong numbers.

01
Measuring activity, not outcomes
Lines of code. Tickets closed. Hours logged. All cheap to count, all trivial to game, all uncorrelated with whether the business moves.
02
Goodhart's Law
When a measure becomes a target, it ceases to be a good measure. Engineers respond to incentives faster than the incentives can be repaired.
03
Vibes-based management
"How's the team doing?" answered from memory, filtered through hallway impressions and the optimism of the last all-hands. The data doesn't survive a hard quarter.
04
Surveillance vs. diagnostics
If a metric creates anxiety instead of clarity, it's the wrong metric. Surveillance hardens defences. Diagnostics invites collaboration. The data is often the same; the framing decides the outcome.

The root cause

The four pitfalls share a single underlying confusion: they conflate activity with outcomes and individuals with systems. A team that ships sixty pull requests a week looks more productive than one that ships twenty — until you notice that the sixty-PR team is churning through bug fixes while the twenty-PR team is shipping a quarter's worth of strategic features.

They also share a tone: punitive. Surveillance creates a defensive posture; defensive teams don't improve. Diagnostics — measurement aimed at understanding the system — creates a collaborative one. Same data, different framing, opposite outcomes.

§02The Baseline

The baseline — what DXI shows, and where it stops.

Later runs the Developer Experience Index survey via GetDX every quarter. It's a strong baseline — but it tells you about experience, not system shape. A snapshot taken on January 4, 2026, with all 72 engineers responding.

DXI is a sentiment instrument. It captures what engineers experience — how they feel about planning, documentation, debugging, ease of release. That experience is invaluable: people who feel blocked usually are blocked. But sentiment lags reality by weeks, and the score alone never reveals the underlying system that produced it.

Figure 2.1 · DXI snapshot, Later.comDXI 59 · −5 vs Industry P50
Jan 4, 2026 · 72 of 72 responded (100%) · 67% triaged
DriverScorevs P50Rank
Documentation36−8#1
Clear direction61−14#2
Planning process66+1#3
Cross-team collaboration55−4#4
Production debugging46−6#5
Build and test59−1#6
Deep work61−4#7
Code maintainability70−10#8
Incremental delivery43−8#9
Customer focus45−9#10
Review turnaround72−8#11
Change confidence81+1#12
Ease of release63−9#13

Reading the table

Only two drivers post a positive delta against the industry median — Planning process (+1) and Change confidence (+1). Everything else trails. A score of 59 against a P50 of 64 is not catastrophic, but the pattern of weakness (documentation, customer focus, incremental delivery, ease of release) suggests a system, not just a mood.

The number is the headline. The shape of the deltas is the diagnosis. And the diagnosis is exactly where DXI hands the problem to whatever you build next.

§03The Custom Layer

LaterBot — the homegrown layer that closes the gap.

GetDX tells you about experience. Later also needed real-time operational visibility. No vendor knows your org shape — your teams, your domains, your definition of healthy. So Later built one. With AI generation tools.

A bespoke layer used to be an expensive choice. AI-native iteration changed the economics. What used to require a six-month internal-tooling sprint can now be drafted in a week and refined in another. That shift is what makes the next section in this report possible at all — and is the reason the speakers stress it as a precondition rather than an option.

The four questions LaterBot exists to answer

  1. What's actually moving right now?
  2. What's stuck — and where is the WIP piling up?
  3. Where is quality degrading?
  4. Which teams are overloaded?

None of these can be answered by a quarterly survey. None can be answered by a third-party tool that doesn't know your domains. They demand a layer custom to the organisation that uses it.

This system exists for two reasons — to improve the lives of the people who build software, and to improve the software we ship. If it isn't doing both, we change it.
The LaterBot Manifesto

The intellectual sources

The system is consciously borrowed from four bodies of work — each a school of thinking about how creative knowledge work actually flows:

ReinertsenQueue theory
GoldrattTheory of constraints
BrooksEssential complexity
DemingImprove the system

The four design principles

  1. Metrics illuminate, they don't surveil.
  2. We measure outcomes and flow, not activity.
  3. Trends matter more than snapshots.
  4. Everyone sees the same numbers.
§04Structure

The five pillars, one ladder, no black boxes.

No single number captures engineering health. Five do. LaterBot organises its view around five orthogonal dimensions — and unifies them onto a single five-tier ladder so anyone in the company can read any pillar at a glance.

The five pillars earn their place because each answers a different question and none can be collapsed into the others. A single number lies by omission. Five numbers tell a story.

Linear Hygiene
Clean data as foundation. Without trustworthy issue tracking, every downstream metric is noise dressed up as signal.
WIP Health
Focus constraints: ≤ 5 issues, single project. Limit work-in-progress; expose the queue rather than hiding it.
Project Health
Dual-source: self-reported status combined with velocity trajectory. Two independent signals catch what one would miss.
Productivity
TrueThroughput: merged PRs weighted by complexity. Not raw PR count — outcomes weighted by what they actually cost to ship.
Quality
Bug debt trend: load, net change, age. The trajectory matters more than the absolute number on any given day.

One ladder, five tiers

Every pillar uses the same five-tier status system, so anyone in the company can read any pillar at a glance without learning a new vocabulary. Learn the scale once, read every dashboard.

Peak Flow
Strong Rhythm
Steady Progress
Early Traction
Low Traction

The ladder is deliberately verbal rather than numeric. Engineers respond differently to "low traction" than to "score: 2/5". Words carry shame less easily than numbers, and a label invites a question — "what would move us up one rung?" — that a raw score does not.

§05AI Impact

Stacking it together — measuring AI's precise impact.

How do you measure and maximise the impact of AI-assisted development? Layer the systems. Each layer adds a different lens. The combination is what makes "AI is helping" auditable instead of anecdotal.

The trick isn't a single new metric. It's stacking four lenses — operational, sentiment, system, and threshold — so that a claim of impact has to survive all four to count. Each layer in isolation can be gamed or mis-read. The combination is robust.

Layer 01 Baseline

GitHub + GetDX

Percentage of AI-assisted lines. Cohort of adopters vs. non-adopters tested against operational metrics.

TrueThroughput 7.62 vs 3.54
2.15× lift
Layer 02 Sentiment

DXI driver comparison

Quarterly sentiment layered on operational data. Are adopters more confident, more focused?

Change confidence 79 vs 30
+49 pts
Layer 03 System

LaterBot adoption cohorts

High vs. low adoption against all five pillars. WIP health, project health, quality trajectory.

Change fail rate 4.3% vs 7.3%
−41%
Layer 04 Threshold

Beyond adoption — impact

Not yes/no adoption. Percentage of engineers using AI above an effectiveness threshold informed by the synthesis of the three prior layers.

39%
28 of 71 engineers
39%
AI effectiveness score
Engineers active on tracked AI tools above the effectiveness threshold.

Cursor and Claude Code, fourteen-day rolling window. Twenty-eight of seventy-one engineers are currently using AI effectively — not just installed, not just clicking accept, but operating above a synthesised cross-layer threshold.

Tab acceptance rate · 19.6 %   |   Accepted lines rate · 83.4 %

Why threshold beats adoption

"Did you install Cursor?" is a binary question. "Is your work measurably better because of it?" is a continuous one. Tab acceptance and accepted-lines rates are informational breakdowns — they describe one modality (tab completions) and miss agent-mode workflows entirely. The effectiveness threshold survives Goodhart's Law because it cannot be gamed by clicking accept more often.

A note from the wider industry

The summit-partners framing — borrowed from Coinbase's internal reporting — quantifies AI output in normalised units of work for an expert engineer, showing the proportion generated by AI versus human-written code.

Reference · Coinbase, via Summit Partners
"Quantifying AI's precise impact is hard but we have to try."
Approximation, two weeks ago
416.5units/week
Baseline
Approximation, today
514.6units/week
59th percentile · +23.6%
The metric is imperfect — the speakers admit it. The exact number is not the point. The point is that two weeks of disciplined measurement made the system's trajectory visible at all.
§06Takeaways

Eight things that actually worked.

The principles that survived contact with reality. Pick one and try it this week. You don't need the whole system to start — publishing one formula or sharing one ugly number will change how your team talks about its work.
Start with philosophy, not dashboards.
Measure conditions and flow, not individual output.
Publish the formulas. No black boxes.
Diagnose the system, not the person.
Respect the mirror. Share the ugly numbers.
Build for your org shape. AI-native iteration makes it practical.
Everyone sees the same numbers.
If it isn't improving lives and software, change it.

The list is shorter than it looks. Items one through three set the posture. Items four and five set the social contract. Items six and seven set the architecture. Item eight is the kill-switch that prevents the whole system from drifting into surveillance over time.

Developer impact is notoriously challenging to measure, but it's achievable — with the right abstractions and processes that are collaborative, not punitive.
Sean O'Brien & Justin Foley · Later.com · March 11, 2026

For DocuSketch — three actions

Later's situation rhymes with ours. Seventy-odd engineers across several domains. AI tools rolling out — Cursor and Claude Code there, Claude Code via Bedrock here. Leadership that wants to know if the investment is paying off, honest enough to admit that the obvious metrics don't work.

One — write the manifesto. Before another dashboard is built, document what we believe about how engineering work works at DocuSketch. One page. Four sentences. Signed.

Two — choose the five pillars. They don't have to be Later's, but they have to be non-collapsible and reportable on one shared ladder.

Three — build the four-layer AI-impact stack. Baseline (GitHub) → sentiment (a four-question pulse, monthly) → system (LinearB cohorts) → threshold. The threshold percentage becomes the KPI for the AI-first transformation, replacing "is the tool installed?" as the lead indicator.