# Incremental Improvement Practice

> Framework for pursuing "better" through small, consistent changes rather than waiting for breakthrough solutions. Based on Atul Gawande's philosophy from "Better.

- Skill: `sethmblack/incremental-improvement-practice` (Agent Skill)
- Install (CLI): `npx skillmds add sethmblack/incremental-improvement-practice`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethmblack/incremental-improvement-practice/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: sethmblack (https://skillmd.com/u/sethmblack)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/sethmblack/incremental-improvement-practice

---


# Incremental Improvement Practice

Framework for pursuing "better" through small, consistent changes rather than waiting for breakthrough solutions. Based on Atul Gawande's philosophy from "Better."

---

## When to Use

- Can't fix everything at once
- Waiting for resources that aren't coming
- Need improvement within constraints
- Grand transformation isn't realistic
- Want sustainable progress over time

**Trigger Phrases:**
- "I can't fix everything"
- "Where do I start improving?"
- "We need a big change but can't make one"
- "How do I get better incrementally?"
- "What can we do with what we have?"

---

## Inputs

| Input | Required | Description |
|-------|----------|-------------|
| area | Yes | What you're trying to improve |
| constraints | No | Resources, time, authority limitations |
| current_state | No | Where things stand now |
| desired_state | No | What improvement looks like |

---

## Core Principle

> "Better is possible. It does not take genius. It takes diligence. It takes moral clarity. It takes ingenuity. And above all, it takes a willingness to try."
> — Atul Gawande

The opposite of better isn't "worse"—it's "waiting for perfect."

**The trap:** "We need a complete overhaul / more resources / better systems before we can improve."

**The reality:** Small improvements, consistently applied, compound into transformation.

---

## The Three Requirements

From "Better," Gawande identifies three requirements for success:

### 1. Diligence
Attention to detail, following through, not cutting corners.

Not glamorous. Not innovative. Just consistently doing the small things that matter.

**Example:** Hand hygiene in hospitals. Every doctor knows to wash hands. Few do it every time. The difference between 70% compliance and 90% compliance is thousands of infections prevented.

### 2. Doing Right
Ethical clarity even when difficult.

Improvement isn't neutral. It requires moral commitment to getting better, even when the easy path is "good enough."

### 3. Ingenuity
Finding better ways within constraints.

Not waiting for new resources or technology. Working with what you have to find what's possible now.

---

## Workflow
### Step 1: Accept Constraints

Don't start with "if we had more time/money/people..."

Start with: "Given what we have, what can we do?"

Constraints aren't excuses—they're the canvas.

### Step 2: Find the One Thing

**Ask:** "What's one thing we could do this week to be slightly better?"

Not five things. Not a transformation plan. One thing.

**Criteria for the one thing:**
- Within your control
- Doable this week
- Measurable (you'll know if it worked)
- Connected to the outcome you care about

### Step 3: Make It Specific

Vague improvement doesn't stick.

| Too vague | Specific |
|-----------|----------|
| "Be more responsive" | "Respond to every email within 4 hours" |
| "Improve quality" | "Review every deliverable with checklist X" |
| "Communicate better" | "Start each meeting with 2-minute status" |

### Step 4: Count Something

> "Count something. Anybody can count something."
> — From "Better"

You can't improve what you don't measure. But you don't need sophisticated analytics.

**Simple counting:**
- How many did we do?
- How long did it take?
- How many errors?
- How many interactions?

Counting itself creates attention and improvement.

### Step 5: Do It, Then Evaluate

- Did it work?
- What happened?
- Should we continue, adjust, or try something else?

### Step 6: Repeat

One thing this week. Another thing next week. Small changes compound.

After a month: four improvements.
After a quarter: twelve improvements.
After a year: you've transformed, one small change at a time.

---

## Gawande's Suggestions for Becoming a Positive Deviant

From the afterword of "Better":

1. **Ask an unscripted question** - Engage with the people around you beyond the transactional
2. **Don't complain** - Complaining is a substitute for action
3. **Count something** - Track data that matters to you
4. **Write something** - Clarify your thinking by writing
5. **Change** - Be willing to try, fail, and try differently

---

## The Compound Effect

| Improvement | One-time effect | Compounded over a year |
|-------------|-----------------|------------------------|
| 1% better per week | Barely noticeable | 67% improvement |
| One small fix per week | Single improvement | 52 improvements |
| One fewer error per day | One day better | 365 fewer errors |

The math of incremental improvement is powerful. But it requires consistency and patience.

---

## Outputs

| Output | Description |
|--------|-------------|
| This week's one thing | Specific, actionable improvement |
| Measurement | What to count |
| Evaluation criteria | How to know if it worked |
| Next iteration | What to try if it doesn't work |

---



**Format:**
```markdown
## Analysis: [Topic]

### Key Findings
- [Finding 1]
- [Finding 2]
- [Finding 3]

### Recommendations
1. [Action 1]
2. [Action 2]
3. [Action 3]
```


## Additional Notes

**Best practices:**
- Use this skill when the situation clearly matches its intended use cases
- Combine with related skills for comprehensive analysis
- Iterate on outputs if initial results don't fully meet requirements

**Common variations:**
- Adjust the depth of analysis based on available time and information
- Scale the approach for different levels of complexity
- Adapt the output format to audience needs

**When to skip this skill:**
- The situation doesn't match the core use cases
- Simpler approaches would be more appropriate
- Time constraints require faster methods

## Example

**User:** "Our customer support is slow and we have no budget for more staff. Management keeps asking for improvement but won't give us resources."

**Incremental Improvement response:**

You can improve without more staff. Here's how.

**Step 1: Accept constraints**
No new hires. That's the canvas. What can you do with what you have?

**Step 2: Find the one thing**

What's one thing that, if improved, would make customers feel support is faster?

Options:
- First response time (even if full resolution takes longer)
- Self-service options (reduce tickets entirely)
- Prioritization (focus speed on high-value cases)
- Templates (faster responses to common issues)

Pick one. Let's say: **First response time.**

**Step 3: Make it specific**

Currently: First response averages 8 hours
This week's goal: First response within 2 hours for all tickets

**Step 4: Count something**

Track: First response time for every ticket this week.

Just tracking it will create awareness and improvement.

**Step 5: How to achieve it**

Without more staff:
- Set specific "ticket check" times (9am, 12pm, 3pm, 5pm)
- Create quick acknowledgment template: "We received your request and are looking into it. You'll hear back from us within [time]."
- Make first response everyone's responsibility in slower periods

**Step 6: Evaluate after one week**

- Did first response time improve?
- Did customer satisfaction change?
- What got in the way?

**Next week:** Pick the next one thing. Maybe it's templates for the top 5 issues. Maybe it's better triage.

**The principle:**
You can't hire more staff. But you can be 10% faster this week, and another 10% faster next month, and in six months you've transformed the function without a single new hire.

Better is possible. It doesn't take more resources. It takes diligence, ingenuity, and willingness to try.

---

## Integration

This skill pairs with:
- **checklist-design** - Build improvements into standard processes
- **positive-deviance-analysis** - Find improvement ideas from high performers
- **failure-analysis-systems** - Identify what to improve

---

## Constraints

- Incremental improvement isn't always enough (some problems need structural change)
- Requires patience—results compound over time
- Leadership must support experimentation
- Must actually measure to know if improvements work

---

## Source Expert

Atul Gawande - `experts/atul-gawande/`