Cutting Pick Time by 23% in a 3PL Warehouse

From baseline bottlenecks to pilot results

Breno Alves Beirigo, PhD

Assistant Professor of Stochastic Operations Research

Behavioral, Management and Social Sciences (BMS) Industrial Engineering and Management Science (BMS-HBE-IEMS) University of Twente Enschede, Netherlands

Date

2026-08-14

A warehouse improvement case and a showcase of the shared slide system.

Introduction

Part I

Setting the frame and objectives

01

Section divider: announce the story arc and reset attention. Keep titles short and bold to anchor the audience.

Today’s Agenda

Today’s Flow

1. West Hub baseline performance
   
2. Time-motion findings
   
3. Intervention options
   
4. Pilot design and outcomes
   
5. Financial impact and risks
   
6. Recommendation and rollout

Decision Needed

Approve an 8-week rollout across Zones A-C

• Budget: $180k (software + training)
  
• Expected payback: < 6 months

Key assumptions

• 2 shifts, 180 pickers, 120k lines/day
  
• Peak season +18% volume in Q4
  
• Target: -20% travel time, +1.5% accuracy

Use the agenda as a cognitive map. The callout states the decision.

Learning Objectives

By the end of this presentation, you will be able to:

1. Diagnose the main drivers of pick time
2. Quantify the impact of slotting and batching changes
3. Interpret pilot KPIs (throughput, accuracy, rework)
4. Decide the rollout plan and governance model

Keep objectives measurable and limited to 3-5 items.

The Core Constraint

Travel time drives 62% of total pick cycle time.

Use a minimal slide to shift attention to the speaker for a key point.

The Hook: Why This Matters

What if we could ship 14,000 more orders per week without adding headcount?

Hook with a single, specific outcome. Keep the slide clean.

About the Speaker

FulfillCo Logistics - West Hub

Breno A. Beirigo

Director of Operations Analytics
FulfillCo Logistics

What I do

• Optimize slotting, batching, and labor planning
• Build decision-support tools for ops leaders
• Lead pilots across 8 fulfillment centers

Why listen

• 12 years in warehouse operations + analytics
• Led 5 network-wide efficiency rollouts
• Delivered $9.4M annualized savings in 2025

Contact

breno@fulfillco.com

Keep bio factual and short. Anchor credibility with outcomes.

Stakeholder Goals

Operations

• Reduce travel and congestion
• Stabilize pick accuracy
• Improve labor planning
• Reduce training time

Finance and Customer Experience

• Lower cost per order
• Increase on-time shipment rate
• Reduce rework and returns
• Improve peak-season readiness

Use a simple two-column split to show stakeholder alignment.

Branding & Identity

FulfillCo Optimize 2026

Program goals

• Standardize picking across zones
• Deliver 20% faster cycle times
• Reduce training ramp from 4 weeks to 2
• Create a repeatable playbook for new sites

Branding slide works best early, after the hook.

Core Slide Types

Part II

Layouts, purpose, and high-impact patterns

02

Section divider: shift from context to structure.

Baseline Performance - West Hub

West Hub ships ~24k orders/day across three zones. Current performance shows:

• 6.8 min average pick cycle time (goal: 5.2)
  
• 1.9% mis-pick rate (target: < 0.8%)
  
• 7.4 km walked per picker per shift

Constraint

62% of cycle time is travel + search, not picking.

Use a short title and 3-4 bullets maximum for clarity.

Root Causes We Can Fix

See Figure 1 for the typical pick flow diagram.

Root causes observed

• Replenishments block main aisles at peak
• Fast movers scattered across zones
• Batch sizes inconsistent by shift

These issues create congestion and extra walking (see Figure 1)

Figure 1: Typical pick flow in West Hub before pilot.

Two-column layout pairs explanation with a diagram for quick alignment.

The Three Levers (Slotting, Batching, Pathing)

Slotting

Move fast movers closer to pack

Top movers: A/B/C runners

Batching

Group orders by aisle affinity

Wave size: 12-16 orders

Pathing

Dynamic routes reduce cross-aisle trips

Rule: one-way aisles during peak

Use three columns for mutually exclusive options.

Pilot Result - Cycle Time

23%

Reduction in average pick cycle time after pilot

Measured across Zones A-C over 6 weeks

Use a single metric and one short context line.

Executive Quote

Small changes in travel paths create big gains in throughput.

• Priya Nair, VP Operations, FulfillCo

This is why we focused on travel distance before adding labor.

Keep quotes under ~30 words and add attribution.

The Hidden Tax

Every extra step is a tax on throughput, accuracy, and morale.

Design principles for this slide:

• One idea: the slide is the sentence; everything else is spoken.
• No visible title: the H2 exists for structure/navigation, but CSS hides it on this slide.
• Controlled line length: keep the sentence to ~10-16 words and avoid punctuation-heavy phrasing.
• Emphasis is selective: bold/italic only 1-3 words so the eye lands where you want.
• Pause: show the slide, stop talking for 2-3 seconds, then elaborate with one concrete example.

Pilot Delivery Plan (8 Weeks)

See Figure 2 for the pilot delivery roadmap.

Figure 2: Pilot delivery roadmap

Timeline works well for fixed date sequences.

Prioritization - Effort vs Impact

See Figure 3 for the effort vs impact decision matrix.

Figure 3: Effort vs impact decision matrix

Use a 2x2 only when you can define both axes clearly.

Visual Content

Part III

Figures, tables, and diagrams that teach

03

Section divider: shift from narrative to evidence.

Baseline vs Pilot KPIs

Table 1: Baseline, target, and pilot KPIs

Metric	Baseline	Target	Pilot
Orders per hour	420	520	515
Pick cycle time (min)	6.8	5.2	5.2
Mis-pick rate	1.9%	0.8%	0.7%
Travel per shift (km)	7.4	5.5	5.6

Use tables when the audience needs exact values.

Throughput Trend - Orders per Hour

Readout

Throughput increased steadily after slotting changes in Week 3.

Keep the interpretation close to the chart.

Where People Walk - Travel Heatmap

Hot spots cluster near Aisle 7 and the cross-dock lane.

Use a single callout to highlight the key insight.

Before/After - Travel Pattern

BEFORE: Unoptimized

AFTER: Slotted + Batched

Side-by-side layouts emphasize contrast quickly.

Visual Evidence: Context

Use a 3-up gallery for quick visual context.

Code & Technical Content

Part IV

Readable code, reproducible results

04

Introduce technical depth with a clear transition.

Inputs (Orders + Layout)

# Orders, distances, and aisle metadata
orders = load_orders("west_hub_orders.csv")
layout = load_layout("west_hub_layout.json")
travel_time = build_time_matrix(layout)

Keep code short; use only the essential lines.

Heuristic (Nearest Neighbor + 2-Opt)

# Build initial tour
current = 0
route = [current]
visited = {current}

while len(visited) < len(orders):
    next_stop = min(
        (i for i in range(len(orders)) if i not in visited),
        key=lambda i: travel_time[current, i]
    )
    route.append(next_stop)
    visited.add(next_stop)
    current = next_stop

# Improve with 2-opt
route = two_opt(route, travel_time)

Use a small code window to avoid cognitive overload.

Code Animation: Line Highlight Steps

When to use

• Walk through a function without reading it all at once
• Keep attention on the “now” line(s)
• Pair each click with one sentence of explanation {.fragment}

Tip

Syntax: add code-line-numbers="|1|3-6|8" to a code block.

def two_opt(route, dist):
    improved = True
    while improved:
        improved = False
        for i in range(1, len(route) - 2):
            for j in range(i + 2, len(route) - 1):
                delta = (
                    dist[route[i-1], route[j]]
                    + dist[route[i], route[j+1]]
                    - dist[route[i-1], route[i]]
                    - dist[route[j], route[j+1]]
                )
                if delta < -1e-10:
                    route[i:j+1] = reversed(route[i:j+1])
                    improved = True
                    break
            if improved:
                break
    return route

Use this slide type when you want to teach code, not just show it. Advance one step per sentence: signature → loop → delta → swap → early exit. If someone asks “why this heuristic?”, bridge to the KPI slide (cycle time impact).

Code Animation: Auto-Animate (Build the Solution)

# Step 1: Build an initial route (fast, not optimal)
current = 0
route = [current]
visited = {current}

while len(visited) < len(orders):
    next_stop = min(
        (i for i in range(len(orders)) if i not in visited),
        key=lambda i: travel_time[current, i]
    )
    route.append(next_stop)
    visited.add(next_stop)
    current = next_stop

Auto-animate is best for “before/after” code where you want the audience to see the change. Prompt: “Now we’ll keep the same skeleton, but add one improvement pass.”

Code Animation: Auto-Animate (Build the Solution)

# Step 1: Build an initial route (fast, not optimal)
current = 0
route = [current]
visited = {current}

while len(visited) < len(orders):
    next_stop = min(
        (i for i in range(len(orders)) if i not in visited),
        key=lambda i: travel_time[current, i]
    )
    route.append(next_stop)
    visited.add(next_stop)
    current = next_stop

# Step 2: Improve it (small effort, big travel-time gain)
route = two_opt(route, travel_time)

Call out the “shape” stays the same; we just add a refinement step. Explain why this is safe for ops: predictable runtime + measurable gain.

Highlighted Code (Advanced)

def two_opt(route, dist):
    improved = True
    best = route
    while improved:
        improved = False
        for i in range(1, len(best) - 2):
            for j in range(i + 2, len(best) - 1):
                delta = (
                    dist[best[i-1], best[j]] + dist[best[i], best[j+1]] -
                    dist[best[i-1], best[i]] - dist[best[j], best[j+1]]
                )
                if delta < -1e-10:
                    best[i:j+1] = reversed(best[i:j+1])
                    improved = True
                    break
            if improved:
                break
    return best

Use folding for advanced details not needed in the live flow.

Pilot Run Output (Terminal)

$ python run_pilot.py --zone A --weeks 6
Loaded 4,812 orders
Baseline avg cycle time: 6.8 min
Pilot avg cycle time: 5.2 min
Accuracy: 99.3%

Live Demo

Run a 50-order batch, show route overlay, then export KPIs.

Use terminal slides for reproducibility and trust.

Cycle Time Model (Simple)

The pick cycle time can be expressed as:

\[ T = T_{travel} + T_{search} + T_{pick} + T_{scan} \]

Where \(T_{travel}\) dominates during peak congestion.

Math slides should be short and readable.

Teaching Slides

Part V

Active learning and guided practice

05

Teaching section: break down complex ideas into steps.

Quick Check (Decision Question)

Question: Which action reduces travel time the most?

A. Add two scanners per aisle
B. Re-slot top 300 SKUs to fast aisles
C. Add one extra picker per shift
D. Increase pack station count

Use quizzes to check comprehension before moving on.

Exercise: Slotting Plan

Activity (10 minutes)

In teams of 3, map your zone’s top 20 SKUs to A/B/C slots.

Deliverable

• Top 20 SKU list
• Proposed A/B/C placement
• One risk to address (space, replenishment, safety)

Timebox

• 6 minutes work
• 4 minutes share-out

Exercises work best with a clear deliverable and timebox.

Callouts in Action (Risk, Lever, Critical)

Risk: Aisle Congestion

Replenishments overlap with pick waves between 9-11 AM.

Best Lever: Re-slot Top 300 SKUs

Moves 41% of volume into 2 fast aisles.

Critical: Scanner Latency

2-3s delays add ~18 minutes per shift.

Use callouts for emphasis, not for long paragraphs.

Common Mistakes

Mistake: Mixing A/B/C in one aisle

Creates cross-traffic and breaks batching logic.

Mistake: Untrained temps in fast aisles

Quality drops before cycle time improves.

Mistake: Ignoring replenishment windows

Re-slotting can backfire without timing controls.

Tight callouts work well for lists of pitfalls.

Workflow Slides

Part VI

Breaks, Q&A, and audience flow

06

Workflow slides manage pace and audience energy.

Parking Lot (Capture Questions)

Parking Lot

Questions we will return to later:

• Should we standardize tote sizes before re-slotting?
• What change-management approvals are required?
• How will we measure pick accuracy during the pilot?

Use a parking lot for important but off-track questions.

Switch to Demo

Live demo: running the heuristic and visualizing the tour

Plan

• Run: python run_pilot.py --zone A
• Show: output, runtime, and KPI export

Signal mode changes explicitly to keep the audience oriented.

Coffee Break

15:00

Please return by 10:45 AM

Break slides should be simple and easy to read at a distance.

Questions?

Email: breno@fulfillco.com

Ops Wiki: ops.fulfillco.com/west-hub

End Q&A slides with clear follow-up channels.

Conclusion

Part VII

Synthesis and next steps

07

Use the conclusion to restate the decision and next steps.

Summary of Findings

1. Travel time is the dominant driver of pick cycle time
2. Slotting + batching reduced cycle time by 23% in pilot
3. Accuracy improved to 99.3% with revised training
4. ROI < 6 months on a full rollout

Keep summary to 3-5 items.

Recommendation

Approve Rollout

Proceed with an 8-week rollout across Zones A-C, starting March 1.

Owner: Ops Excellence
Budget: $180k
Success metrics: -20% cycle time, >99% accuracy

Place the recommendation on a single slide to avoid ambiguity.

References

Quarto generates this from bibliography: references.bib.

Thank You!

Breno A. Beirigo

breno@fulfillco.com

Slides created with Quarto and reveal.js

Keep the closing slide simple and visual.

Media & Engagement

Part VIII

Video, imagery, and social proof

08

Media section: show evidence of impact and adoption with minimal text.

Social Proof (Operations Quote + Metrics)

“The new slotting rules cut training time in half. New hires hit target pace by day 10.”

• Maria Lopez, West Hub Operations Lead

Pilot KPIs

Training ramp: 4 weeks -> 2 weeks : 98.1% -> 99.3%

Signals of adoption

• 82% of pickers prefer new route guidance
  
• 3 supervisors requested the playbook for other sites
  
• Replenishments rescheduled to avoid peak congestion

Pair a short quote with a small metric panel for credibility.

Social Media Integration

Why include social proof?

• Demonstrates adoption
  
• Shows stakeholder buy-in
  
• Signals momentum
  
• Reinforces credibility

Quote

“Pilot results changed how we schedule labor.”

Sarah Johnson, DC Manager

West Hub - 2w

The slotting changes reduced congestion immediately. Pickers report less backtracking and faster training.

Like 127 - Comment 23 - Share 45

Use social posts sparingly and with permission.

Pilot Walkthrough

Local video is more reliable for offline presentations.

Peak Hour - Zone B

Peak hour footage

Pilot in action

Pick density increased while congestion decreased.

Use muted, dimmed video to keep text legible.

West Hub at Peak

Floor-level still

The floor feels calmer, faster, safer.

Image slides work best with a single short line.

Visual Evidence: Assets

Throughput trend

Throughput

Travel heatmap

Travel heatmap

Warehouse photo

Floor context

Use a gallery to provide visual context without heavy text.

Customer Voice

“Orders ship earlier now, and we see fewer partials at the docks.”

• Regional Carrier Partner

Keep testimonials short and attributable.

Appendix

Appendix

Risks, rollout plan, and operating model

A

Use appendix slides for backup during Q&A.

Risk Register

Table 2: Risk register with mitigations

Risk	Impact	Likelihood	Mitigation
Poor data quality	High	Medium	Data audit + validation checks
Overfitting to benchmarks	Medium	Medium	Holdout instances + ablations
Low adoption	High	Low	Pilot, training, human-in-the-loop

Risk tables are common in executive reviews.

Rollout Phases

See Figure 4 for the implementation plan diagram.

Figure 4: Implementation plan in four phases

Pair with owners, timelines, and KPIs for realism.

Key Idea: Decision Leverage

Key idea

Optimization is not the goal.
Better decisions are the goal.

Optimization creates decision leverage: faster planning, fewer errors, and more consistent outcomes.

Overlay panels work best with short, punchy text.

Method Walkthrough

Part IX

Step-by-step method walkthrough

09

Use the walkthrough for complex methods or algorithms.

The method: Large Neighborhood Search (LNS)

The method comprises 5 steps:

• Step 1: Define the problem
• Step 2: Define a neighborhood move
• Step 3: Evaluate the move efficiently
• Step 4: Acceptance rule
• Step 5: Stopping conditions

Step 1: Define the problem

We solve a routing problem with constraints:

• Each order is picked once
• Travel time is minimized
• Aisle rules (one-way) are respected

Feasible Solution

Meets capacity, time windows, and route length limits.

Start with the problem definition before the algorithm steps.

Step 2: Define a neighborhood move

Common moves:

• 2-opt (reverse a segment)
• relocate (move one node)
• swap (exchange two nodes)

Connect move choice to the warehouse layout.

Step 3: Evaluate the move efficiently

Instead of recomputing the full tour cost, use a delta evaluation:

\[ \Delta = (c_{a,c} + c_{b,d}) - (c_{a,b} + c_{c,d}) \]

Explain which edges are removed and added.

Step 4: Acceptance rule

• Greedy: accept if \(\Delta < 0\)
• Threshold: accept if \(\Delta < \tau\)
• Probabilistic: accept sometimes even if \(\Delta > 0\)

Use this slide to motivate metaheuristics.

Step 5: Stopping conditions

Stop when:

• no improving move exists
• time limit reached
• max iterations reached
• stagnation threshold reached

Practical default

Always enforce a hard time limit for reproducibility and fairness.

Close with a practical rule-of-thumb.

Layout Components

Part X

Reusable layout building blocks

10

Layout components are reusable across decks.

West Hub Snapshot

West Hub

• 3 zones
• 180 pickers
• 24k orders/day

Why this site?

• Highest congestion in the network
• Most variable cycle time
• Best data coverage for pilot

We chose West Hub because changes here generalize to other sites.

Sidebar layouts are good for context blocks.

Rollout Workstreams (2x2)

Baseline

Baseline

• Time study
• Heatmap
• KPI audit

Interventions

Interventions

• Slotting
• Batching
• Pathing

Pilot

Improvement

• Local search
• Delta evaluation
• Stopping rules

Rollout

Deployment

• Pilot
• Monitoring
• Adoption plan

Use grids for scanable summaries.

Operating Model Checklist (3x2)

People

Roles and ownership

Process

Routines and SOPs

Data

Inputs and quality

Model

Optimization logic

Tech

Tooling and uptime

Change

Training and adoption

Keep tile text short and parallel.

Key Idea: Standardization Scales

Key idea

Standardization unlocks scale.

Once West Hub is stable, we can replicate the playbook across 7 sites.

Overlay panels should be brief and high-contrast.

Chart Visualization

Part XI

Common questions → charts that answer them

11

Transition to “how we show the work”: chart choice + clear visual encoding.

Chart Visualization: Common Themes That Carry Meaning

Use the question to pick the chart. Most stakeholder asks fit one of these:

• Change over time → line chart (annotate events; show baseline vs after)
• Compare categories → sorted dot/bar (common scale; label values)
• Show distribution → histogram/box (show variability + outliers)
• Show relationships → scatter + trend (avoid overplotting; show uncertainty)
• Show composition → 100% stacked bars or small multiples (avoid pies)
• Show intensity in a matrix/space → heatmap (use perceptual color scales)

Design Rules of Thumb

• Prefer position/length encodings; use color for emphasis, not precision.
• Direct-label key series/categories; keep legends secondary.
• Remove non-essential ink; keep one clear takeaway per slide.

Design: start with a compact “chart selection” mental model so each following slide feels inevitable. We use a consistent Seaborn theme + a single font (via pyfonts) to reduce visual noise across figures.

Change Over Time: Annotate the Moment the Story Changes

Figure 5: Orders per hour over eight weeks, with the intervention annotated.

Readout

The slope and the step-change matter more than week-to-week noise — annotate the intervention.

Design: line charts are the default for time. We keep a single series, add a subtle baseline/after cue, and annotate the event so the audience doesn’t have to infer “what happened when”. Gridlines are light, and the title states the takeaway.

Compare Categories: Sort + Use a Common Baseline

Figure 6: Average pick cycle time by zone (sorted).

Readout

Sorting makes rank obvious; highlight only the category you want the room to discuss.

Design: comparisons are fastest when categories are sorted and share the same baseline. We use a single accent color to draw attention to the worst-performing zone, and we label values directly to reduce legend-hunting.

Distribution: Show Spread, Not Just the Average

Figure 7: Pick cycle time distribution (baseline vs pilot).

Readout

Medians can improve while the tail stays risky — distributions reveal where the pain remains.

Design: averages hide variability. A two-distribution overlay makes the shift visible while still showing spread and tails. We keep colors consistent (gray baseline, green pilot) and mark medians with dashed lines for quick reference.

Relationship: Use Scatter + Trend to Explain Drivers

Figure 8: Travel distance vs pick time with an overall trend.

Readout

When you argue about drivers, show the raw points and the trend — not just a correlation coefficient.

Design: scatterplots make relationships and outliers tangible. Transparency reduces overplotting, and a single trend line summarizes direction without hiding the variability. Keep units explicit on axes.

Composition: Check Mix Before Comparing Performance

Figure 9: Order mix (100% stacked) for baseline vs pilot.

Readout

If the mix changes, KPI changes may be case-mix, not process improvement.

Design: composition is easiest to compare with 100% stacked bars (or small multiples). We format the y-axis as percentages and keep categories in a consistent palette. Legend moves right to protect data.

Heatmaps: Use Color for Patterns, Not Precise Values

Figure 10: Correlation heatmap across operational metrics (illustrative).

Readout

Heatmaps are for pattern recognition — use them to find questions, then follow up with detailed charts.

Design: a heatmap trades precision for speed. We use a diverging scale for signed values (negative↔︎positive), lock the range to [-1, 1] for honest comparisons, and annotate to support quick scanning. Avoid rainbow maps; prefer perceptually ordered palettes.