Shadowing in Service Design: Method, Process & Practical Example

Shadowing is an ethnographic observation method where a researcher accompanies a person — customer or employee — through their real service day over an extended period, without intervening in the process [1]. Instead of reconstructing what someone experienced in a retrospective interview, you are there in real time: you see the wait times nobody mentions. You see the workarounds nobody has documented. You see the moment the customer gives up — not because they tell you, but because you are standing right there when it happens.

The method originates from ethnographic social research and was substantively codified as a standalone research instrument by Barbara Czarniawska in “Shadowing: And Other Techniques for Doing Fieldwork in Modern Societies” (2007) [2]. In service design, agencies such as IDEO and Engine Service Design adapted the technique in the early 2000s as part of their human-centered design toolkit [3]. Today, shadowing is part of the standard repertoire of user research in service design.

What distinguishes shadowing from other observation methods: you follow a person, not a process. A Gemba Walk observes the workplace and its processes from a management perspective. A contextual inquiry observes the use of a specific system or product. Shadowing accompanies a concrete individual through their entire service day — across departmental boundaries, system transitions, and waiting periods. That is why shadowing uncovers things that process-oriented methods cannot see: the gaps between processes, the moments when nothing happens, and the informal strategies people use to bridge system boundaries.

This article gives you everything you need to use shadowing in a service design project: the methodological background, a complete step-by-step protocol, a practical example from healthcare, the five most common mistakes, and a systematic comparison with related methods.

Where Does Shadowing Come From? The Academic Roots

The technique of “going along” has its roots in the participant observation of cultural anthropology — Bronislaw Malinowski’s fieldwork with the Trobriand Islanders (1922) is considered an early example of immersive long-term observation [4]. The systematic transfer to organizational contexts was achieved by Henry Mintzberg in 1973 with his groundbreaking study “The Nature of Managerial Work,” in which he accompanied five CEOs for one week each, documenting every activity, every conversation, and every context switch [5]. Mintzberg proved that what managers do differs fundamentally from what they say they do.

Barbara Czarniawska formalized the method as “shadowing” in 2007 and distinguished it from related techniques [2]. Her central insight: shadowing reveals the gap between the official narrative and lived practice — and it is precisely this gap that is the most valuable source of insight for service designers.

In service design, Stickdorn et al. (2018) established shadowing as one of the central research methods in This Is Service Design Doing [6]. They emphasize the difference between shadowing and contextual interviews: “Shadowing goes beyond conversation — you also observe the moments no one talks about because they are taken for granted” [6].

Segelstrom (2013) examined in his dissertation how service designers communicate and visualize insights from qualitative research — including shadowing [7]. His finding: the richest insights from shadowing sessions emerge not during the planned interaction moments but in the transitions and wait times between them — precisely the moments that remain invisible in interviews and surveys.

When Is Shadowing the Right Tool?

Shadowing is most valuable when you want to understand the lived reality of a service experience — not the documented version, not the remembered version, but what actually happens while it happens. The method is particularly suitable when the discrepancy between “how it should be” and “how it is” is unknown or underestimated.

Use shadowing when:

• You want to understand how a customer experiences a complex, multi-stage service — e.g., a hospital stay, an insurance claims process, or onboarding to a new financial product
• You want to investigate the transitions between touchpoints — the moments when the customer switches from one channel to the next, is handed from one department to another, or simply waits
• You want to uncover tacit knowledge — workarounds, informal strategies, unwritten rules that no process manual describes
• You need a realistic basis for personas or journey maps that goes beyond interview data
• You want to capture the emotional experience in the service moment — frustration, uncertainty, relief, confusion — at the moment it occurs

Use a different tool when:

Situation	Better alternative	Why
You want to understand the workplace and its processes from a management perspective	Gemba Walk	Gemba Walk observes the place; shadowing follows a person
You want to understand how someone uses a specific system or product	Contextual Inquiry	Contextual Inquiry focuses on human-system interaction with think-aloud
You want to understand many users simultaneously, not one individual in depth	Survey or diary study	Shadowing does not scale — 1 researcher per person
You want to capture recalled experiences over a long period	Diary study	Diary studies cover weeks or months; shadowing typically hours to one day
You want to benchmark service quality from the customer perspective	Mystery Shopping	Mystery Shopping measures standards; shadowing understands experiences

Step by Step: How to Conduct Shadowing

A shadowing project has six phases. The total effort per participant is typically 4-8 hours of observation plus 2-4 hours of preparation and debriefing.

Step 1: Prepare — Define Research Question and Scope

Formulate the research question: Shadowing without a question is a walk. Define what you want to understand — but keep the question open enough to allow for surprises. Example: “How does a patient experience the admission process from entering the hospital to reaching their room?” is better than “How long does a patient wait at admission?”

Define the scope: Determine the start and end point of the observation. In the service context, this means: at which touchpoint does the shadowing begin? At which does it end? Plan enough time — most shadowing sessions are too short because teams underestimate the wait times and transitions.

Create an observation framework: Prepare a structured note-taking system covering the following dimensions:

• Actions: What is the person doing? (Activities, movements, interactions)
• Context: Where is the person? What is happening around them?
• Emotions: What mood is visible? (Body language, facial expressions, tone of voice)
• Artifacts: What objects, documents, or systems is the person using?
• Time: When does what happen? How long do wait times last?
• Anomalies: What surprises you? What deviates from the expected flow?

Ethics and consent: Inform all participants about the purpose, duration, and type of observation. Obtain written consent when documenting observations that can be traced back to individuals. For employee shadowing in German companies, early engagement with the works council (Betriebsrat) is advisable (see GDPR notes in the “Common Mistakes” section).

Step 2: Select Participants

Customer shadowing: Select participants who use the service in a real context — no special conditions, no VIP treatment. Ideally, accompany customers who are using the service for the first time (onboarding perspective) AND customers who use it routinely (everyday perspective). The combination of both perspectives reveals both the learning curve and habituation effects.

Employee shadowing: Select employees from different roles and experience levels. An experienced claims adjuster will show you the optimized workarounds; a new employee will show you the points where the system does not work without tacit knowledge.

Sample size: The Nielsen Norman Group recommends 5-8 participants for qualitative observation studies [8]. In shadowing practice, 3-5 sessions often suffice because the observation depth per session is higher than with other methods. You recognize saturation when new sessions no longer reveal new patterns.

Step 3: Observe — The Shadowing Day

Clarify the role: Explain to the observed person at the beginning: “I am accompanying you today to understand how your day unfolds. You do not need to do anything differently than usual. I will take notes but will not intervene. If you have questions, we can talk at the end.”

Observe, do not intervene. The golden rule: you are a shadow, not a consultant. You do not step in when something goes wrong. You do not offer help. You do not evaluate. You observe and document. The only exception: if the observed person addresses you directly and expects a response.

Document in real time: Record observations immediately — not from memory. Brief keywords suffice as long as they contain enough context to reconstruct the situation later. Timestamp every observation. Use the observation framework from Step 1 as a guide, but do not let it constrain you — the most important insights are often those that fit no schema.

Pay particular attention to:

• Wait times: What does the person do while waiting? How do they react?
• Transitions: What happens between two touchpoints? Who explains the next step?
• Information gaps: At which points does the person search for information? Whom do they ask?
• Emotional turning points: When does the mood change? What triggers it?
• Workarounds: Where does the person bypass the intended flow?

Step 4: Debrief — The Post-Observation Conversation

After the observation (not during it), conduct a short debrief conversation with the observed person. This conversation serves to validate observations and understand contexts that remained hidden to you as an outsider.

Three question types for the debrief:

1. Confirmation questions: “I noticed that you waited ten minutes at station 3. Does that happen regularly?”
2. Deepening questions: “You printed the documents from the system even though they were available digitally. What was the reason?”
3. Reflection questions: “If you were to describe today’s process as a whole — what was the most difficult moment?”

Important: The debrief is not a retrospective correction of your observations. If the person says “That was atypical today” and your observation reveals a systemic pattern, the observation is the stronger evidence. People tend to rationalize their daily routines in hindsight — that is precisely what makes shadowing more valuable than interviews alone [2].

Step 5: Analyze — Synthesize Patterns and Insights

Transcription and preparation: Transfer your field notes into a structured format within 24 hours. The memory of contextual details fades rapidly after that. If you have observed multiple people, create one shadowing protocol per person.

Pattern analysis: Search for recurring patterns across all shadowing sessions:

• Systemic wait times: Points where all participants had to wait
• Shared workarounds: Informal strategies that multiple people developed independently
• Information breaks: Points where all participants lacked relevant information
• Emotional patterns: Moments that triggered similar reactions in all participants

Link to the service design project: Transfer the shadowing insights into existing project artifacts — Customer Journey Map, Service Blueprint, personas. Shadowing data is particularly valuable for filling the backstage layer of a blueprint with reality.

Step 6: Synthesize — From Observations to Design Impulses

Identify design opportunities: Every systemic observation is a potential design impulse. Formulate insights as “How might we” questions: “How might we use the wait time between admission and room assignment, rather than merely shortening it?”

Create an evidence catalog: For each design impulse, document the associated shadowing observations — date, participant, situation, quote. This evidence catalog is your strongest argument in stakeholder discussions because it is based not on opinions but on documented observations.

Comparison: Shadowing vs. Interview vs. Diary Study vs. Gemba Walk

Dimension	Shadowing	Interview	Diary Study	Gemba Walk
Focus	One person through their service day	Memories and evaluations	Self-documentation over time	Processes at the workplace
Perspective	Observer perspective in real time	Participant perspective from memory	Participant perspective in real time	Management perspective on-site
Data type	Behavior, context, emotions, artifacts	Statements, evaluations, narratives	Diary entries, photos, reflections	Process observations, questions
Time investment	4-8 hours per participant	30-90 minutes per participant	1-4 weeks duration	30-60 minutes per walk
Strength	Reveals the gap between saying and doing	Efficient, scalable, accessible	Captures longitudinal patterns	Quick, leadership-friendly, regular
Weakness	High time investment, Hawthorne effect	Recall bias, social desirability	Self-selection bias, incomplete	Focus on process, not on person
Best combination	Shadowing + Interview (debrief)	Interview + Shadowing (validation)	Diary + Shadowing (deep phase)	Gemba Walk + Shadowing (individual)

Decision guide: If you want to know what people say, conduct interviews. If you want to know what people document over time, use diary studies. If you want to know what people actually do, use shadowing. If you want to understand the process environment, use a Gemba Walk. The strongest combination: shadowing for depth, interviews for breadth, diary studies for longitudinal perspective.

Practical Example: Shadowing a Patient Through the Hospital Admission Process

Starting Situation

A hospital network in the DACH region wants to improve its elective admission process — the journey from the patient’s arrival to reaching their room. Internal dashboards show an average admission duration of 2 hours and 15 minutes. The patient satisfaction survey yields a mediocre score. The quality management team has documented the process and finds no obvious bottlenecks. The Chief Patient Officer commissions a shadowing project.

Execution

Three researchers each accompany two patients through the entire admission process (6 shadowing sessions, 2-3 hours each). The patients were informed in advance and gave consent. The researchers carry a clipboard, no camera — physical unobtrusiveness is critical.

Observations (Excerpt from One Session)

Patient Mrs. M., 62, elective knee surgery:

Time	Station	Observation	Type
07:30	Main entrance	Mrs. M. spends 4 minutes looking for the correct entrance. Signage points to the main entrance, but admissions is at the side entrance.	Information break
07:38	Admissions desk	Wait time 12 minutes. Mrs. M. fills out a form she already completed online — the system did not transfer the online data.	System break / Workaround
07:55	Admissions counter	Clerk explains the day’s schedule verbally. Mrs. M. asks three times about the sequence — no written schedule provided.	Information break
08:15	Waiting area	25-minute wait. Mrs. M. does not know what she is waiting for. She asks a nurse, who says: “You need to wait for the anesthesiologist.”	Wait time without information
08:40	Anesthesia consultation	Conversation lasts 10 minutes. Well explained, Mrs. M. appears relieved.	Positive moment
08:55	Waiting area	Another 20-minute wait. Mrs. M. stands up, sits down, goes to the vending machine, comes back. Body language: increasingly tense.	Wait time with rising frustration
09:20	Room assignment	An orderly takes Mrs. M. to her room. No orientation conversation — “You can go ahead and settle in.”	Missing handover moment

Insights Across All 6 Sessions

Pattern 1: The “black hole” effect. In 5 of 6 sessions, there were at least two waiting phases during which patients did not know what they were waiting for, how long it would take, or whether they could leave the waiting area. These wait times generated disproportionate frustration — not because they were long, but because they were information-empty.

Pattern 2: Duplicate data entry. 4 of 6 patients had to re-enter information they had already submitted online. The IT system did not transfer the data, and the clerks knew it — they had developed a workaround (printing the online submission + manually re-typing).

Pattern 3: The missing first impression. In none of the 6 sessions was there a defined “welcome moment” — a point at which someone said: “Welcome, here is your schedule, I am your contact person.” The patients navigated themselves through the process.

Implementation

Based on the shadowing insights, three measures were derived: (1) A one-page schedule handed to every patient at admission — with time estimates and a contact person. (2) System integration between the online pre-registration and the admission system as an IT project. (3) A “welcome protocol” as a defined first contact with a named nurse.

Note: This example is illustratively constructed to demonstrate the method in a service context. The observations are based on typical industry patterns in healthcare.

5 Common Mistakes in Shadowing

1. Observing for too short a period

What goes wrong: The shadowing session is limited to 1-2 hours “because we do not want to burden the patient for too long.” The result: you see the optimized core process but miss the transitions, wait times, and peripheral phenomena that make up the bulk of the experience.

Why it hurts: McDonald (2005) demonstrates in her analysis of shadowing as a research method that the most relevant observations often appear only after an “acclimatization phase” of 30-60 minutes, once both observer and observed have adapted to the situation [9]. Short observations capture only the beginning — the phase when the Hawthorne effect is strongest.

Solution: Plan the shadowing session for the entire duration of the service experience — from entry to departure. If that takes 4 hours, it takes 4 hours.

2. Ignoring the Hawthorne effect

What goes wrong: The observed person behaves differently because they know they are being observed. Employees work more carefully. Customers complain less. You see the showcase version.

Why it hurts: The Hawthorne effect is particularly pronounced in shadowing because the observation occurs 1:1 — there is no “group protection” as in team observations. Czarniawska (2007) recommends not trying to eliminate the effect but rather documenting it and incorporating it into the analysis [2].

Solution: (1) Begin with a warm-up phase (15-30 minutes) during which you deliberately take few notes to normalize the situation. (2) Observe the same person on two different days if possible — the effect is typically weaker on the second day. (3) Triangulate: compare shadowing observations with data from other sources (process logs, complaint statistics).

3. Observer bias — seeing what you expect

What goes wrong: You enter the observation with a hypothesis and see only what confirms it. If you believe wait times are the problem, you meticulously note every wait time — but miss the information breaks that are actually the larger cause.

Why it hurts: Confirmation bias is the most common source of distortion in qualitative research [10]. In shadowing, it is particularly insidious because the abundance of impressions favors selective perception.

Solution: (1) Record EVERYTHING, not just what seems relevant — selection happens during analysis, not during observation. (2) Observe in pairs and compare protocols. Differences between observers reveal bias points. (3) Use the structured observation framework from Step 1 as a corrective.

4. Intervening instead of observing

What goes wrong: The researcher helps the customer when they get lost. Or explains a function to the employee they do not know. Every intervention changes the observed flow and renders the data unusable.

Why it hurts: Intervening destroys exactly the situation you want to observe. If you show the patient the way, you do not know how they would have navigated without help — and that is precisely the relevant information.

Solution: Agree on a clear intervention rule before the session: you intervene only if there is immediate danger. Everything else is observed and documented. When the urge to help is strong, note the situation as “intervention needed” instead — that is a particularly strong design insight.

5. Not obtaining consent

What goes wrong: The team begins shadowing without fully informing the observed persons or without obtaining the required consents. In German companies, this can cause problems with GDPR (DSGVO) and the works council (Betriebsrat).

Why it hurts: Without informed consent, the collected data cannot be used — and participants’ trust is permanently damaged. With employee shadowing in particular, the works council can retroactively intervene if it was not engaged early.

Solution: Create a consent form explaining purpose, duration, type of documentation, data storage, and right of withdrawal. Inform the works council before the first employee shadowing session. Anonymize all observations during analysis.

Variants of Shadowing

Customer Shadowing

The researcher accompanies a customer through their entire service experience — from first contact to completion. Focus: experience, emotions, information gaps, waiting moments. Particularly valuable for customer journey mapping and the Identify phase in service design.

Employee Shadowing

The researcher accompanies an employee through their workday. Focus: workarounds, system boundaries, information needs, handovers to colleagues. Particularly valuable for the backstage layer of a Service Blueprint.

Cross-Functional Shadowing

The researcher accompanies the same case from the perspective of different participants — e.g., first the customer, then the frontstage employee, then the backstage employee. This variant reveals how different actors experience the same service moment differently.

Digital Shadowing

For digital services: the researcher observes via screen sharing how a user navigates the service. What is lost in digital shadowing: body language, physical context, ambient sounds. What works: 1:1 screen sharing with narration and webcam.

Frequently Asked Questions

What is the difference between shadowing and an interview?

In an interview, someone tells you what they experienced — from memory, with all the distortions that memory brings. In shadowing, you are present while it happens. You see the behavior, not the narrative. The strength of the interview is efficiency and scalability. The strength of shadowing is authenticity: you observe the gap between what people say and what they do [2].

How many shadowing sessions do I need?

3-5 sessions typically suffice for a service design project because the observation depth per session is high [8]. You recognize saturation when new sessions no longer reveal new patterns. Plan 4-8 hours of observation plus 2-4 hours of preparation and debriefing per session.

Can I conduct shadowing remotely?

Yes, for digital services. Via screen sharing, you observe how a user navigates the service. What is lost: body language, physical context, spontaneous peripheral observations. What works: 1:1 screen sharing with webcam and narration. For physical services, remote shadowing is not a substitute.

How do I handle data privacy in shadowing?

In German companies, GDPR (DSGVO) requirements apply to all personal observation data. Specifically: informed consent of all participants, anonymization during analysis, purpose limitation of data. For employee shadowing, we recommend early engagement with the works council (Betriebsrat), even if the observation does not fall under formal co-determination obligations.

How does shadowing differ from a Gemba Walk?

The Gemba Walk observes a place and its processes — the leader goes to the place where value is created to understand work processes. Shadowing accompanies a person — the researcher follows an individual through their service day, across places and processes. The Gemba Walk is process-centric; shadowing is person-centric.

Related Methods

A typical sequence in service development: With shadowing, you observe how a customer or employee actually experiences the service. The insights flow into a Customer Journey Map and a Service Blueprint. With a Gemba Walk, you validate the process observations on-site. The overarching method selection guide is in the Service Design Methods Overview.

• User Research in Service Design: Shadowing as part of the user research repertoire — placing it in the broader context of qualitative and quantitative methods
• Gemba Walk: When you want to observe not a person but a workplace and its processes
• Service Blueprint: Shadowing insights provide the raw material for the backstage layer of the blueprint
• Service Design: The overarching discipline in which shadowing is embedded as a research method

---

Research Methodology

This article synthesizes findings from Czarniawska’s foundational work on shadowing as a research method (2007), Mintzberg’s early application of structured observation (1973), Stickdorn et al.’s framing in the service design context (2018), Segelstrom’s research on insight communication (2013), McDonald’s methodological analysis (2005), and the practitioner literature from IDEO and the Nielsen Norman Group. The practical example (hospital admission) is illustratively constructed based on typical industry process patterns.

Limitations: Academic literature on shadowing in service design is thin — most studies treat shadowing as part of a larger research program, not as an isolated method. Empirical comparative studies on the effectiveness of shadowing versus other observation methods in service contexts are nearly nonexistent.

Disclosure

SI Labs offers consulting services in the area of service innovation. In the analysis phase of the Integrated Service Development Process (iSEP), we use shadowing as one of several research methods to observe actual service delivery at our clients’ organizations. This practical experience informs the framing of the method in this article. Readers should be aware of potential perspective bias.

References

[1] Stickdorn, Marc, Markus Edgar Hormess, Adam Lawrence, and Jakob Schneider. This Is Service Design Doing: Applying Service Design Thinking in the Real World. Sebastopol: O’Reilly Media, 2018. [Practitioner Handbook | Shadowing method description | Citations: 1,500+ | Quality: 88/100]

[2] Czarniawska, Barbara. Shadowing: And Other Techniques for Doing Fieldwork in Modern Societies. Malmo: Liber / Copenhagen Business School Press, 2007. [Foundational work | Shadowing methodology | Citations: 1,200+ | Quality: 92/100]

[3] IDEO.org. The Field Guide to Human-Centered Design. 2015. URL: https://www.designkit.org/resources/1.html [Practitioner Toolkit | 57 design methods including Shadowing | Quality: 85/100]

[4] Malinowski, Bronislaw. Argonauts of the Western Pacific. London: Routledge & Kegan Paul, 1922. [Foundational work | Participant observation | Citations: 10,000+ | Quality: 90/100]

[5] Mintzberg, Henry. The Nature of Managerial Work. New York: Harper & Row, 1973. [Foundational work | Structured observation of managers | Citations: 8,000+ | Quality: 90/100]

[6] Stickdorn, Marc et al. “Shadowing.” This Is Service Design Doing — Method Library. URL: https://www.thisisservicedesigndoing.com/methods/shadowing [Practitioner Method Card | Step-by-step protocol | Quality: 85/100]

[7] Segelstrom, Fabian. Stakeholder Engagement for Service Design: How Service Designers Identify and Communicate Insights. PhD thesis, Linkoping University, 2013. [PhD Thesis | Insight communication from qualitative research | Citations: 100+ | Quality: 80/100]

[8] Nielsen Norman Group. “How Many Participants for UX Research?” Accessed February 25, 2026. URL: https://www.nngroup.com/articles/how-many-test-users/ [Practitioner Article | Sample size recommendations | Quality: 85/100]

[9] McDonald, Seonaidh. “Studying Actions in Context: A Qualitative Shadowing Method for Organizational Research.” Qualitative Research 5, no. 4 (2005): 455-473. DOI: 10.1177/1468794105056923 [Academic Article | Methodological analysis of Shadowing | Citations: 400+ | Quality: 85/100]

[10] Flick, Uwe. An Introduction to Qualitative Research. 6th edition. London: SAGE, 2018. [Textbook | Qualitative research methodology including observer biases | Citations: 5,000+ | Quality: 88/100]