SURVEY RESEARCH · QUANTITATIVE + MIXED METHODS
Could a high-touch service move to self-serve — and where couldn't it?
A high-touch service wanted to move work onto consumers, but only if consumers would accept it. This study went to the whole market, not just existing customers, to find where self-service could hold and where it could not.
METHOD
Survey · mixed methods
RESPONDENTS
588
OPEN-ENDED CODED
536
CONTEXT
Relocation services
01
THE PROBLEM
Move consultants were doing a heavy share of the work on behalf of the people they relocated. The business wanted to free them up and improve the direct-to-consumer experience by shifting toward self-service.
But that only works if consumers actually want it. So the real question wasn't internal — it was whether a self-service shift was viable across the market, and whether removing consultant touch-points would help or hurt.
The goal was never to strip consultants out everywhere. It was to find where people genuinely needed a human, versus where reaching out was just a gap the product could close.
02
MY ROLE
Sole researcher. I designed the survey and question hierarchy, defined the recruitment strategy, fielded the study, analyzed the quantitative data, and built the codebook to analyze 536 open-ended responses across two questions.
03
RESEARCH QUESTIONS
Is a more self-directed experience actually wanted across the market — not just tolerated by existing customers?
Where do consumers still genuinely need a human in the process?
Where can automation or self-service carry the load without hurting the experience?
04
DESIGN DECISIONS THAT SHAPED WHAT THE SURVEY COULD SEE
Sampled the market, not just customers
A quota would have capped who could respond and produced a customer view. Sampling the broader market — including competitors' customers — produced a market view, the only way to answer whether the market wants self-service. A company's own customers self-selected into its current high-touch model, so they can't answer that question alone. The trade: the sample self-selects on who responded, so it's directional on market composition, not projectable to it. I took that deliberately.
Preference before experience
I asked how people wanted to interact before asking how they actually did, so their lived experience couldn't overwrite their stated preference.
Behavior before evaluation
Frequency and channel questions first, concrete counts next, satisfaction scales last — so evaluation couldn't color the behavioral record.
Coding chosen per question
Deductive on issue themes, where prior research had mapped the problem space, so the job was quantifying it.
Inductive on positive feedback, because nobody had mapped what people actually valued — and a deductive scheme there finds only what you go looking for.
05
HOW I ANALYZED IT
Likert-scale questions were analyzed quantitatively as response distributions and cross-tabbed against categorical variables. Open-ended responses were coded into an 8-theme, 29-subtheme taxonomy. The most decisive move was a cross-tab: satisfaction against how long reimbursement took, which turned a vague complaint into a precise threshold.
Satisfaction held near 92% through a week, then fell away past 8–10 days. The threshold came from the data, not a guess — it's the point where the trend breaks. (The 20+ bucket is the smallest, so that figure is directional.)
06
WHAT I FOUND
People wanted control — with help within reach
Asked how they wanted to interact with the service, the market didn't split cleanly into DIY versus done-for-me. The center of gravity was hybrid.
How people wanted to interact with the service.
51% wanted to run their own move; only 36% wanted it fully handled. But the single largest group wanted mostly self-serve with help available when needed. Reading the 51% as a mandate for pure self-service would have built the wrong product — people wanted to drive, with someone reachable.
Reimbursement delay drove the dissatisfaction
The cross-tab above located it precisely: confidence held for about a week, then fell off once reimbursement passed 8–10 days. That gave the business a concrete service target rather than a vague "be faster."
What people asked to fix
Coding the open-ends, speed and automation together made up about 65% of requests — faster processing and less manual entry. This is the qualitative data quantified: the open-ends held the specificity, and the coding made it countable.
07
RECOMMENDATION
Each recommendation traced back to a finding. The reimbursement cliff pointed to clearer expense guidance and a real-time view of expense status, so people aren't left waiting in the dark. The autonomy finding pointed to a hybrid service model — self-service tools with on-demand consultant support. Every recommendation did the same double duty: unburden consultants, the original goal, while keeping human help exactly where the data said people still needed it.
08
IMPACT
Two recommendations moved directly into build: a clearer guide to eligible expenses and timelines, and a real-time view into expense status. The hybrid service model went onto the roadmap.
The study gave the business what it needed to act on the consultant-workload problem — without guessing where automation would help and where it would break the experience.
09
WHAT I'D DO DIFFERENTLY
No quotas is a real trade
Sampling the whole market was the right call for the question, but it means the sample self-selects on who responded. The findings are directional on market composition, not projectable to it — and I present them that way.
Single-coder open-ends
I coded the open-ended responses alone. A second coder with an inter-rater reliability check — comparing how consistently we categorized the same responses — would strengthen confidence in the theme percentages.