{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-11-04T15:23:40.651Z"},"content":[{"type":"blog","id":"ca81bc3e-4197-428d-9f6a-06a59fdf67c5","slug":"research-team-of-one-2026","title":"The Research Team of One: How Solo UX Researchers Scale Impact in 2026","url":"https://www.koji.so/blog/research-team-of-one-2026","summary":"Solo UX researchers face an arithmetic problem, not a skill problem: median output is about one study per month, while the most common staffing ratio is one researcher per five designers (NN/g), giving each designer roughly 2.4 studies per year. Nearly 14% of researchers in 2025 worked at companies with no other dedicated researcher, and 63% cite time and bandwidth as their biggest constraint. The fix is structural — triage on decision value, templatize recurring study shapes, delegate shallow work to supported non-researchers, and automate moderation with AI interviews.","content":"**If you are the only researcher at your company, your problem is not skill — it is arithmetic.** The median researcher completes about six studies every six months. At the most common staffing ratio of one researcher per five designers, that works out to roughly 2.4 studies per designer per year. No amount of craft closes a gap that size. The solo researchers who succeed in 2026 stop trying to personally run every study and instead redesign how research gets produced: triage ruthlessly, templatize the repeatable, delegate the shallow, and automate the interviews themselves.\n\nThis guide covers what the staffing data actually says, the four traps that consume a team of one, and a concrete operating model — including where AI-moderated interviews change the math.\n\n## What the staffing data actually says\n\nThe reference point most teams quote comes from Nielsen Norman Group. In a survey of 557 respondents (377 of whom answered the ratio questions), the most common researcher-to-designer ratio was **one researcher per five or fewer designers, reported by 37%**. Extended across the product organization, NN/g described the typical shape as **1:5:50** — one researcher, five designers, fifty developers. That ratio means researchers make up roughly 11% of all UX staff.\n\nTwo details from that study matter more than the headline. First, **54% reported at least one researcher per ten designers**, and **8% reported one researcher per fifty or more designers** — so the distribution has a long, painful tail. Second, and most usefully, NN/g found that a higher designer-to-developer ratio *alone* did not correlate with greater organizational maturity or business impact. Headcount is not the variable that determines whether research matters.\n\nThe second data point is that the team of one is becoming *more* common, not less. In the User Interviews *State of User Research* report — 485 qualified responses collected between 25 July and 9 August 2025 — the share of UX researchers working at companies with **zero dedicated researchers was 19% in 2019, fell to 6% in 2022, and rose again to nearly 14% in 2025**. The function contracted after the 2022–2024 correction, and a lot of research work landed on individuals.\n\n![Column chart: the share of UX researchers at companies with zero dedicated researchers was 19% in 2019, 6% in 2022 and nearly 14% in 2025.](/visuals/ux-researchers-team-of-one-share-2019-2025/ux-researchers-team-of-one-share-2019-2025.png \"After falling to 6% in 2022, the team of one is common again. · Source: User Interviews, State of User Research report, 2025\")\n\nThe third data point explains what those individuals are actually up against. In Maze research covering 800+ researchers, designers and product managers, **63% named time and bandwidth as their biggest challenge**, rising to **70% at enterprise technology and software firms**. Not budget. Not stakeholder buy-in. Capacity.\n\n## The arithmetic nobody writes down\n\nPut the three numbers together and the situation becomes concrete rather than emotional.\n\nThe User Interviews data reports median output over the previous six months as **2 mixed-method studies, 3 qualitative studies, and 1 quantitative study** — six studies per half-year, or about **one study per month** for a working researcher.\n\n| Ratio you support | Studies per year | Studies per person supported |\n|---|---|---|\n| 1 researcher : 5 designers | 12 | 2.4 per designer |\n| 1 researcher : 10 designers | 12 | 1.2 per designer |\n| 1 researcher : 5 designers : 50 engineers | 12 | 0.24 per engineer |\n\nAt the modal ratio, each designer you support receives about **two and a half studies a year** of your dedicated attention. Every other decision they make that year is made without research, or with research someone else ran, or with no research at all.\n\nThis is the honest framing of the team of one: you are not under-performing, and you are not going to fix a 5x demand gap by working faster. You are running a production system whose throughput is capped at roughly one study a month, against demand from an entire product organization. The fix is structural.\n\n## The four traps of the team of one\n\n**1. Becoming an order-taker.** When demand exceeds capacity, the default triage rule becomes whoever asked most recently or most loudly. The result is a queue optimized for stakeholder anxiety rather than decision value.\n\n![A researcher working alone at a laptop, the only person fielding research requests from a whole product team](https://images.unsplash.com/photo-1758598305369-4cab4c1bc364?crop=entropy&cs=tinysrgb&fit=max&fm=jpg&ixid=M3w5NDUzMDN8MHwxfHNlYXJjaHw5fHxwZXJzb24lMjB3b3JraW5nJTIwYWxvbmUlMjBsYXB0b3AlMjBvZmZpY2V8ZW58MXwwfHx8MTc5MzcyMTAxM3ww&ixlib=rb-4.1.0&q=80&w=1080)\n\n_Photo by [Vitaly Gariev](https://unsplash.com/@silverkblack?utm_source=koji&utm_medium=referral) on [Unsplash](https://unsplash.com/?utm_source=koji&utm_medium=referral)_\n\n**2. Becoming the bottleneck.** If every piece of customer contact must route through you, you have made yourself the rate limiter on organizational learning. Teams that cannot wait will simply proceed without evidence — and they will be right to.\n\n**3. Doing all the work at the same depth.** A pricing decision worth seven figures and a copy change on a settings page both arrive as research requests. Treating them alike is the single most expensive habit in a one-person practice.\n\n**4. Never building the second-order assets.** Templates, a question bank, a repository, and a decision log take time you do not have — and they are the only things that produce compounding returns. Skipping them guarantees year three feels exactly like year one.\n\n## The operating model: triage, templatize, delegate, automate\n\n![Framework for a solo researcher with four parts: triage requests on decision value, templatize repeat studies, delegate shallow work and automate interviews.](/visuals/research-team-of-one-operating-model/research-team-of-one-operating-model.png \"One researcher can't run every study. This is how the work gets produced anyway.\")\n\n### Triage on decision value, not requester\n\nAsk two questions of every incoming request: *what decision does this change?* and *what does being wrong cost?* If the answer to the first is \"none\" — a surprisingly common outcome — the correct response is to decline and say why. If a request cannot name a decision, it is not a research request; it is a request for reassurance.\n\nPublish your triage rule. A visible rule converts every individual \"no\" from a personality conflict into a policy outcome. This is the single highest-leverage change most solo researchers can make in a week.\n\n### Templatize the repeatable\n\nRoughly 70% of the requests a team of one receives are variations on a handful of recurring shapes: why did this customer churn, why is this flow failing, what do people think of this concept, why do trials not convert. Each of those deserves exactly one canonical study design that you reuse.\n\nStart from a [user interview script template](/docs/user-interview-script-template) and a [research report template](/docs/ux-research-report-template) rather than a blank page. Keep a running question bank so that writing a study becomes assembly rather than authorship. See the [structured questions guide](/docs/structured-questions-guide) for how to standardize the measurable parts of every study.\n\n### Delegate the shallow, keep the deep\n\nDemocratization is not a threat to the team of one — it is the only way the team of one survives. The Maze data shows it is already happening: **designers (70%) and product managers (42%)** were running studies in 2025, and in the 2026 data **product managers (39%), market researchers (35%) and marketers (23%)** all run their own studies alongside dedicated researchers.\n\nThe question is whether that work is supported or feral. Maze found **61% of organizations provide non-researchers with access to tools and templates**, 49% maintain research libraries, 46% offer structured training, 45% provide dedicated researcher support — and **13% provide no support at all**. That last 13% is where bad research gets made and bad decisions follow.\n\nYour job shifts from producing all research to setting the floor: which methods non-researchers may use unsupervised, which require your review, and which are yours alone. Our [research democratization guide](/blog/research-democratization-scaling-insights-2026) and the [democratization playbook](/docs/research-democratization-playbook) cover how to draw those lines.\n\n![Table for a solo researcher: PMs and designers self-serve quick tests, AI runs always-on churn interviews, and you keep pricing, strategy and high-risk work.](/visuals/research-team-of-one-who-runs-what/research-team-of-one-who-runs-what.png \"Your job moves from running every study to setting the floor for the rest.\")\n\n| Work type | Who runs it | Your role |\n|---|---|---|\n| Concept checks, copy tests, quick reactions | PMs and designers, self-serve | Template owner, spot review |\n| Churn, onboarding, win/loss interviews | AI-moderated, always-on | Design the guide, own the analysis |\n| Pricing, positioning, strategy, segmentation | You | Everything |\n| Compliance, accessibility, high-risk populations | You | Everything, no exceptions |\n\n### Automate the interviews, not the thinking\n\nThe step that has actually changed since 2024 is moderation. Scheduling, running and transcribing interviews is the part of the pipeline that consumes a solo researcher and produces the least differentiated value — and it is now automatable.\n\nThis is also where AI adoption has concentrated. Maze reports **69% now use AI in at least some research projects**, with data analysis (76%), transcription (57%), study planning and drafting (56%), generating research questions (55%) and synthesis and reporting (47%) as the leading applications.\n\nAutomating moderation does not mean automating judgment. Deciding what to study, which trade-off matters, and what the findings mean for the roadmap remains entirely yours — and is the part of the job that justifies the headcount.\n\n## Where Koji fits for a team of one\n\nKoji is built for exactly this constraint: a small number of people who need research output far beyond their personal hours.\n\n![Koji voice interview in progress with an AI orb asking a follow-up question and a microphone button](/email/shots/shot-voice-interview.jpg \"An AI-moderated voice interview runs and probes without you in the room.\")\n\n- **AI-moderated voice interviews run without you in the room.** Launch a study on Monday, have 40 conversations completed by Wednesday. You are not scheduling, not attending, and not transcribing. See [how AI research agents work](/docs/ai-research-agent) and our comparison of [AI versus human moderators](/docs/ai-vs-human-moderators) for where each is appropriate.\n- **Six structured question types** — open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — mean one study returns both the quotes and the numbers. You stop running a qualitative study and a separate survey to answer one question. Details in the [structured questions guide](/docs/structured-questions-guide), [scale questions](/docs/scale-questions-guide) and [open-ended questions](/docs/open-ended-questions-ai-interviews).\n- **Automatic thematic analysis** turns 40 transcripts into themes with supporting quotes, removing the multi-day synthesis block that usually forces a team of one to run fewer, larger studies.\n- **One-click reports** give stakeholders something to read without you building a deck for every request.\n- **No moderator bias and no interviewer fatigue.** Human moderators degrade after four to six interviews in a day; the fortieth Koji interview is conducted exactly like the first.\n- **Customizable AI consultants** let you encode your own probing style once and apply it across every study.\n- **Recruitment you do not have to run.** Import a churn list or a CRM segment so each person gets a personalised link and the report is pre-segmented, share the generic anonymous link when candour matters, or open Panel recruit when the question needs people you have no list for: describe the audience, approve the price per completed interview in credits, and Koji finds them across 130+ countries, on paid plans. Sourcing participants is the other calendar-eater for a team of one, and it stops being yours.\n- **No per-seat pricing.** A team of one is not penalized for inviting stakeholders to read results — see [per-seat versus usage-based pricing](/blog/per-seat-vs-usage-based-research-tools-2026).\n\nThe practical effect on the arithmetic above: the constraint moves off your calendar. Studies stop being rationed by your available hours and start being rationed by what is actually worth knowing — which is the correct constraint.\n\n## A 90-day plan for a new team of one\n\n| Days | Focus | Concrete output |\n|---|---|---|\n| 1–30 | Map demand and stop the bleeding | Written triage rule, inventory of every open request, one decision log |\n| 31–60 | Build the reusable layer | Three canonical study templates, a question bank, a shared repository |\n| 61–90 | Shift the ratio | Two always-on AI-moderated studies running continuously; non-researcher self-serve tier defined |\n\nBy day 90 the goal is not more studies. It is that the studies which happen are the ones that change decisions, and that routine questions no longer require your calendar.\n\n## Prove the impact before someone asks\n\nThe CMX community data makes a point that transfers directly: teams whose data was connected to a system of record were **over twice as likely to rate their impact as extremely successful**. The lesson is not about tooling for its own sake — it is that impact you cannot evidence is impact you will be asked to justify.\n\nKeep a decision log from week one: the decision, the evidence, the date, and what changed. When budget season arrives, that log is the argument. Our [research ROI guide](/docs/research-roi-guide) covers how to frame it, and [UX researcher salary and team cost benchmarks](/docs/ux-researcher-salary-team-cost-benchmarks) gives you the comparison figures.\n\n## Frequently Asked Questions\n\n### What is a research team of one?\nA research team of one is a single dedicated researcher supporting an entire product organization, with no other research staff. It is common: User Interviews found that nearly 14% of UX researchers in 2025 worked at companies with zero dedicated research headcount besides themselves, up from 6% in 2022.\n\n### How many designers should one researcher support?\nNielsen Norman Group found the most common ratio is one researcher per five or fewer designers (37% of respondents), often described as 1:5:50 across researchers, designers and developers. However, 8% of respondents reported one researcher per fifty or more designers, so the tail is long.\n\n### How many studies can a solo researcher realistically run?\nMedian reported output is about six studies per six months — roughly one per month, split across mixed-method, qualitative and quantitative work. At a 1:5 ratio that is about 2.4 studies per designer per year, which is why triage and automation matter more than working faster.\n\n### Should a team of one let non-researchers run studies?\nYes, with guardrails. Democratization is already happening — 70% of designers and 42% of product managers were running studies in 2025 — so the real choice is whether it is supported or unsupervised. Define which methods are self-serve, which need your review, and which are yours alone.\n\n### Does AI moderation replace a solo researcher?\nNo. It replaces scheduling, moderating and transcribing — the highest-volume, lowest-differentiation parts of the job. Choosing what to study, interpreting trade-offs and influencing the roadmap still require a researcher, and those are the activities that justify the role.\n\n### How do I prove my impact as the only researcher?\nKeep a decision log recording each decision, the evidence behind it, and what changed as a result. Evidence you can point to is far more persuasive than study counts, and teams that connect research data to their systems of record are measurably more confident about their impact.\n\n## Run more research than your calendar allows\n\nIf you are a team of one, the constraint is your hours — not your judgment. Koji removes the moderation and synthesis bottleneck so the studies that matter actually happen.\n\n**[Start free with 10 credits](https://www.koji.so)** — no credit card, no research expertise required. Design a study in minutes, launch AI-moderated voice interviews, and get themes and quotes back in hours rather than weeks.\n","category":"Research","lastModified":"2026-11-03T15:50:13.536068+00:00","metaTitle":"The Research Team of One: How Solo UX Researchers Scale Impact (2026)","metaDescription":"At the modal 1:5 researcher-to-designer ratio and a median of one study a month, each designer gets ~2.4 studies a year. The operating model — triage, templatize, delegate, automate — that fixes the math for solo researchers.","keywords":["research team of one","solo ux researcher","one-person research team","ux research ratio","scaling user research","research democratization","ai moderated interviews"],"aiSummary":"Solo UX researchers face an arithmetic problem, not a skill problem: median output is about one study per month, while the most common staffing ratio is one researcher per five designers (NN/g), giving each designer roughly 2.4 studies per year. Nearly 14% of researchers in 2025 worked at companies with no other dedicated researcher, and 63% cite time and bandwidth as their biggest constraint. The fix is structural — triage on decision value, templatize recurring study shapes, delegate shallow work to supported non-researchers, and automate moderation with AI interviews.","aiKeywords":["research team of one","solo researcher","researcher to designer ratio","research capacity","research democratization","ai moderated interviews","research operating model"],"aiContentType":"guide","faqItems":[{"answer":"A research team of one is a single dedicated researcher supporting an entire product organization, with no other research staff. It is common: User Interviews found that nearly 14% of UX researchers in 2025 worked at companies with zero dedicated research headcount besides themselves, up from 6% in 2022.","question":"What is a research team of one?"},{"answer":"Nielsen Norman Group found the most common ratio is one researcher per five or fewer designers (37% of respondents), often described as 1:5:50 across researchers, designers and developers. However, 8% of respondents reported one researcher per fifty or more designers, so the tail is long.","question":"How many designers should one researcher support?"},{"answer":"Median reported output is about six studies per six months, roughly one per month, split across mixed-method, qualitative and quantitative work. At a 1:5 ratio that is about 2.4 studies per designer per year, which is why triage and automation matter more than working faster.","question":"How many studies can a solo researcher realistically run?"},{"answer":"Yes, with guardrails. Democratization is already happening (70% of designers and 42% of product managers were running studies in 2025), so the real choice is whether it is supported or unsupervised. Define which methods are self-serve, which need review, and which are researcher-only.","question":"Should a team of one let non-researchers run studies?"},{"answer":"No. It replaces scheduling, moderating and transcribing, which are the highest-volume and lowest-differentiation parts of the job. Choosing what to study, interpreting trade-offs and influencing the roadmap still require a researcher.","question":"Does AI moderation replace a solo researcher?"},{"answer":"Keep a decision log recording each decision, the evidence behind it, and what changed as a result. Evidence you can point to is more persuasive than study counts, and teams that connect research data to their systems of record are measurably more confident about their impact.","question":"How do I prove my impact as the only researcher?"}],"relatedTopics":["research team of one","solo researcher","research democratization","research operations","ux research capacity"]}],"pagination":{"total":1,"returned":1,"offset":0}}