What Is a Research Virtual Assistant? The Role Quietly Deciding Which Businesses Pull Ahead
It is 4:15 on a Tuesday afternoon. Three hours ago you opened a browser tab to answer one question: what are our three closest competitors charging now, and has anything shifted since March?
You currently have forty-one tabs open. Two are pricing pages. Six are pricing pages you already read but cannot remember reading. One is a PDF from 2021 that looked authoritative until you checked the date. Four are Reddit threads where someone almost answers the question. There is a spreadsheet, half-built, with a column header that says “NOTES???” and nothing under it. Somewhere in there is a LinkedIn profile of a person you meant to look up for a completely different reason.
The question is still unanswered. The proposal you were supposed to write today has not been started. And tomorrow, when a client asks you what the market is doing, you will say something vague and confident, because the alternative is admitting that you spent an afternoon researching and came away with tabs instead of answers.
This is the most expensive habit in small business, and almost nobody puts it on a budget line. Research is not a task that gets skipped. It gets absorbed — by founders, by operations managers, by whoever was closest to the question when it landed. It never appears on a payroll report. It just quietly eats the hours that were meant for the work that actually moves revenue.
A research virtual assistant is the correction to that. And the gap between businesses that have made this correction and businesses that have not has become wider than most owners realise.
In this article
- So What Is a Research Virtual Assistant, Exactly?
- The Invisible Bill: What Unassigned Research Already Costs You
- The Seven Research Jobs Businesses Hand Over First
- What the Alternatives Actually Cost
- The Human in the Loop: Why Research Is the Last Place to Remove a Person
- The South African Advantage
- Managed, Not Matched: Why the Delivery Model Decides the Output
- The Competitive Gap, Stated Plainly
- DIY vs Generic Freelancer vs VAConnect: A Side-by-Side
So What Is a Research Virtual Assistant, Exactly?
A research virtual assistant is a trained remote professional who owns the gathering, verifying and structuring of information your business needs to make decisions — so that the decision-maker receives a brief instead of a browser history.
That definition is doing more work than it looks. Notice what it does not say. It does not say “someone who Googles things for you.” It does not say “a junior who copies data into a spreadsheet.” And it does not say “an analyst.”
The distinction matters. An analyst is paid to interpret and to recommend. A research VA is paid to establish what is actually true, from sources you would be comfortable defending, and to present it in a form you can act on in under ten minutes. Interpretation stays with you, because you are the one who knows what the business is trying to do. The research VA removes the part of the job that is mechanical, repetitive and enormously time-consuming — and leaves you with the part that requires your judgement.
In practice, a research VA sits across several of the service lines you already know. At VAConnect, research shows up inside general and administrative support (“inbox management, data entry, scheduling, research, and day-to-day admin”), inside sales support as lead and prospect research, and inside paralegal support as legal research and document preparation. The agency’s own service description lists market research alongside data entry, CRM management and customer support as core delivery areas.
The reason it rarely gets its own job title is the same reason it rarely gets its own budget: research feels like something everyone does incidentally, rather than something one person should own deliberately. That assumption is the problem.
Think about the last five significant decisions your business made. A pricing change. A supplier switch. A new market or suburb. A hire. A tool. Every one of those decisions rested on a body of information that somebody had to assemble. If you cannot name who assembled it, the honest answer is that it was assembled badly, in fragments, late at night, by someone who was already tired.
The Invisible Bill: What Unassigned Research Already Costs You
Here is where the numbers get uncomfortable.
Research firm IDC has long put the figure at roughly two and a half hours per day — about 30% of the working day — spent by knowledge workers on locating information. McKinsey Global Institute’s often-cited estimate lands at around 1.8 hours daily, or 9.3 hours a week. The framing McKinsey used has stuck with a lot of people who read it: a business hires five employees, but only four turn up to do the work. The fifth spends the week hunting for answers and produces nothing.
The UK-based research from Interact arrived at a similar place from a different angle: close to 20% of business time — one day out of every working week — is lost to employees searching for the information they need to do their jobs. A Forrester study commissioned by Airtable found large organisations running an average of 367 separate software applications, with knowledge workers spending 30% of their time simply looking for data across them. Atlassian’s own research suggests teams can lose up to a quarter of their time hunting for answers, and that more than half of employees regularly resort to asking a colleague or booking a meeting just to find something.
Asana’s State of Work Innovation research puts a name to the whole category: 60% of work time now goes to “work about work” — searching, switching, chasing, coordinating, tracking down decisions that were already made somewhere.
Then there is the compounding cost, which is the one people underestimate most. Dr Gloria Mark’s well-known University of California, Irvine research established that it takes an average of 23 minutes and 15 seconds to fully regain deep focus after an interruption. Harvard Business Review has estimated that knowledge workers toggle between applications and websites around 1,200 times a day. Research consistently suggests frequent switching can consume up to 40% of productive time.
Put those together and the picture of your Tuesday afternoon stops being a personal failing and starts being an entirely predictable outcome.
The research you are doing yourself is not free. It is the most expensive labour in your business, because it is being performed by the person whose hour is worth the most — and performed badly, because that person is doing it between other things.
There is a second-order cost too, and it is harder to measure but easier to feel. Decisions get delayed because the research is not ready. Then decisions get made anyway, without it, because waiting was worse. Then the business absorbs the consequences of a decision made on a hunch, and nobody connects that consequence back to the forty-one tabs.
The Seven Research Jobs Businesses Hand Over First
When businesses do finally assign research to a dedicated person, the same categories come up again and again. These are the ones that produce the fastest, most visible relief.
1. Prospect and lead research
Building and enriching lists of the right people at the right companies — verifying job titles, checking that the business still exists in the form your CRM thinks it does, finding the decision-maker rather than the generic inbox, and noting the trigger events (funding, expansion, new leadership, a job posting) that make an approach timely rather than random.
This is the single most commonly delegated research job, and the reason is simple: it is enormously valuable, entirely mechanical, and every hour a founder or salesperson spends on it is an hour not spent talking to a human being.
2. Competitor and pricing monitoring
Standing, recurring surveillance of what the competition is charging, launching, hiring for and saying publicly. Not a one-off panic audit when a client mentions a rival’s new offer — a weekly or fortnightly rhythm that means you are never surprised.
Dedicated competitive intelligence platforms exist for this, and they are not cheap. Crayon runs somewhere in the region of $25,000 to $40,000 a year; Klue between roughly $16,000 and $45,000. A trained research VA delivering a structured fortnightly brief covers a meaningful share of what those platforms promise, at a fraction of the outlay, with the advantage of actually being read.
3. Supplier, vendor and partner sourcing
Finding candidates, comparing on real criteria rather than whoever ranks first, requesting and normalising quotes into a comparable format, checking registration and references, and flagging the things a spreadsheet does not show — like a supplier whose reviews all appeared in the same fortnight.
4. Market, location and expansion research
Sizing an opportunity before you commit to it. What does demand look like in this suburb, this city, this country? Who is already serving it? What does the regulatory picture require? What did the last three businesses that tried this get wrong?
CB Insights’ well-known analysis of startup post-mortems found that the leading cause of failure — cited in 42% of cases — was building something with no market need. That is not a research problem in the abstract. That is a research problem in the specific.
5. Candidate and talent research
Sourcing and pre-screening for a role, verifying claimed credentials, checking public professional history, and assembling a shortlist with reasons attached rather than a pile of CVs with no order.
6. Content, SEO and topic research
Keyword and topic mapping, competitor content audits, gathering the source material and statistics that turn a thin blog post into something with authority, and building the reference pack a writer needs before they start rather than while they are writing.
7. Due diligence and verification
The unglamorous one. Confirming that a prospective client is solvent, that a partner’s claims hold up, that the statistic you are about to put in an investor deck comes from somewhere real. This category has grown considerably in importance in the last two years, and the next section explains why.
What the Alternatives Actually Cost
Before looking at what a research VA costs, it is worth being honest about the price of every other route.
Hiring an in-house researcher. Analysis published by intelligence firm Elevated Signal in 2026 puts a fully loaded mid-level market research analyst at $150,000 to $175,000 per year once you add benefits, retirement matching, paid leave, payroll taxes, recruitment costs, training and the tool stack. Their point about the tool stack is worth repeating: a basic research setup runs $6,000 to $8,000 annually before you touch enterprise-grade platforms, at which point the software bill alone can pass $50,000.
Commissioning agencies. The same analysis puts a boutique research project at $5,000 to $40,000, mid-market firms at $40,000 to $100,000, global research houses at $50,000 to $500,000-plus, and strategy consultancies at $150,000 upward. A single focus group — eight to ten people, one location — comes in at $7,000 to $20,000 once you account for the facility, the moderator and the incentives. Syndicated industry reports offer a cheaper middle ground: IBISWorld charges roughly $745 to $1,095 per report.
Hiring a freelancer. Upwork’s median market research rate sits around $38 an hour, with the broader band running $25 to $100. The Elevated Signal team flag the structural risk honestly: single point of failure. One person gets sick, misreads the brief, or oversells a methodology they barely know, and the project stalls with no backup and no accountability.
Doing it yourself. Free, on paper. In practice, priced at whatever your own hour is worth, multiplied by the 20 to 30% of the week that research consumes, plus the 23-minute refocusing penalty every time you switch back.
Against that, a managed research VA from South Africa costs roughly $4 to $6 an hour for entry-level admin research, $6 to $10 for experienced professionals, and $12 to $15 for senior or specialist work, according to 2026 rate guides from providers including Cherry Assistant. VAConnect’s own positioning puts the annual saving against a comparable local hire at over $25,000.
A single focus group costs more than a year of dedicated research support from South Africa. That is not a marginal difference in value. That is a different category of decision.
The point is not that agencies and analysts have no place. For a regulated multi-market study with board-level scrutiny, they absolutely do. The point is that most businesses do not have that problem. Most businesses have a continuous, unglamorous, never-ending stream of medium-stakes questions that need reliable answers this week — and they have been solving that problem by not solving it.
The Human in the Loop: Why Research Is the Last Place to Remove a Person
The obvious objection to everything above is: surely AI has made this whole role obsolete?
It is a fair question, asked in good faith, and it deserves a real answer rather than a defensive one. The honest answer is that AI has changed the research job substantially — and in doing so has made the trained human more necessary, not less.
Start with what the tools do brilliantly. Drafting a first pass at market sizing, mapping a competitive field, pulling together a trend summary — work that cost $5,000 to $15,000 from a desk research firm five years ago now takes an afternoon and a $20 subscription. Elevated Signal’s assessment is that AI handles roughly 30 to 40% of what traditional research firms do. That is an enormous, genuine gain, and any research VA who is not using these tools daily is doing the job with one hand tied.
Now look at the other 60 to 70%, because that is where businesses are getting hurt.
The accuracy data is sobering. AIMultiple’s 2025 benchmark found hallucination rates between 17% and 45% across general-purpose language models. In a 600-prompt accuracy test run by Neil Patel’s team, the best-scoring model returned fully correct answers 59.7% of the time. Citation-specific research is worse: a 2026 arXiv study of source attribution in deep research agents notes documented citation hallucination rates ranging from 11% to 57% across commercially deployed models, while the GhostCite benchmark of 13 models across 40 domains recorded rates spanning 14% to 95% depending on the domain.
This is not a theoretical problem confined to labs. A study covered by STAT News in May 2026 examined over two million papers and 97 million citations and found fabricated references climbing steeply — from roughly one in 2,828 papers in 2023 to one in 458 in 2025, and one in 277 during the first seven weeks of 2026. Separately, an analysis of 4,841 papers accepted to NeurIPS 2025 identified at least 100 confirmed hallucinated citations across 53 papers, despite peer review.
If rigorous academic review is catching this imperfectly, the odds that a busy founder catches it while skimming a report at 9pm are not good.
And it does reach the outside world. A February 2026 survey of marketers by Neil Patel’s team found that 43% reported hallucinated AI content had actually gone public, with a further 39.8% describing a near miss. Developer forums have been describing the same pattern from the receiving end for a while now — the curl project famously stopped its bug bounty programme over a flood of AI-generated submissions, with maintainer Daniel Stenberg’s complaints drawing hundreds of comments across Hacker News and LinkedIn. One commenter’s summary of why the material is so corrosive is the best short description of the risk anyone has written: the output looks just right enough to seem legitimate.
There is a structural finding underneath all of this that explains why the human does not go away. Stanford’s AI Index, using the RE-Bench framework, found that on tasks constrained to two hours, AI agents outperformed human experts by a factor of four. But as the time budget stretched to 32 hours, human performance overtook AI by roughly two to one. Speed is the machine’s advantage. Sustained synthesis, and knowing when something does not smell right, is not.
The bottleneck in research has moved. It is no longer finding information — it is verifying it. And verification is a human skill that scales with training and judgement, not with compute.
This is what a good research VA actually is in 2026: not someone who searches instead of a machine, but someone who directs the machine and then holds it to account. They know which claim needs a primary source before it goes near a client deck. They notice when three sources are all quietly citing the same original press release. They open the PDF rather than trusting the snippet. They check the date. They flag the thing they could not confirm instead of smoothing over it.
Roughly 90% of research professionals now use AI tools regularly, and 83% of research organisations plan to increase AI investment. Almost nobody serious is arguing for a pure-automation model. ESOMAR’s code, which governs the profession, still treats human oversight as non-negotiable. The consensus that has settled is a hybrid: machines for breadth and speed, trained humans for design, verification and interpretation.
The businesses being embarrassed right now are the ones that skipped the second half.
The South African Advantage
If the argument above holds — that research is fundamentally a verification and communication job rather than a data-retrieval job — then where your researcher sits matters far more than it does for purely asynchronous work. Two attributes decide it: whether they can talk to you inside your working day, and whether they can write.
South Africa is unusually strong on both.
The clock. South Africa runs on GMT+2 and observes no daylight saving, which makes the offset predictable year-round. That places a South African researcher one to two hours ahead of the UK, giving six to eight hours of shared working day with London, near-complete overlap with Western Europe, and a solid live morning with US Eastern and Central time. Sourcefit’s assessment of the market puts it directly: the GMT+2 alignment gives near-complete business-hour overlap with the UK and Western Europe, a strategic advantage neither the Philippines nor India can match for European-focused operations.
For research specifically, this is not a convenience — it is a quality mechanism. Research is iterative. The first brief always produces a follow-up question. “This is useful but I actually need it split by region” is a two-minute conversation and a same-day turnaround when your researcher is awake, and a 48-hour round trip when they are eight hours out of phase. Multiply that across a quarter and the difference in output is not small.
The language. English is an official language of South Africa and the default language of business, higher education and government. South African professionals write business English — reports, briefs, summaries, client-facing documents — that reads naturally to British and American audiences without the editing pass that output from some other offshore markets requires. Sourcefit’s write-up makes the same observation about written proficiency specifically, which is the relevant one here, because a research deliverable is a written artefact. A brief that is technically accurate but awkwardly phrased still costs you an hour to rewrite, and that hour was the entire point of delegating.
Spoken fluency with a neutral, widely understood accent matters too, for the research jobs that involve picking up a phone: verifying a supplier, requesting a quote, confirming a detail that no website will tell you.
The cost. Estimates cluster tightly. HireSava puts savings at 40 to 70% against US staffing costs. Sourcefit puts them at 50 to 65% against UK and European salaries. Cherry Assistant’s 2026 rate guide places South Africa at $4 to $15 an hour through managed providers, noting that the country carries a modest premium over the Philippines which buys native-level English and the timezone overlap — and that where work is real-time, that overlap is often worth more than the price gap.
The depth. This is not a boutique talent pool. BPESA figures for 2025 put South Africa’s professional services and business process sector at $5.3 billion with more than 270,000 workers. There is a mature, established industry here, not a handful of individuals with good internet.
The South African trade-off is not cheap-versus-good. It is a market where the timezone, the language and the cost all happen to point the same direction — which is rare enough that businesses who find it tend to stop looking elsewhere.
Add the cultural piece, which is difficult to quantify and impossible to ignore: South African professionals work within Western business conventions around deadlines, escalation, directness and feedback. A researcher who will tell you “I could not verify this and I do not think the source is reliable” is worth several who will hand you something confident and wrong.
Managed, Not Matched: Why the Delivery Model Decides the Output
Everything above describes what a good research VA does. It does not yet explain how you reliably get one — and that is where most businesses come unstuck.
The freelancer marketplace model hands you a profile and a rating and wishes you luck. You interview, you hope, you onboard, you train, you build a rhythm — and then that person takes a better-paying contract, or goes quiet, or turns out to have been three simultaneous clients deep the whole time. You are back to the profile page, starting over, and the institutional knowledge they built about your business, your competitors, your sources and your standards has walked out with them.
The managed model exists specifically to remove that risk from the client’s side of the table.
VAConnect began as Lime Tree Consulting in 2008 and became VAConnect in 2014, when founder Karen van Zyl concluded that the managed model — rather than a marketplace or a freelancer pool — was where the industry was going. The company describes the distinction as one where every VA is recruited, trained, monitored and supported by the agency, so that the client never has to manage the manager. Seventeen years and four continents later, the infrastructure behind that claim has been built out into four proprietary platforms: a dedicated South African talent portal for sourcing and pre-vetting, an upskilling academy with real-world skills testing (VAVarsity), plus the anti-burnout and two-way happiness accountability frameworks — Atomic Energy and VAPI — that sit behind the company’s reported 98% retention figure.
For a research role in particular, three parts of that model earn their keep.
Vetting before you ever see a shortlist. Skills testing, background checks and cultural-fit assessment happen in the pipeline, before a candidate reaches you. You are choosing between pre-qualified people, not filtering a marketplace.
Continuity engineered rather than hoped for. Research support compounds. A researcher in month six is dramatically more valuable than the same researcher in week one, because they know your sources, your standards, your competitors and what “good” looks like to you. A model built around retention protects that compounding; a model built around placements does not. VAConnect’s published position is that if a VA is not performing, they replace them — no fees, no friction — and that most matches are filled within about two weeks.
Backup cover. Freelancer research has a single point of failure by definition. A managed agency carries the team behind the individual.
The client evidence VAConnect publishes runs in the direction you would expect. A London SaaS co-founder’s verified Clutch review describes the VA as an extension of the team rather than an outsourced service, with 15-plus hours a week reclaimed in the first month and the placement retained over two years. A New York CEO’s review notes the handover was seamless and the quality had not dipped in two years. A Dubai fintech founder credits a VA-led social strategy with taking LinkedIn from 11,000 to 28,000 followers. A Cape Town maritime software client singles out the calibre of the screening process itself.
None of those are research-specific testimonials. They are something more useful: evidence that the delivery model produces people who stay long enough to become genuinely good at a specific business.
The Competitive Gap, Stated Plainly
Here is what makes this worth taking seriously rather than filing away.
Two businesses in the same sector, same size, same market. Business A does its own research — in gaps, at night, between other work, with forty-one tabs and no verification step. Business B has a trained research VA in a compatible timezone producing structured briefs on a schedule, using AI tools for breadth and human judgement for accuracy.
Business B does not simply save time, although it saves a lot of it. Business B knows things. It knows what the competition changed last fortnight. It knows which of its suppliers has quietly become uncompetitive. It knows the founder’s name and the trigger event before the sales call. It knows the market number in the investor deck is real, because someone opened the original source and checked the date.
Business A knows roughly what it knew last year, plus whatever surfaced by accident.
That gap does not announce itself. It shows up as Business B pricing more confidently, entering markets earlier, losing fewer deals to a competitor’s new offer, and making decisions in days rather than weeks. By the time it becomes visible in the numbers, it has usually been widening for a year or more.
The mechanics of closing it are not complicated. It costs less than one focus group. It does not require a new department, a new platform subscription, or a strategic overhaul. It requires deciding that research is a job somebody owns, and then giving it to someone built to do it properly.
If you want to see what that looks like in your business specifically, VAConnect will scope the role around your actual questions rather than a generic job description — start with a discovery call.
DIY vs Generic Freelancer vs VAConnect: A Side-by-Side
| DIY / In-House Coordination | Generic Freelancer or AI Tool Alone | VAConnect Managed Research VA | |
|---|---|---|---|
| Who does the work | You, or whoever was nearest the question | A marketplace hire, or an unsupervised AI output | A vetted, trained researcher who owns the function |
| Real cost | 20–30% of your week, at your hourly value | $25–$100/hr freelance; $20/mo AI plus your verification time | Roughly $4–$15/hr managed; $25,000+ annual saving vs a local hire |
| Verification step | Skipped when time is short — which is always | Depends entirely on one individual’s habits | Explicit standard: sources opened, dates checked, gaps flagged |
| AI hallucination exposure | Whatever slips past a tired reader | High — 17–45% model hallucination rates, 11–57% on citations | Managed: AI used for breadth, human verification before delivery |
| Timezone overlap | N/A | Frequently 7–11 hours out of phase | GMT+2, no DST — 6–8 hrs shared with UK, live US East mornings |
| Written English quality | Yours | Variable; often needs a rewrite pass | Native-level, neutral, British- and US-readable business English |
| Continuity | Continuous but always deprioritised | Single point of failure; no backup cover | Backup cover, replacement guarantee, 98% reported retention |
| Consistency of output | Sporadic — driven by crisis, not cadence | Format changes with whoever is available | Standard brief format, recurring cadence, compounding context |
| Ramp-up | Zero, but capacity is also zero | Weeks of unpaid trial and error | Pre-vetted shortlist; most matches filled within about 2 weeks |
| Ongoing training | None | The freelancer’s own initiative | VAVarsity upskilling; Atomic Energy and VAPI accountability |
| What you actually receive | Forty-one tabs and a half-built spreadsheet | A document you still have to fact-check | A brief you can act on in ten minutes |
VAConnect is Africa’s largest managed virtual assistant agency, placing rigorously vetted South African professionals with businesses across the UK, US, Europe, Canada and Australia since 2008. Explore our services or book a discovery call to scope a research role around the questions your business actually needs answered.