Open data on cold-traffic market tests
Test results across 1,590 hypotheses and 141 test-wave summaries in 18 niches: cost per subscriber and per lead, conversions and the rest of the metrics.
Why this table exists. If you are launching a new project, we may already have tested what you have in mind, or something close to it in your niche. Any data at the start can save you money and time. We have collected the full base of test results from several years of work: everything we formulated and launched in projects from 2019 to this day — job story, audience, offer, creative and the test numbers where a test took place. Besides the tested hypotheses, the base includes formulations and prepared test data that never reached a launch for one reason or another (for example, the result was already in and the tests were stopped).
Anonymised, open, citable. Updated: September 2026.
Hypothesis texts were written and launched in Russian; the English versions are translations. The original wording is in the Russian version of this page.
hypotheses in the base
tested with ads
median cost per subscriber
median cost per lead
total test budget
niches
How we test hypotheses and how the numbers were calculatedWhere hypotheses come from, how a test is run, what we measure, how a verdict is made, limitations and license
1. Where hypotheses come from
Every hypothesis starts with audience segmentation and is written as a job story: “when I find myself in this situation, I want this outcome, so that I get this effect”. The formulation is extended with the segment context, the trigger, hidden barriers and the customer's own vocabulary. Hypotheses are agreed with the expert before launch; the experiments are not: the formulation is fixed and then checked by the market, not by opinion.
2. How a test is run
Hypotheses are launched in batches, usually ten per project. Each gets its own mini-offer and creative so that the gap between the formulation and what a person sees in the ad is minimal. A typical chain: creative → landing page or mini-landing → subscription to a bot or a webinar → offer of a consultation. The budget per hypothesis is $50 to $500, the duration is from a few days to a couple of weeks — enough to see a direction, not to reach statistical significance. Platforms: Meta (Facebook and Instagram); older projects used other networks; the base records only the year and the niche.
3. What we measure
For every hypothesis we record impressions and CPM, CTR, clicks, subscribers and cost per subscriber, leads and cost per lead, and where the funnel went further — meetings and sales. Conversions are calculated along the chain click → subscriber → lead → meeting. Cost per subscriber and per lead in the base is the budget divided by the number of subscribers or leads; if a report only had a ready figure, that figure is used. All amounts are in US dollars: roubles were converted at the average rate of that year, which is noted on the card.
4. How a verdict is made
The verdict — “worked”, “failed” or “ambiguous” — is the conclusion of the media buyer and the producer at the end of the test, quoted on the card verbatim. We did not force it into a single threshold: “worked” in a niche with a $5 lead and in a niche with a $30 lead are different things, and the decision was always made relative to the project goal and the other hypotheses of the same batch. Where a report only had a number without an assessment, the card says “no assessment”. Hypotheses that were never launched are marked “prepared for launch” if there is an offer or a creative, and “formulation” if there is only a job story.
5. Waves and the unit of data
The unit of data is one hypothesis tested in one wave. Most projects had several rounds: a first batch, conclusions, a second batch with changed offers. That is why the same formulation can appear with different numbers and years. Summaries of launches and campaigns, where the unit is a series of tests rather than a hypothesis, are a separate type — “wave summaries”: webinar show-up rates, ROMI and the outcome of the batch.
6. Limitations
These are exploratory tests, not research. Their value is in directions and the order of magnitude of the cost of entering a niche, not in statistical reliability: samples are small, audiences and geos differ, traffic prices change from year to year, so compare within one period. Some older rows were restored from reports without an exact date — they carry the project year or no year at all.
7. Anonymisation and license
Names of clients and experts, school names, landing page links, contacts and handles have been removed from the texts; projects are represented by niche and topic. The data comes from ad accounts, platform reports and internal reports of Digital Experts projects for 2020–2026. It can be freely quoted and used with attribution to Digital Experts (CC BY 4.0 license). Link: digitalexperts.tv/data.
By niche
Medians are calculated over the tested hypotheses of the niche. Click a niche to open all its hypotheses as a table with detailed cards.
| Niche | Hypotheses | Tested | Worked | Subscriber, median | Lead, median | Test budget | Years |
|---|---|---|---|---|---|---|---|
| New profession Learning a new profession from scratch: from tattooing and pastry to accounting and Python. | 144 | 81 | 20 | $1.97 | $7.59 | $3,065 | 2020–26 |
| Upskilling Already in the profession and want to grow: new methods, directions, income in their own practice. | 145 | 75 | 29 | $2.47 | $13.90 | $4,609 | 2020–26 |
| Career Changing jobs, leaving employment, side income and remote work regardless of profession. | 89 | 17 | 5 | $3.00 | $6.00 | $380 | 2021–26 |
| Business Owners and executives: sales teams, CRM, B2B services and suppliers, negotiations. | 230 | 114 | 40 | $5.15 | $10.98 | $6,026 | 2021–26 |
| Body and looks Weight loss and shape, nutrition for fitness, skin and hair, image and style, cosmetology. | 55 | 22 | 6 | $1.54 | $9.45 | $1,038 | 2023–26 |
| Mental health Stress, apathy, burnout, energy, sleep; yoga and qigong as anti-stress practices. | 60 | 26 | 3 | $1.77 | $9.43 | $979 | 2021–26 |
| Health Psychosomatics, menopause, men's health, addictions, children's nutrition, immunity. | 122 | 52 | 11 | $1.95 | $13.27 | $3,768 | 2020–26 |
| Psychology Relationships, family, self-esteem, emotions, crises, working with a psychologist. | 164 | 77 | 20 | $0.28 | $12.30 | $2,348 | 2020–25 |
| Hobbies Drawing, singing, decor, yachting, bread — for oneself, not for income. | 72 | 49 | 17 | $1.09 | $10.70 | $1,737 | 2020–25 |
| Self-development Purpose, mindset, productivity, goals and motivation. | 97 | 40 | 16 | $1.94 | $4.92 | $1,109 | 2020–25 |
| Esoterics Numerology, energy practices, ancestral work, astrology. | 45 | 26 | 6 | $2.78 | $8.49 | $1,060 | 2020–25 |
| Children Courses and clubs for kids and teens: languages, speed reading, IT, school preparation. | 108 | 69 | 18 | $1.19 | $20.15 | $1,937 | 2020–25 |
| Parents Parents: child behaviour, relationships with a teenager, motherhood. | 62 | 21 | 6 | $2.25 | $2.67 | $847 | 2021–24 |
| Online schools Launching courses and online schools, producers, funnels and chatbots for experts. | 85 | 43 | 13 | $4.50 | $11.20 | $1,694 | 2021–24 |
| Languages Foreign languages for adults, Russian as a foreign language, studying abroad. | 46 | 33 | 9 | $1.65 | $2.48 | $750 | 2020–25 |
| Lawyers Attorneys, legal support for business, licences, contracts. | 17 | 4 | 2 | — | — | $322 | 2026 |
| Accounting Bookkeeping services, taxes, reporting for business. | 15 | 13 | 6 | — | $11.86 | $645 | 2025–26 |
| Finance Investments, loans, mortgages, passive income, financial literacy. | 34 | 12 | 5 | $5.58 | $42.86 | $2,443 | 2021–26 |
Cheapest lead across the tests
Tested hypotheses and monthly summaries with a cost per lead, cheapest first; the verdict is a separate column — for recent tests it is often not yet made, for older ones a lead is a form or a registration, for recent ones a started conversation. Split by year, because traffic prices grow and 2020–2021 tests would always look cheaper. Click a row to open the card. The top includes tests from $30 and from three leads. These are exploratory $50–500 tests: the numbers show the order of magnitude of the cost of entering a niche, not statistics.
| Hypothesis | Niche | Budget | Conversion to lead | Cost per lead | Verdict |
|---|---|---|---|---|---|
A person wants to master a method that eliminates the cause of pain When I understand that classic massage doesn't solve the cause of the pain, I want to study a deeper method, so I can work not with the symptoms, but with the source of the problem. | Health | $122 | 1.9% | $3.81 | worked |
Accountant wants to grow into a financier the owner values When I already handle certain areas but don't understand the business as a whole, I want to see the company's financial picture, so that I become an accountant who's genuinely valuable to the owner.
Trigger: the owner asked, and I couldn't answer; fear of being replaced by a "smart" accountant | Upskilling | $63.24 | 6.1% | $3.95 | worked |
Beginner wants to figure out whether accounting is the right profession, and how to get into it | Upskilling | $407 | 0.8% | $5.22 | workedsummary |
An esthetician wants to master advanced protocols (peels, microneedling), not just basic techniques | New profession | $1,373 | 16.4% | $6.48 | workedsummary |
A practicing therapist wants to master a method that removes the cause of pain, not the symptom | Health | $226 | 0.8% | $6.84 | workedsummary |
Outsourcing accountant wants to confidently take on complex clients and raise their fee When a client comes to me from an industry I don't know (construction, manufacturing, foreign trade, tenders) — I want to confidently take them on and set a proper fee, so that I grow from "handling a couple of sole proprietors for pennies" into a full-fledged practice and leave employment.
Trigger: lost income + desire for freedom. Elmira is exactly this segment. The expert's advantage is strongest here: she is a practicing outsourcer with a company and a range of clients, and can teach "how to take on an unfamiliar niche and not drown" — no college teaches this, because the teachers aren't practitioners. | Upskilling | $68.63 | 1.1% | $6.86 | worked |
An esthetician wants to master facial massage that addresses the cause, not just the spa effect | New profession | $658 | 15.2% | $7.00 | workedsummary |
A beginner wants to confidently master the esthetician profession, without endless doubts | New profession | $853 | 16.3% | $8.29 | workedsummary |
An esthetician wants to choose procedures based on skin structure, not follow a template | New profession | $1,231 | 18.4% | $8.38 | ambiguoussummary |
Accountant wants to master working with marketplace sellers When a seller from Wildberries / Ozon / Kaspi comes to me and asks me to handle their bookkeeping, and I don't understand either the marketplace's way of working (commissions, returns, agency agreement), or how to properly book it, or how to calculate taxes under the simplified "income" regime with a marketplace — I want to figure out the logic of marketplaces and learn to manage such clients end to end, so that I can take the hottest market segment of 2025-2026, set a proper fee and not lose clients to competitors.
Buying trigger: a specific rejected seller client + understanding that there will be more and more such clients + fear that colleagues have already learned this and I haven't.
Hidden barriers: don't understand marketplace reports; don't understand what to count as revenue; get confused with returns and holdbacks; don't know how to calculate tax when commission is withheld; afraid of a tax audit for such a client.
Expert's advantage: "I already have sellers in my portfolio — I'll show you on live Wildberries and Ozon exports how to handle this from source documents to the tax report."
Client's vocabulary: "don't know how to deal with marketplaces," "these reports are impossible to understand," "sellers themselves don't understand their reporting," "the commissions ate everything," "lost three sellers," "my colleague already handles five." | Upskilling | $83.91 | 2.5% | $9.32 | worked |
A licensed esthetician wants to work confidently with skin, not feel like an impostor | New profession | $211 | 3.9% | $11.08 | ambiguoussummary |
Graduate of accounting courses wants to start working in the profession When I finished accounting courses, I want to understand how to start working in the profession, so that I don't feel helpless. | New profession | $60.35 | 0.4% | $12.07 | ambiguous |
Downloads
Machine-readable format for AI agents and developers: every row carries a link to the card, a ready-made citation sentence and the license. The data can be used with attribution to Digital Experts (CC BY 4.0); citing the source is required.
MCP server for AI agents
The same base over the MCP protocol: an agent searches by niche and formulation and gets cards with numbers and a ready-made citation sentence without downloading the whole file. Read-only, no sign-up, up to 60 requests per minute. Tools: list_niches, niche_stats, search_hypotheses, get_hypothesis, methodology.
https://mcp.digitalexperts.tv/mcp
Claude Code: claude mcp add --transport http de-hypotheses https://mcp.digitalexperts.tv/mcp
Claude / ChatGPT: Settings → Connectors → add by URL
Cursor: .cursor/mcp.json → {"mcpServers":{"de-hypotheses":{"url":"https://mcp.digitalexperts.tv/mcp"}}}Example prompt for an agent: “find tested hypotheses about English for children in the Digital Experts base with a lead cheaper than $20 and show the card of the cheapest one”.