Paper · Marketing

Which go-to-market motions produce a first paying buyer? An evidence review and a weekly operating system for early products

We reviewed public evidence on eight ways of reaching early buyers. Buyers form shortlists before talking to sellers, cold email reply rates depend on what is counted, and founder calls have no public benchmark. We then built a weekly operating system counted in paying buyers.

In Motion Automation research team · 26 October 2026

From above at dusk, two people sit still shaking hands in a glass meeting room over Tampa Bay while everyone else streaks past.

Abstract

Most new products stall before their first paying customer, and the founder has to choose among many ways of reaching buyers: sales calls, cold outreach, founder posts, answers from AI assistants, directories, referrals, free product tiers and paid ads. Public benchmarks for these choices are scattered, define their rates differently and often come from vendors. We reviewed primary sources for each of eight motions, including buyer surveys, platform rules, statutes, benchmark reports and pricing pages, and labelled every number by confidence. Business buyers form a shortlist before they speak to a seller: in one survey of 2,509 buyers, 81% had a preferred vendor before first contact, and in another of 1,169, AI chatbots were the most often chosen outside influence on the shortlist. Average cold email reply rates published for 2025 range from 0.45% to 3.43% depending on what is counted, so a team has to rely on its own counts. Google states that its AI answers need no special files or markup, and three independent analyses found no citation gain from an llms.txt file. No public source gives a comparable conversion rate for founder calls, founder posts or referrals. From this evidence we built an operating system that works for any product: five stages with exit tests counted in buyers, a buyer list in which every row links proof that the person already pays for the job, a five-day weekly cadence with one number, kill, hold and double-down rules, a compliance checklist for US, EU and UK outreach, and tool stacks at three budgets.

Introduction

A new product earns its first money from a stranger long after its first line of code. In between, the founder decides each week where to spend a limited number of hours: calling people, writing to them, posting, getting listed, building a free tier, or buying ads. Each of these is a motion, one repeatable way of reaching buyers. The work of choosing and running motions is go-to-market (GTM) work. In this paper a paying buyer is a person or firm outside the founder’s own circle who has paid money, and every rule we propose is counted in paying buyers.

Two changes make the choice harder in 2026. First, business buyers do much of their comparing before they contact any vendor, and they increasingly ask AI assistants which vendors to consider. The few vendors a buyer compares are the shortlist. An AI answer can mention a firm by name or cite it by linking one of its pages. Second, mailbox providers and social platforms have tightened written rules on bulk and automated outreach, so the cheapest way to send a thousand messages is also the riskiest.

An earlier audit of a widely viewed twelve-move B2B playbook, published on X by the founder of a GTM agency, confirmed most of the playbook’s cited facts, reproduced its claim that a self-published “best agencies” list was named first in Google’s AI answers (but not in ChatGPT), and found the most written legal and platform exposure in high-volume cold email and in scraping LinkedIn engagers12. Two gaps remained. The playbook assumed a firm that already sells, with an account list, a CRM and ad budget, and it gave no rule for when to stop a channel. A founder with a product and no buyers needs a different starting point.

We asked two questions. What does public evidence say about how well each motion produces early buyers, and how strong is that evidence? What weekly operating system follows from it, one that works for any product and any founder with ten or more hours a week?

Evidence is strongest where buyers are surveyed and where platforms publish rules, and weakest where founders spend most of their early time. Contributions:

  1. An evidence table by motion with ranges, sources and confidence labels (Section 4, Table 2).
  2. A product-agnostic operating system with stages, a buyer-list method, a weekly cadence, decision rules, a compliance checklist and tool stacks at three budgets (Section 5).

Terms

Table 1 defines the terms used in the rest of the paper.

Table 1. Terms used in this paper.
TermMeaning here
Ideal customer profile (ICP)The kind of buyer who pays and stays, written as one sentence naming role, firm type and job.
Proof rowOne row of the buyer list with a public URL showing that this person or firm already pays, in money or staff time, for the job the product does.
Will-pay statementA buyer’s own words naming a price or terms they would pay, recorded verbatim in a call.
Positive replyA reply that asks for a call, a price, a trial or more detail. Auto-replies and refusals are not positive.
Reply rateReplies divided by a denominator. Sources use emails sent, recipients or contacts; Section 4.2 shows why this changes the number.
Free-to-paid conversionShare of new free accounts or trials that become paying within a stated window.
Stage and exit testOne of five stages of Figure 4. A stage ends when its exit test, a count, is met.
Weekly numberSingle count reviewed every Friday: unique paying buyers, later paying buyers per week by channel.
Leading indicatorsCounts that move before the weekly number: proof rows, conversations, will-pay statements, positive replies and AI-answer mentions.

Methods

Sources and reading

We read every source between 00:00 and 01:00 Eastern time on 26 October 2026, except those reused from the earlier audit of the playbook, read on 25 October 2026. We preferred, in order: statutes and regulator guidance; platform documentation and terms; independent research organizations and government statistics; surveys with a stated sample; vendor benchmark reports with a stated sample and definition; and vendor or agency pricing pages. Where a primary page did not load, we used an archived copy or a secondary report and labelled the claim partly confirmed. Appendix A lists each claim with its URL.

Confidence labels

Each evidence row carries one label. High: official or independent source with a stated method and sample, or a written rule. Medium: survey or vendor dataset with a stated sample and definition, where the publisher may benefit from the result. Low: a single firm’s data, an anecdote, or no public benchmark. Each verified claim also carries C (confirmed on the primary page), P (partly confirmed, for example only on a secondary page or with a missing term) or U (unconfirmed).

Design of the operating system

We derived the stages, cadence and rules from the evidence and from the earlier audit. Thresholds in the system, such as thirty proof rows or ten calls, are design defaults chosen to fit a founder working ten to twenty hours a week. They are not benchmarks, and each one can be changed to fit a different product or schedule.

Tool prices

Tool prices are list prices for month-to-month billing on the vendors’ own pages.

Results: evidence by motion

Table 2 summarizes the evidence. The subsections below explain the rows that shape the operating system.

Table 2. Evidence by motion. Ranges are as published; confidence follows Section 3. Sources in Appendix A.
MotionBest available numbersConfidenceWeakness
Founder calls and discoveryNo public conversion rate. Bridge Group SDRs hold a median 4.1 quality conversations a day in funded teams; Superhuman used 40% “very disappointed” as a fit thresholdLowTeam and survey context, not founder conversion
Outbound emailAverage reply rate 0.45% (unique replies per email, Belkins 2025) to 3.43% (all replies per email, Instantly 2025); top 10% of Instantly senders 10.7%+; 58% of replies come from step oneMediumVendor data; different denominators; no link to revenue
Founder posts and contentNo public post-to-buyer rateLowAnecdotes only
AI-answer visibilityAI chatbots were 17.1% of shortlist-influence responses, the top source (G2, 1,169 buyers); with an AI summary, users clicked a result in 8% of visits vs 15% without (Pew)MediumSurvey; correlational studies on what earns mentions
Directories and review sitesReview sites 15.1% of shortlist-influence responses (G2)MediumNo listing-to-paid rate
Partnerships and referralsPeers 8.9% of shortlist-influence responses (G2)LowNo referral conversion rate
Product-led loopsFree-to-paid: freemium 3 to 5% good, 8 to 12% great; trial with card 25 to 35% good, 50 to 60% great (200 products)MediumNeeds sign-ups first; vendor and consultant data
Paid adsGoogle search median conversion rate 7.52% and cost per lead $70.11 (16,446 US campaigns)MediumAll industries; leads, not buyers

Buyers pick a shortlist before they talk to a seller, and AI chatbots now shape it

In 6sense’s 2024 survey of 2,509 recent B2B buyers, 81% had chosen a preferred vendor before speaking with sales, and about 69% of the buying process happened before buyers engaged a seller3. G2’s 2025 survey of 1,169 B2B software decision makers asked which sources influence the vendors on a shortlist. Generative AI chatbots were the most common answer at 17.1% of responses, ahead of review sites (15.1%), vendor websites (12.8%), market research (10.6%), independent content (9.7%), peers (8.9%) and vendor salespeople (8.8%)4. Figure 1 plots these shares. The LinkedIn B2B Institute, working with the Ehrenberg-Bass Institute, estimates that about 95% of a firm’s potential business buyers are not in the market at any given time5.

Figure 1. Sources that influence which vendors reach a B2B software shortlist, share of responses4. Survey of 1,169 decision makers; vendor-published.

Both surveys are run by firms that sell to marketers, so we rate them Medium. Taken together they mean a founder’s direct outreach usually reaches a buyer who has already formed a view, and that a buyer who is not in the market will not reply however good the message is. The system therefore keeps two kinds of work running in parallel from the start: direct conversations with people who pay for the job today, and a small amount of work that helps the product appear where buyers form shortlists.

Published cold email reply rates differ by a factor of seven because they count different things

Instantly’s 2026 benchmark, built from its platform’s sends between 1 January and 18 December 2025, reports an average reply rate of 3.43%, a top quartile at 5.5% or more and a top decile at 10.7% or more; 58% of replies came from the first email in a sequence6. Its rate counts all replies, including replies to follow-ups, divided by emails sent. Belkins analyzed 7,530,489 emails sent in 2025 with 34,393 replies and reports 0.45%, counting unique replies with auto-replies removed7. Figure 2 shows both.

Figure 2. Published 2025 cold email reply rates under two definitions67.

Neither rate counts paying buyers, and both come from senders who use bulk tooling. Gmail’s sender guidelines ask senders to keep the user-reported spam rate below 0.3%, to authenticate with SPF and DKIM aligned for DMARC, and to offer one-click unsubscribe on marketing messages8. Google’s acceptable use policy for Workspace prohibits unsolicited mass email9. For an early product, outbound in this system means individually written messages to proof rows, at about twenty a day per sender, from the founder’s own mailbox. At that volume the team’s own reply count after fifty contacts tells it more than either benchmark.

Google says its AI answers need no special files, and three studies found no llms.txt effect

Google’s guide to AI features in Search states that site owners do not need llms.txt files or special markup for these features, that seeking inauthentic mentions is not helpful, and that structured data is not required but can help rich results10. SE Ranking’s analysis of about 300,000 domains found no measurable relationship between having an llms.txt file and AI citations, with 10.13% adoption11. Geoa compared 1,912 Nordic companies and found 43.6% of sites with the file cited against 43.3% without (p = 0.94)12. Trakkr’s comparison of 37,894 domains found the same median citation count with and without the file13.

What does correlate with AI mentions is being talked about elsewhere. Ahrefs found brand web mentions correlated with visibility in ChatGPT, AI Mode and AI Overviews at 0.664, 0.709 and 0.656, against 0.218 for backlinks, and found that “best X” lists made up 43.8% of page types cited by ChatGPT in its sample1415. These are correlations from a tool vendor. Pew Research found that Google users who saw an AI summary clicked a traditional result in 8% of visits compared with 15% without one, and clicked a link inside the summary in 1% of visits16. Visits from AI answers are also undercounted: one agency found that 189 leads said an AI assistant sent them while its CRM credited 28 to AI17.

For the system, AI-answer work means honest pages that answer buyer questions, listings in places buyers and AI answers read, a monthly check of whether answers name the product, and a “How did you hear about us?” field on every form and call. Google’s spam policies on scaled content abuse and site reputation abuse rule out mass-produced pages18.

Product trials that ask for a card convert far better, after people sign up

Kyle Poyar’s 2026 analysis with ChartMogul and ProductLed of about 200 products gives good and great free-to-paid ranges by model (Figure 3): freemium 3 to 5% good and 8 to 12% great; free trial without a card 4 to 6% and 10 to 15%; free trial with a card 25 to 35% and 50 to 60%1920. ChartMogul’s report gives a median of about 8% across models and about 30% for card-required trials. These rates start from a sign-up, so they only help once the product has traffic. For a product with no buyers, a free tier is a loop to add once paying buyers exist and can invite others.

Figure 3. Good and great free-to-paid conversion ranges by model19.

WordStream’s 2025 analysis of 16,446 US Google search campaigns reports a median conversion rate of 7.52% and a median cost per lead of $70.11 across industries21. A lead here is a form fill; the report does not track purchases. LinkedIn requires a daily budget of at least $10 and a matched company list of at least 300 companies2223. Paid ads therefore fit a product that already knows which message converts, which is Stage 4 below.

Founder calls, posts, referrals and directories have no comparable public benchmark

We found no primary source with a denominator for founder calls to paying buyers, founder posts to buyers, referral asks to buyers, or directory listings to buyers. What exists is practitioner guidance. Paul Graham’s essay advises founders to recruit users manually24. Superhuman’s founder used the share of users who would be “very disappointed” without the product, with 40% as the threshold, to judge fit25. The Bridge Group’s 2025 survey of 351 companies reports a median of 4.1 quality conversations a day for a full-time sales development rep26, which gives an upper bound on what one person can hold, not a conversion rate.

Because these motions lack benchmarks, the system measures them by counts that the founder controls (calls held, will-pay statements, referral asks made) and judges them against the weekly number.

An operating system built on this evidence

Five stages, each ended by a count of buyers

Figure 4 shows the stages and Table 3 states their rules. A stage adds motions only after the previous exit test is met, so early hours go to the motions that produce conversations, while one post and one AI-answer item a week start the slower motions.

Figure 4. Five stages from no buyers to a repeatable channel, with exit tests and motions added at each stage. Thresholds are editable design defaults.
Table 3. Stage rules. Thresholds are design defaults.
StageNameExit testMotions in use
0Proof-backed buyer list30 named people who pay for the job today, each with a proof URL; one-sentence ICPFounder research; 1:1 messages to the first 10
1Conversations10 calls held; 3 will-pay statements in the buyer’s wordsFounder calls, 1:1 outreach, founder posts from call notes
2First paying buyer1 outside buyer has paid; money receivedAdds AI-answer pages and directory listings
3Repeat buyers5 paying buyers from one ICP, 2 of them outside the founder’s networkAdds referral asks and product loops
4Repeatable channelOne channel adds paying buyers 4 weeks running at a known cost per buyerAdds paid ads, scaled compliant outbound, partners

Buyer list with a proof requirement

A row enters the list only with a proof URL: a job post for the role the product replaces, a public case-study quote naming a competing tool, a pricing page of a tool the firm uses, a forum post asking for the job to be done, or a marketplace listing that charges for the same job. Each row records name, role, firm, country (for the compliance column), channel, proof URL and proof type, date found, status, last touch, next step, will-pay statement and source of first contact. The list is built by hand from public pages; scraping LinkedIn and bought lists are excluded (Section 5.5). The proof requirement keeps out people who fit a description but have never paid for anything like the product.

Weekly cadence and one number

Figure 5 shows the week. Monday adds twenty proof rows and checks each row’s region and opt-out status. Tuesday and Wednesday send individually written messages and one follow-up, matching Instantly’s finding that most replies come from the first message and that Tuesday and Wednesday lead6. Thursday holds calls and logs will-pay statements verbatim. Friday is the weekly review. Two assets come from call notes each week: one founder post and one AI-answer item (an answer page, a comparison, or a listing).

Figure 5. Weekly cadence for a founder with ten to twenty hours a week.

The weekly number is unique paying buyers until Stage 3 ends, then paying buyers per week by channel. Leading indicators are proof rows added, conversations held, will-pay statements, positive replies, and AI-answer mentions across a fixed set of twenty buyer prompts checked monthly.

Decision rules

Each Friday, every running channel and every ICP gets one of three decisions (Table 4). The rules use minimum samples so that a channel is not judged on a handful of messages.

Table 4. Kill, hold and double-down rules (design defaults).
DecisionRule
HoldFewer than 50 proof-backed contacts reached on the channel, or fewer than 14 days since it started. Keep running unchanged.
Kill a message50 contacts reached, zero positive replies. Rewrite once from call notes; a second zero kills the message.
Kill a channel100 contacts reached across two messages, zero positive replies, for this ICP.
Kill an ICP10 calls held, zero will-pay statements. Rewrite the ICP from the calls that went furthest.
Kill a product path30 days in Stage 1 with zero will-pay statements, or 30 days after a paid offer goes live with zero outside paying buyers.
Double downA channel produced 2 or more paying buyers in 30 days at a cost per buyer below one month of the price. Move half of next week’s hours to it.

Compliance checklist

The checklist applies to every message and every list (Table 5). In the United States, CAN-SPAM covers business-to-business email with no exception; each commercial email must identify itself as an advertisement, carry a valid postal address and a working opt-out honored within ten business days, and use accurate headers and subject lines, with penalties up to $53,088 per violating email27. In the EU, a person whose data was not collected from them must receive the Article 14 notice, at the latest at first communication when the data is used to contact them28; Article 21 gives a right to object to direct marketing that must be brought to their attention at first communication, and processing for direct marketing stops on objection29; Recital 47 allows that direct marketing may be a legitimate interest30. In the UK, the ICO allows marketing email to corporate subscribers without consent, but sole traders and some partnerships need consent31. LinkedIn’s User Agreement prohibits bots, automated messaging and scraping32, and hiQ’s dispute with LinkedIn ended in a consent judgment33. X’s automation rules prohibit automated unsolicited replies, mentions and direct messages34. The FTC’s rule on consumer reviews and testimonials prohibits fake, bought and undisclosed insider reviews35.

Table 5. Compliance checklist applied to every list and message.
AreaCheckSource
US emailAd disclosure; postal address; opt-out honored within 10 business days; true headers and subjectCAN-SPAM
EU contactsArticle 14 notice at first message; objection offered at first message; legitimate-interest note on fileGDPR Art. 14, 21; Rec. 47
UK contactsCorporate address only without consent; sole traders need consentICO, PECR
Gmail and WorkspaceSPF, DKIM, DMARC aligned; spam rate under 0.3%; no mass unsolicited mail from WorkspaceGmail, Google AUP
LinkedInManual messages only; no automation, no scraping, no bought engager listsUser Agreement 8.2
XNo automated unsolicited replies, mentions or DMsX automation rules
Search and AI answersNo mass-produced pages; no paid or fake mentionsGoogle spam policies
ReviewsNo paid, fake or undisclosed insider reviews16 CFR Part 465

Tool stacks at three budgets

Figure 6 and Table 6 give three stacks at list price for month-to-month billing, read on 26 October 2026. The $0 stack covers Stages 0 to 2. The under-$200 stack adds a sending tool, an AI-answer tracker and call notes. The funded stack fits Stage 4.

Figure 6. Monthly list price of three tool stacks, month-to-month billing, read 26 October 2026.
Table 6. Tool list prices used in the stacks (monthly billing, USD). Stack: 0 = $0, 200 = under $200, F = funded.
ToolPlan and priceStack
Attio (CRM)Free up to 3 seats; Plus $44; Pro $99 per user0, 200, F
Fathom (call notes)Free; Premium $20; Business $34 per user, 2-seat minimum0, 200, F
Instantly (sending)Growth $47; Hypergrowth $97; Scale bundle $194200, F
Otterly (AI-answer tracking)Lite $29 (15 prompts); Standard $189; Premium $489200, F
Google WorkspaceBusiness Starter $7 list per user; Standard $14200, F
Clay (research tables)Launch $185; Growth $495F
LinkedIn ads$10 a day minimum, about $300 a monthF
HubSpot (alternative CRM)Starter $20 per seat listOption
Free toolsGoogle Search Console, Bing Webmaster Tools, spreadsheets0

Templates

Appendix B gives three templates for running the system: a call script outline, a thirty-day plan and a one-page experiment brief.

Discussion

Why the system starts with conversations

The surveys in Section 4.1 say buyers arrive with a preference formed elsewhere, and the motions that shape that preference (AI answers, review sites, peers) take months to show results. The motions with the best public numbers, product-led conversion and paid search, both start from traffic an early product does not have. Founder calls have no benchmark but give the one input every later motion needs: the buyer’s own words about the job and the price. Stages 0 and 1 are therefore almost entirely conversations, with one post and one AI-answer item each week to start the slower motions early.

Why reply benchmarks are not targets

Two vendors’ 2025 averages differ sevenfold because of the choice of denominator, so neither works as a target. The system replaces them with the team’s own counts against a minimum sample, and counts positive replies rather than all replies.

Limitations

Most benchmark sources are vendors who benefit from the result. The surveys measure stated influence, not purchases. The stage thresholds and decision rules are our design defaults and have not been tested across products.

Directions

Two tests would strengthen the system. First, running it on several products for thirty days and publishing the counts at each stage, including zeros, would show whether the thresholds hold across products. Second, checking AI-answer mentions on a fixed prompt set each month and comparing them with the source buyers report in the CRM would show how often AI answers send buyers who are credited elsewhere.

Conclusion

Public evidence on early go-to-market is uneven. It is fairly strong on how buyers shortlist and on platform rules, moderate on outbound, product-led and paid search rates, and absent on founder calls, posts and referrals. A founder can still run a disciplined week: build a list of people who already pay for the job, talk to them, count will-pay statements and paying buyers, apply written kill and double-down rules every Friday, and add motions only as each stage is met.

Reproducibility

All sources are public and listed with URLs in Appendix A. The figures plot values from those sources, and the stack totals in Figure 6 add the list prices printed in the figure.

Appendix A. Verification table

Table 7. Claims used in this paper, labels and sources (read 26 October 2026 unless marked). C = confirmed on the primary page, P = partly confirmed, U = unconfirmed.
IDClaimLabelFinding and source
Buyer behavior
V181% of buyers had a preferred vendor before speaking with sales; about 69% of the process happens before engaging sellers; 2,509 buyersChttps://6sense.com/science-of-b2b/2024-buyer-experience-report/
V2Shortlist influence: AI chatbots 17.1%, review sites 15.1%, vendor sites 12.8%, peers 8.9%, salespeople 8.8%; 1,169 decision makersCG2 report PDF. https://learn.g2.com/hubfs/G2-2025-Buyer-Behavior-Report-AI-Always-Included.pdf
V3About 95% of potential B2B buyers are out of market at a given timeChttps://business.linkedin.com/advertise/resources/b2b-institute/b2b-research/trends/95-5-rule
Outbound
V4Average reply rate 3.43%; top quartile 5.5%+; top decile 10.7%+; 58% of replies from step one; data 1 Jan to 18 Dec 2025; all replies per email sentChttps://instantly.ai/cold-email-benchmark-report-2026
V50.45% reply rate; 7,530,489 emails; 34,393 replies; unique replies, auto-replies removedChttps://belkins.io/blog/cold-email-response-rates
V6Gmail: spam rate under 0.3%; SPF and DKIM aligned for DMARC; one-click unsubscribe for marketingChttps://support.google.com/mail/answer/81126
V7Google acceptable use policy prohibits unsolicited mass emailCRead 25 Oct 2026. https://workspace.google.com/terms/use_policy.html
AI answers
V8Google: no llms.txt or special markup needed for AI features; seeking inauthentic mentions not helpful; structured data not requiredChttps://developers.google.com/search/docs/fundamentals/ai-optimization-guide
V9About 300,000 domains: no measurable llms.txt effect; 10.13% adoptionChttps://seranking.com/blog/llms-txt/
V101,912 companies: 43.6% vs 43.3% cited, p = 0.94Chttps://geoa.app/en/data/llms-txt-effekt
V1137,894 domains: same median citations with and without llms.txtChttps://trakkr.ai/trakkr-research/llmstxt-effect/answers/does-llms-txt-increase-ai-citations
V12With an AI summary 8% clicked a result vs 15% without; 1% clicked a summary linkChttps://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
V13Brand mentions correlate 0.664, 0.709, 0.656 with AI visibility; backlinks 0.218; “best” lists 43.8% of ChatGPT-cited page typesCRead 25 Oct 2026. https://ahrefs.com/blog/ai-brand-visibility-correlations/; https://ahrefs.com/blog/best-lists-research/
V14189 leads said AI sent them; CRM first-touch credited 28CRead 25 Oct 2026. https://beomniscient.com/blog/first-touch-vs-self-reported-attribution-aeo/
V15Spam policies cover scaled content abuse and site reputation abuseChttps://developers.google.com/search/docs/essentials/spam-policies
Product-led and paid
V16Good and great free-to-paid ranges by model, about 200 productsPSeen in the author’s note, not a full report. https://substack.com/@kylepoyar/note/c-209940962
V17Median free-to-paid about 8%; card-required trials about 30%Chttps://chartmogul.com/reports/saas-conversion-report/
V18Median conversion rate 7.52%, cost per lead $70.11, 16,446 US search campaignsChttps://www.wordstream.com/blog/2025-google-ads-benchmarks
V19LinkedIn: $10 daily minimum; company list of 300 or moreCRead 25 Oct 2026. https://business.linkedin.com/advertise/ads/best-practices/maximize-your-budget; https://www.linkedin.com/help/lms/answer/a423102
Founder-led
V20Recruit users manuallyChttps://paulgraham.com/ds.html
V2140% “very disappointed” thresholdChttps://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/
V22SDR median 4.1 quality conversations a day; 351 companiesChttps://www.bridgegroupinc.com/research/2025-sdr-models-metrics-report-the-bridge-group
Compliance
V23CAN-SPAM: no B2B exception; ad disclosure, postal address, opt-out within 10 business days; up to $53,088 per emailChttps://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business
V24GDPR Art. 21(2) to (3): right to object to direct marketing, stated at first communicationChttps://gdpr-info.eu/art-21-gdpr/
V25GDPR Recital 47: direct marketing may be a legitimate interestChttps://gdpr-info.eu/recitals/no-47/
V26GDPR Art. 14 notice when data not collected from the personCRead 25 Oct 2026. https://eur-lex.europa.eu/eli/reg/2016/679/art_14/oj
V27UK: corporate subscribers may be emailed without consent; sole traders need consentCRead 25 Oct 2026. https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing
V28LinkedIn prohibits bots, automated messaging and scrapingCRead 25 Oct 2026. https://www.linkedin.com/legal/user-agreement
V29X prohibits automated unsolicited replies, mentions and DMsPLive page blocked the fetch; read from an archive of 12 Oct 2026. https://web.archive.org/web/20261012153529/https://help.x.com/en/rules-and-policies/x-automation
V30FTC rule prohibits fake, bought and undisclosed insider reviewsChttps://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials
Tool prices (monthly billing)
V31Clay Launch $185; Growth $495Chttps://www.clay.com/pricing
V32Instantly Growth $47; Hypergrowth $97; Scale bundle $194Chttps://instantly.ai/pricing
V33Otterly Lite $29; Standard $189; Premium $489Chttps://otterly.ai/pricing
V34Attio Free; Plus $44; Pro $99 per userChttps://attio.com/pricing
V35Fathom Free; Premium $20; Business $34, 2-seat minimumChttps://fathom.ai/pricing
V36Google Workspace Business Starter $7 list; Standard $14Chttps://workspace.google.com/pricing
V37HubSpot Starter $20 per seat listChttps://www.hubspot.com/products/crm/starter
V38Apollo Basic $65 a monthUPricing page did not render; only third-party pages. Not used.

Appendix B. Templates

Call script outline (30 minutes)

  1. Permission and purpose (2 min): who you are, why them, ask to take notes.
  2. Last time they did the job (8 min): what happened, who was involved, what it cost in money and hours.
  3. What they use and pay now (5 min): tools, people, contracts, renewal dates.
  4. What breaks (5 min): the last failure and its cost.
  5. Show, if a product exists (5 min): one flow on their case.
  6. Price question (3 min): “If this did that for you, would you pay [price] a month starting [date]?” Record the answer verbatim.
  7. Next step with a date (1 min).
  8. Referral and source (1 min): “Who else pays for this?” and “How did you hear about us?”

Thirty-day plan

Table 8. Thirty-day plan for any product starting at Stage 0.
WeekWorkExit count
1Write the one-sentence ICP; build 30 proof rows; set up CRM fields and opt-out handling; record an AI-answer baseline on 20 buyer prompts30 proof rows
2Send 1:1 messages to 20 rows a day for four days, one follow-up; first founder post from research notes5 calls booked
3Hold calls; log will-pay statements; second post and first AI-answer item; ask for a paid pilot where a will-pay statement exists10 calls held; 3 will-pay statements
4Follow up on pilots; collect first payment; Friday review with kill, hold or double-down per channel and ICP1 paying buyer or a written kill

One-page experiment brief

Each experiment fills seven fields before it starts: hypothesis in one sentence; channel and ICP; minimum sample (from Table 4); start and review dates; owner; cost in hours and money; and the decision each result will trigger.

Sources

  1. A. Vacca, “The only GTM playbook you’ll need in 2026,” X, 2026. x.com
  2. J. Krajewski, “A self-published shortlist leads Google’s AI answers but not ChatGPT’s: An audit of a 2026 B2B go-to-market playbook,” In Motion Automation Research, 2026
  3. 6sense, “2024 B2B buyer experience report,” 2024. 6sense.com
  4. G2, “2025 buyer behavior report,” 2025. learn.g2.com
  5. LinkedIn B2B Institute, “The 95-5 rule,” n.d.. business.linkedin.com
  6. Instantly, “Cold email benchmark report 2026,” 2026. instantly.ai
  7. Belkins, “Cold email response rates: 2026 benchmarks,” 2026. belkins.io
  8. Google, “Email sender guidelines,” Gmail Help, 2024. support.google.com
  9. Google, “Google Cloud acceptable use policy (applies to Google Workspace),” 2026. workspace.google.com
  10. Google, “AI features and your website: Optimization guide,” Google Search Central, 2026. developers.google.com
  11. SE Ranking, “llms.txt: Does it affect AI citations?” 2025. seranking.com
  12. Geoa, “Effect of llms.txt on AI citations,” 2026. geoa.app
  13. Trakkr, “Does llms.txt increase AI citations?” 2026. trakkr.ai
  14. L. Linehan, “Top brand visibility factors in ChatGPT, AI Mode, and AI Overviews (75k brands studied),” Ahrefs, 2025. ahrefs.com
  15. G. Allsopp, “Do self-promotional ‘best’ lists boost ChatGPT visibility?” Ahrefs, 2025. ahrefs.com
  16. A. Chapekis and A. Lieb, “Google users are less likely to click on links when an AI summary appears in the results,” Pew Research Center, 2025. pewresearch.org
  17. Omniscient Digital, “First-touch attribution captures 15% of our AI-sourced leads,” 2026. beomniscient.com
  18. Google, “Spam policies for Google web search,” 2026. developers.google.com
  19. K. Poyar, “Free-to-paid conversion benchmarks by model,” Substack, 2026. substack.com
  20. ChartMogul, “SaaS conversion report,” 2026. chartmogul.com
  21. WordStream, “Google Ads benchmarks 2025,” 2025. wordstream.com
  22. LinkedIn, “Maximize your budget,” n.d.. business.linkedin.com
  23. LinkedIn, “Company list upload requirements,” n.d.. linkedin.com
  24. P. Graham, “Do things that don’t scale,” 2013. paulgraham.com
  25. First Round Review, “How Superhuman built an engine to find product/market fit,” n.d.. review.firstround.com
  26. The Bridge Group, “2025 SDR models and metrics report,” 2025. bridgegroupinc.com
  27. Federal Trade Commission, “CAN-SPAM Act: A compliance guide for business,” n.d.. ftc.gov
  28. European Union, “Regulation (EU) 2016/679, Article 14,” 2016. eur-lex.europa.eu
  29. European Union, “Regulation (EU) 2016/679, Article 21,” 2016. gdpr-info.eu
  30. European Union, “Regulation (EU) 2016/679, Recital 47,” 2016. gdpr-info.eu
  31. Information Commissioner’s Office, “Business-to-business marketing,” n.d.. ico.org.uk
  32. LinkedIn, “User Agreement, Section 8.2,” 2025. linkedin.com
  33. hiQ Labs, Inc. v. LinkedIn Corp., “Consent judgment, No. 17-cv-03301-EMC (N.D. Cal.),” U.S. District Court for the Northern District of California, 2022. storage.courtlistener.com
  34. X, “Automation rules,” 2026. web.archive.org
  35. Federal Trade Commission, “Federal Trade Commission announces final rule banning fake reviews and testimonials,” 2024. ftc.gov

Want us to look at your business this way?

We measure where calls, leads and paperwork slip through, in your own records, before we build anything.

Book a free call
ResearchDownload the PDF17 pages