Beam Discovery
Go-to-market · initial pipeline
Iteration 1 — building the first pipeline for Process Discovery
Two motions — a waitlist and an outbound sales pipeline — pointed at 100 MQLs and 10 qualified customers by the end of September. This is the working version for the cofounder conversation: what the plan says, what the arithmetic says, what is missing, and the six decisions that have to be made in the room.
Where this lands
The plan names the right two motions and the right two personas. It is not yet buildable, because it is a list of deliverables rather than a funnel with arithmetic behind it. Four things are load-bearing and three of them are absent from the plan entirely.
- 1The window is 7 weeks, not 10. Today is 11 August; 30 September is 50 days away. Two to three of those weeks are consumed by email-infrastructure warmup that has not been started. That leaves about four weeks of sending — but only the first ten days of it can realistically mature into a qualified opportunity before 30 September.
- 2The $100 LinkedIn budget is off by roughly two orders of magnitude for the target it is attached to. At published 2026 enterprise benchmarks, 100 MQLs from paid LinkedIn costs $15k–25k. $100 buys under one lead. It is a creative test, not a channel.
- 3Cold outbound alone cannot deliver 10 qualified by 30 September. The arithmetic needs ~1,350–2,700 sequenced contacts and 9–18 warmed mailboxes, and enterprise deals do not mature inside three weeks. Warm intros, the existing prospect list, and founder network are the only channels that can close the gap — and none of them appear in the plan.
- 4The offer is the whole conversion mechanism and it is one line long. "Do a call and get a gap analysis" carries both streams. Until it has a name, a scoped deliverable, a turnaround, a sample artifact, and a weekly delivery cap, neither number can be planned or defended.
#01 The plan as it stands
Restated as written, with the stream titles corrected to match their outcomes. Two items are done; six are open.
- Website — shares.beam.ai/a/beam-discovery-landing-page-v2-product
- LinkedIn personalised advertisement — $100
- ICP definition — Process Owner, AI Transformation Head
- Freebie — a call, and you get a gap analysis
Outcome: 100 MQLs on the waitlist.
- ICP definition — Process Owner, AI Transformation Head
- Outbound campaign
- Demo
- Sales deck
Outcome: 10 qualified customers by end of September; land-and-expand ($10k POC → department-wide scale-up).
The personas are the strongest part of the plan, because they are already research-backed in the repo rather than invented here: Process Reality Owner and AI Transformation Sponsor. Both carry documented buying triggers, current pains, and the exact sentence each persona says when the product is working. That is unusual to have this early, and the outbound copy should be written directly from those two files rather than from scratch.
#02 One funnel, not two streams
Drawn as two parallel streams, the plan implies two independent machines with two independent targets. They are not independent. Both streams end at the same place — a person agreeing to a call, and receiving a gap analysis — and both feed the same $10k POC. Waitlist and outbound are two intakes to one funnel, not two funnels.
This matters for the targets. If the waitlist is a source of the September pipeline, then 100 MQLs and 10 qualified customers are two readings of the same chain, and only one of them is a real goal — the other is a leading indicator. If the waitlist is a separate audience being built for a later launch, then it is not competing for the same seven weeks and should be resourced accordingly. As written, the plan does not say which, and the two readings imply very different work. That is decision D1.
#03 What the arithmetic says
Three calculations decide whether this plan is fundable, staffable, and reachable. All three use published 2026 benchmarks, listed in section 10 — they are industry ranges, not Beam's own data, and should be replaced with real numbers the moment there are any.
A. The paid budget does not reach the paid target
Beam's ICP is Director-and-above: process excellence leads, heads of transformation, CIOs, Chief AI Transformation Officers. That is the most expensive audience on LinkedIn. Published 2026 benchmarks put average B2B sponsored-content CPC at $5–12, with VP+ targeting running 3–5× that and C-suite reaching $25–40. Cost per lead through native lead-gen forms sits at $50–130 for general B2B and $150–250 for C-suite enterprise targeting.
| What $100 buys at Beam's ICP | Optimistic | Realistic | Reading |
|---|---|---|---|
| Clicks to a landing page | ~13 | ~4 | Not enough traffic to read a single creative test conclusively. |
| Leads via native lead-gen form | ~0.7 | ~0.4 | Under one lead. The target is 100. |
| Budget required for 100 MQLs | $15,000 | $25,000 | Two orders of magnitude above the line item. |
The conclusion is not "spend $20k". It is that paid LinkedIn is not the channel that produces 100 MQLs at this stage, and the $100 should be redeployed. The same money spent on domains, mailboxes and a sequencer buys an entire outbound channel rather than a fraction of one lead. That is decision D4.
B. Ten qualified customers, worked backwards
Sequencing 1,350–2,700 contacts over roughly four weeks means 5,400–10,800 emails, or 1,350–2,700 emails per week. At the safe modern limit of 25–30 sends per mailbox per day, that is 9–18 mailboxes across 3–6 domains — all of which need two to three weeks of warmup before the first real send. Nothing in the current plan procures them, and the lead time is longer than the runway left after procurement.
C. Delivery capacity caps the funnel before the market does
The freebie is a founder-delivered session plus a written artifact. Assume 60 minutes on the call, ~2 hours to produce the analysis, and a 30-minute readout: roughly 3.5 hours per gap analysis.
| Scenario | Volume | Founder hours | Verdict |
|---|---|---|---|
| Gap analysis for every qualified call held | 20 | ~70h over 5 weeks (14h/wk) | Tight but deliverable if it is one founder's main job. |
| Gap analysis offered to every waitlist MQL | 100 | ~350h | Not deliverable. The promise would break publicly. |
| Capacity-capped, ICP-gated (recommended) | 6–8 / week | ~21–28h/wk | Deliverable, and scarcity makes the offer stronger, not weaker. |
This is the clean argument for gating the freebie rather than broadcasting it: the constraint is real, and stating it — "we run six of these a week" — converts better than an open offer while protecting the quality of the artifact that is meant to prove Beam can do discovery.
#04 What is missing — the gap register
Ranked by whether the September number survives without them. Owner and decide-by are deliberately blank; they are the output of the conversation, not the input.
| # | Gap | Why it blocks | Severity | Owner | Decide by |
|---|---|---|---|---|---|
| 01 | Email sending infrastructure | No domains, mailboxes, warmup, or sequencer named. 2–3 week lead time. This is the single item most likely to make September unreachable, and it expires with every day it is not bought. | Blocking | ||
| 02 | Account-level ICP and a built list | Two personas are defined; an ICP is not. No firmographics, geography, industry, size band, trigger signals, exclusion rules, or target list size. Without them there is no list, and without a list there is no outbound and no ad audience. | Blocking | ||
| 03 | The offer, specified | "Do a call and get a gap analysis" is the conversion mechanism for both streams and is one line long. Needs a name, scope, deliverable format, turnaround, a sample artifact prospects can see before booking, and a weekly cap. | Blocking | ||
| 04 | Definitions of "MQL" and "qualified" | Both targets are unscoreable without them, and every weekly review will relitigate whether a given conversation counted. Section 05 proposes both. | Blocking | ||
| 05 | A system of record for pipeline | No CRM or tracker named. Clay is available for list building and enrichment; nothing is named for holding stage, owner, next step, and date. "10 qualified customers" cannot be counted from inboxes. | Blocking | ||
| 06 | Warm channel — intros, network, prospects in play | Absent from the plan and, per section 03, the only channel that can produce qualified enterprise opportunities inside seven weeks. Ameriprise is already a live prospect in the repo. Enterprise sponsors rarely take a cold meeting with an unknown vendor. | Blocking | ||
| 07 | What we are actually selling — software or outcome | The product is at the internal-dogfooding milestone. A $10k POC implies a founder-led outcome, not software access. This choice changes the deck, the demo, the offer, and who runs the call — and it is currently unstated. | High | ||
| 08 | A public capture surface with a form | The ticked website URL resolved to a Beam workspace sign-in. If there is no public page and no form writing to a durable store, Stream A has nowhere to land. | High | ||
| 09 | Founder-led LinkedIn content | This ICP lives on LinkedIn. Organic founder posting costs nothing, compounds, and warms the same audience the outbound touches. It is the highest-yield unbudgeted channel available and it is not in the plan. | High | ||
| 10 | Proof — one reference or a redacted sample | Reply rates on this ICP move materially on whether the sender can point at a named design partner or show a real artifact. The gap-analysis sample doubles as this. | High | ||
| 11 | Attribution and instrumentation | No UTM convention, no single event schema, no agreement on where a signup is recorded. Without it you cannot tell which intake produced a qualified deal, and every budget argument becomes an opinion. | High | ||
| 12 | What the $10k POC actually is | Scope, deliverables, duration, success criteria, and the expansion price. Needed before the first qualified call, not after it — it is what "qualified" is being qualified against. | Medium | ||
| 13 | Cold-outreach compliance posture | If the list includes DACH or EU contacts, the legitimate-interest basis, suppression handling, and imprint requirements need a decided position before the first send, not after a complaint. | Medium | ||
| 14 | Operating cadence and owners | No named owner per stream, no weekly number, no review ritual. Seven weeks without a cadence produces one panic in week six. | Medium |
#05 Six decisions for this conversation
Each has a recommendation attached. The recommendations are arguable — that is the point of having them written down before the conversation rather than after it.
Is the waitlist a source for the September pipeline, or a separate audience for a later launch?
The waitlist feeds the same gap-analysis offer and the same POC. 100 MQLs becomes a leading indicator, not a second goal.
The waitlist is built for a product launch later in the year. It gets its own timeline and stops competing for these seven weeks.
Are we selling software access or a founder-delivered outcome?
Sell access to Process Discovery. Requires a demo that survives a live enterprise call in the next four weeks.
Sell the discovery result — sessions run, gaps named, a runbook a Beam agent can be built from. The product is the delivery mechanism, not the SKU.
Which persona do we open on?
Holds the $10k budget and the board mandate. Hardest to reach cold, most gatekept, lowest reply rate.
Feels the pain daily, replies far more often, can host a session — but usually cannot sign $10k alone.
Where does the money go?
$100 into LinkedIn ads. Buys under one enterprise lead and no readable signal.
Roughly $1,000–1,400 across the seven weeks buys the entire outbound channel: domains, mailboxes, sequencer, enrichment, verification.
| Line | Indicative cost | Why |
|---|---|---|
| 3–6 sending domains | ~$60 one-off | Protects the primary domain's reputation. Never send cold from it. |
| 9–18 mailboxes | ~$60–130 / mo | Set by the volume arithmetic in section 03, not by preference. |
| Sequencer with warmup | ~$100–200 / mo | Warmup, rotation, reply detection, suppression. |
| Enrichment and verification | ~$200–350 / mo | Clay is already connected. Bounce rate is the fastest way to burn new domains. |
| LinkedIn ads | $0–100 | Hold. If spent at all, spend it retargeting people who already visited, not on cold enterprise reach. |
What counts as an MQL, and what counts as qualified?
A person at an ICP-fit account, holding an ICP-fit title, who gave a work email and asked for the gap analysis. Not: anyone who visited, or gave a personal address.
Call held, and: a named process with stated volume or pain, a named budget owner, a live trigger (mandate, stalled pilot, or an audit/backlog event), and an agreed next step with a date.
How many gap analyses can we deliver per week, and who runs them?
Anyone who books gets one. Breaks at ~100 MQLs — 350 founder hours.
A stated weekly number, ICP-gated, with a short qualifying form before booking.
#06 A draft ICP you can correct
The plan lists two personas, which is who to talk to. An ICP is which accounts to talk to, and it is what a list is actually built from. This is a starting position drafted from the repo's persona research and vision, offered to be argued with rather than adopted.
Size 1,000–20,000 employees. Large enough to have shared services and a transformation mandate; small enough that a decision does not take two quarters.
Geography DACH first, then UK and Benelux, on the strength of existing reach.
Industry Insurance, banking and financial operations, logistics, manufacturing, energy, healthcare administration — high-volume back-office work with real exception handling.
A publicly announced AI or automation initiative in the last 12 months.
Open roles for AI transformation, process excellence, or process mining.
Existing Celonis, UiPath, SAP Signavio or ServiceNow footprint — the process pain is already acknowledged and budgeted.
Recent shared-services consolidation, a stalled pilot, or a publicly reported backlog or audit finding.
Under ~200 employees — no shared services, no process owner distinct from the operator.
Pure software companies — the back-office volume is not there.
Accounts where the only reachable contact is IT with no business-process owner behind them.
Anyone who needs a procurement cycle longer than the POC itself.
Two numbers turn this into a work item: how many accounts does this definition produce, and how many contacts per account. The arithmetic in section 03 needs 1,350–2,700 contacts; at two to three contacts per account that is 500–1,200 accounts. If the ICP as drafted yields fewer, either the definition widens or the target moves — and that trade is better made deliberately now than discovered in week four.
#07 The offer, specified
Everything else in the plan is distribution. This is the thing being distributed, and it is currently one line. A proposed specification, again as a starting position.
A 60-minute working session on one real process they already care about — not a demo, not a discovery call in disguise.
Within 48 hours, a written analysis: how the process actually runs including the exceptions, where the official version and the real version diverge, which steps are automation-ready, which are not, and what evidence is still missing.
A 30-minute readout, and a version of the document built to be forwarded to their sponsor.
A real discovery session run through Process Discovery — dogfooding that produces pipeline instead of competing with it.
A named process, a named budget owner, and a live trigger — which is exactly the qualification bar from D5.
An artifact that, redacted, becomes the sample that makes the next hundred emails credible.
The naming matters more than it looks. "Gap analysis" describes the method; the personas in the repo buy an answer. Something closer to "which of your processes is actually ready for an agent — and what's missing from the ones that aren't" is the sentence the AI Transformation Sponsor is already trying to answer, and it lifts almost verbatim from their documented success criteria.
#08 Seven weeks, laid out
11 August to 30 September. The shape of this chart is the argument: the warmup band is unmovable, it consumes the front third, and the maturation cutoff means sends after roughly 9 September are working on October, not September.
| Week | The one thing that must happen | Everything else |
|---|---|---|
| W1 · 11 Aug | Buy domains and mailboxes; start warmup. Nothing else on this plan has a hard lead time. | Settle D1–D6. Draft the ICP into a Clay table. Start the warm-intro list — every account either founder can reach through someone. |
| W2 · 18 Aug | Produce the sample gap analysis from a session already run. | Enrich and verify the list. Write the sequences from the persona files. Decide the pipeline system of record. Post on LinkedIn. |
| W3 · 25 Aug | First warm meetings held — these are the September number. | Warmup completes. Sequences loaded and tested on a small cohort. Booking link and qualifying form live. |
| W4 · 1 Sep | Cold sending at volume begins. | First cold replies worked same-day. Gap analyses start running at cap. Weekly scoreboard goes live. |
| W5 · 8 Sep | The maturation cutoff falls on 9 Sep. Sends after Tuesday are working on October, so the week pivots from prospecting to converting what has already landed. | First POC proposals go out. Re-cut sequences on early reply data — those rates are what iteration 2 gets planned from. Second content push. |
| W6 · 15 Sep | Convert, do not prospect. Attention moves to the deals that exist. | Gap analyses delivered at cap. Sponsor conversations from forwarded artifacts. Chase the POC decisions. |
| W7 · 22–30 Sep | Close what is closeable and count honestly against the D5 definition. | Write the retro while it is fresh: real conversion rates replace the benchmarks used here, and iteration 2 gets planned from data. |
#09 The weekly scoreboard
One table, reviewed at the same time every week. The point is not reporting — it is that by week three the benchmark assumptions in section 03 get replaced by Beam's real conversion rates, and the plan stops being a guess.
| Metric | Type | Weekly target once sending | Why it is on the list |
|---|---|---|---|
| Contacts sequenced | Leading | 340–680 | The only number fully under your control. If it misses, everything downstream misses four weeks later. |
| Positive replies | Leading | 7–14 | The fastest read on whether the list or the copy is wrong. Below 1%, stop and fix before sending more. |
| Calls booked | Leading | 7 | Booked is not held. Track both or you will over-report. |
| Calls held | The number | 5 | The single weekly number. Everything upstream exists to produce it; everything downstream is decided in it. |
| Gap analyses delivered | Lagging | 6–8 cap | Also the dogfooding count — product progress and pipeline progress in one metric. |
| Qualified (D5 bar) | Lagging | 2–3 | The September target, scored strictly against four criteria. |
| POC proposals sent | Lagging | 1–2 | Distinguishes real interest from a pleasant conversation. |
| Bounce rate | Health | < 2% | Above 3% the domains start burning and the channel dies quietly mid-plan. |
Attribution is worth five minutes of setup now: one UTM convention, and a source field on every record that survives to the qualified stage. Without it, the week-seven retro cannot answer which intake produced revenue, and iteration 2 gets planned from the same guesses as iteration 1.
#10 Assumptions and what I could not verify
Every conversion rate here is a published industry benchmark, not Beam data. They are the right order of magnitude for planning and the wrong thing to defend a number with by week four. Replace each one as real data arrives.
The stream titles are inverted and the outcomes are correct.
Two founders, no dedicated SDR, no marketing hire inside the window.
The $10k POC is a founder-delivered engagement rather than a software licence — this is D2, and if it resolves the other way, sections 07 and 08 change materially.
The landing page. The URL in the plan returned a Beam workspace sign-in page, so its copy, form, and fields could not be checked. If it is genuinely gated, Stream A has no public capture surface.
Existing pipeline. No CRM was available, so any deals already in flight — including Ameriprise, which is documented in the repo as a live prospect — are not counted in the arithmetic. If there is meaningful pipeline already, the September number is closer than section 03 suggests.
Sales deck and demo. Marked done and open respectively; neither was inspected here.
Benchmark sources
LinkedIn CPC, CPM and cost-per-lead by seniority and industry —
Digital Applied, LinkedIn Ads Benchmarks 2026,
Stackmatix, LinkedIn Ads Cost Per Lead Benchmarks,
Meet Lea, LinkedIn Cost Per Lead 2026,
Dupple, LinkedIn Ads B2B Cost in 2026.
Cold-email reply, positive-reply and meeting-booked rates —
Agentic Demand, B2B Cold Email Benchmarks for 2026,
Leadhaste, Meeting Booked Rate Benchmarks 2026,
Reachoutly, Cold Email Response Rate 2026,
Martal, B2B Cold Email Statistics 2026.
Personas and positioning — Process Reality Owner, AI Transformation Sponsor, Vision.