There are two types of improvements worth discussing:

1. Further automations to build

2. Rethinking the overall approach

A. FURTHER AUTOMATIONS TO BUILD

1. Pull a Larger Dataset

52,000 companies sounds like a lot. After filtering, the real target universe is ~5,000. But you can’t find the “needle"“ without pulling the “full haystack” - the 50-row free tier sample isn’t sufficient to build the real pipeline.

2. Add Better Signals

The current signals used are good - company age, director age, accounts type, debt. The next layer would to layer behavioural data to enrich the “officer” data:

• Director LinkedIn activity - posting about retirement, business challenges, or growth fatigue

• Job listings - hiring a manager for the first time suggests the owner is trying to step back

• Trade website reviews - volume and recency as a proxy for revenue activity

• Domain age and website sophistication - a site built in 2009 and never updated is a strong unsophisticated-seller signal

These are useful spreadsheets (not mine) which gives signal ideas:

1. https://docs.google.com/spreadsheets/d/1Y3bbNPJ5tmOk2wlQO7ItWz3wzjM4tglPjlz6vnAqmwA/edit?gid=0#gid=0

2. https://airtable.com/appGc0v7TfWslJP7L/shrX0gSEGYdasEu58/tblBOqYAGCSL9bVRC

3. Use Claygent More Deeply

I used Claygent for B2B/B2C classification. There’s more it can do:

• Broker check - search Daltons, BusinessesForSale.com, and Christie & Co for any active listing

• Employee count estimation - triangulate from LinkedIn, job boards, and website team pages

• Revenue estimation - combine accounts type, website sophistication, and headcount signals

4. Segment Lists More Tightly

The best-performing outreach campaigns have gotten more specific over time. A message written for “UK plumbing company owners aged 60+” will always outperform one written for “UK SME owners.” Tighter segmentation produces better message-to-market fit — and better message-to-market fit is what drives reply rates.

The data infrastructure to do this is already built. It’s just a matter of slicing it more deliberately before hitting send.

5. Build Proper Outbound Infrastructure

None of this works without a properly configured sending environment.

Buy dedicated domains for cold outreach. Warm the inboxes for at least four weeks before sending. Configure DKIM, DMARC, and SPF records correctly. Deliverability is all about the infrastructure.

Multiple providers now offer pre-warmed mailboxes, which removes the warmup period entirely.

6. Build a Feedback Loop

Everything above is a system that runs once and produces output. The next level is a system that learns.

Once one is getting calls with business owners, the next step would be to introduce an AI layer that grades each conversation - which framing generated engagement, which questions triggered openness about exit, which objections came up repeatedly. That analysis feeds back into every upstream step. Better call data refines the copy. Better copy refines the scoring. Better scoring refines the targeting.

This is the difference between a campaign and a machine that compounds over time.

7. MCP Servers for Data Retrieval

One underexplored option: Model Context Protocol (MCP) servers as an alternative to Clay enrichment columns for certain data pulls.

An MCP server connected to Claude can call the Companies House API directly, run enrichment logic, and return structured data - all within a single conversation context. For bespoke data pulls or testing enrichment logic before committing Clay credits, it’s a faster and cheaper path.

B. RETHINKING THE OVERALL APPROACH

1. Extensive Prospect Research

Dan Kennedy made an observation: people don’t buy what you sell, they buy a solution to a problem they feel acutely.

The acquisition equivalent: don’t approach a business owner talking about deals and multiples. Approach them talking about the problem they’re sitting with.

What are the actual concerns of a 62-year-old plumbing company owner? To better understand “pain points”, one should go to where these owners are. Trade magazines, podcasts forums etc. This is market research.

2. Sharpen the Offer

The pipeline finds the right companies. But what am I actually offering?

“I want to buy your business” is one offer, but it might not be the most compelling. The message needs to communicate a specific value proposition.

For example:

• Clean exit, full price - Speed and certainty as the value proposition.

• Phased handover - Reduces the psychological barrier of walking away from something they built.

• Retained equity - Works well when the owner believes in the business but is tired of running it.

• Seller financing - structure part of the purchase price as a loan from the seller.

• Legacy preservation - keep the company name, keep the team, keep the local reputation intact.

• Employment guarantees - commit to retaining staff for a defined period. Often the thing an owner cares most about and the thing buyers never mention.

The offer determines reply rate as much as the targeting does. Different owners respond to different things - which is another argument for segmenting the list before sending.

3. MQL vs SQL - Defining the Objective

A reply isn’t a deal. But it might be sufficient - at least at first.

Most outbound campaigns are implicitly optimised for SQLs - getting a business owner on a call, ready to discuss an exit. That’s a high bar for a cold email. These are people who have never considered selling, have never met you, and are being asked to entertain one of the biggest decisions of their lives off the back of a single message.

An MQL - someone who replies with genuine interest, asks a question, wants to know more - is a more realistic first objective for cold outbound.

The job of the outreach sequence is to generate that initial signal of interest. The job of everything that follows - the follow-up, the education, the offer framing, the calls - is to convert that interest into a real conversation. Trying to collapse the whole funnel into a single cold email is why most acquisition outreach doesn’t work.

MQL first. SQL second.

4. Use Lead Magnets

Cold outreach asks for something. A lead magnet gives something first.

Done well, it builds a warm list of owners who’ve already engaged with relevant content - and signals credibility before any direct approach. AI makes this easy to produce:

• A valuation guide - what UK plumbing businesses sell for and at what multiples. The owner who downloads this is already thinking about exit.

• A benchmarking report - how their business compares on margin and revenue per employee. Business owners are competitive. They want to know where they stand.

• An exit readiness checklist - what buyers look for, what makes a business hard to sell. Frames you as the expert before you’ve spoken.

• A valuation calculator - estimate what a business is worth at 4x, 5x, or 6x EBITDA. Interactive beats PDF every time.

Tools like Lovable and Replit let you build and deploy personalised software and landing pages at scale.

Instead of sending every prospect the same guide, you generate a personalised version for each company, pre-populated with their data from Clay. Or build a hosted tool and send each prospect a unique link with their business already loaded. They click and see their own numbers.

The owner who engages with your content has self-selected.

5. Think in Terms of a Full Funnel

Email opens the conversation.

Thinking in terms of the full funnel changes what you build:

• Top of funnel - cold email, LinkedIn content, trade press, referrals. Objective: awareness and first contact.

• Middle of funnel - follow-up sequences, educational content, lead magnets. Objective: familiarity and trust with owners who’ve engaged but aren’t ready to sell yet.

• Bottom of funnel - direct calls, in-person meetings, personalised letters. Objective: converting a warm conversation into a serious discussion.

Most searchers invest everything at the top and have nothing for the middle or bottom. The owner who replies in March might be ready to talk seriously in September - but only if there’s been consistent, credible contact in the interim. The pipeline should manage the relationship until the owner is ready, not just generate first touches.

6. Apply ABM Principles to the Top Targets

ABM was designed for enterprise B2B sales. The principle translates directly to acquisition outreach.

The idea: identify your highest-priority targets and treat them as markets of one. Instead of sending everyone the same sequence, coordinate multiple touchpoints across multiple channels - email, phone, direct mail, advertising - all pointed at the same 50 companies simultaneously.

Retargeting ads are particularly underused here. Upload your target company domains to Meta or Google and serve ads specifically to people browsing from those organisations. The owner who received a cold email on Monday, can see a LinkedIn post on Wednesday, and then an ad on Friday.

Clay has launched a new “ads” feature to support the above flow.

Step 1 - Data Strategy (Companies House API)

First, I needed to work out my data strategy. Thankfully, the Companies House API is excellent. It’s the UK’s public company registry - every limited company files here: incorporation date, director details, accounts type, charges, insolvency history. All publicly accessible via API, and all free.

I used 3 endpoints.

1 - Advanced Search

Returns a filtered list of companies by SIC code, status, and structure. This is the starting point to populate my dataset with a list of “active” UK plumbing companies from which everything else flows.

Documentation:

https://developer-specs.company-information.service.gov.uk/companies-house-public-data-api/reference/search/advanced-company-search

2 - Company Profile

Returns the full record for a single company - insolvency history, charges, accounts type, incorporation date, and status detail. I use this to pull disqualification signals before spending any credits on enrichment. If a company has insolvency history, active charges, or files dormant accounts, I want to know immediately.

Documentation:

https://developer-specs.company-information.service.gov.uk/companies-house-public-data-api/resources/companyprofile?v=latest

3 - Officers

Returns every officer on record - name, role, appointment date, and year of birth. The director is the person I need to reach. The officer record gives me their name (required for email finding), how long they’ve run the business, and an estimated age - my single most important acquisition signal. A 62-year-old sole director running a debt-free plumbing company for 20 years is exactly who I’m looking for.

Documentation:

https://developer-specs.company-information.service.gov.uk/companies-house-public-data-api/reference/officers/list

Step 2 - HTTP API Import

In Clay, I used “Import data from HTTP API” as the table source.

SIC code filtering

Companies House classifies every business by SIC code. Plumbing falls under 43220 (Plumbing, heat and air-conditioning installation). Passing sic_codes=43220 as a query parameter returns only plumbing companies. Full parameters: sic_codes=43220, company_status=active, size=50, start_index=0.

Authentication

The authentication method is HTTP Basic Auth. Clay requires credentials in Base64 format. I generated mine by running btoa(’API_KEY:’) in the browser console and pasting the result into Clay’s Authorization header as Basic [base64_value]. The same header works across every Companies House endpoint.

Pagination is handled via start_index - increment by 50 to pull beyond the first page. On the free tier the import caps at 50 rows; on a paid tier, one could loops through all pages to pull the full 52,000. No Clay credits are consumed on import.

Fields pulled from the Advanced Search endpoint:

• company_name

• company_number - the primary key for all subsequent enrichment calls

• company_status - filtered to active only

• date_of_creation - used to identify companies incorporated before 2010

• company_type - confirms limited company structure

• registered_office_address.locality - town

The resulting data that was pulled:

Step 3 - Company Profile Enrichment

For each imported company, I run a second Companies House call - this time against the company profile endpoint. Still free. Same API key, no third-party credits.

The three fields I pulled in:

• has_charges - flags outstanding debt against the business

• has_insolvency_history - immediate disqualifier if true

• accounts.last_accounts.type - the key size proxy. total-exemption-full is the sweet spot; micro-entity is too small; dormant is an immediate disqualifier

Running this before any paid enrichment means I’m not spending credits on companies I’m about to eliminate. Every subsequent step - Claygent, email finding, domain enrichment - only fires on rows that pass these checks first.

The raw API response:

The resulting data:

Step 4 - Hard Filter

Before spending a single credit, I eliminate bad rows using deterministic formula checks. Based on what I actually pulled from the API, the filters were:

• has_insolvency_history = true → DISQUALIFIED

• has_charges = true → flagged

• accounts_type is micro-entity, dormant, or group → DISQUALIFIED

• date_of_creation after 2010 → TOO YOUNG

Formula columns in Clay are free and instant. In my 50-row dataset, the majority were knocked out here - mostly on accounts type, with micro-entity being the most common reason. This step typically eliminates 30–50% of rows before any paid enrichment runs.

I built two output columns to make the results readable:

• qualification_status - outputs “qualified” or “disqualified” as a clean label

• Disqualification Reasons - outputs a human-readable reason for each eliminated row (too small, post-2015, dormant, etc.)

Both are free, auditable, and explainable - which matters when you’re trying to understand why a company was skipped.

The disqualification formula:

The resulting data:

Step 5 - Website Finding

Before running any contact enrichment, I need each company’s website domain. The reason is practical: email finding tools work significantly better when you give them a domain to search against. Asking “find me an email for John Smith at chiversplumbing.co.uk” returns far higher match rates than asking for “John Smith at Chivers Plumbing” - the domain acts as a lookup key.

Finding websites for UK plumbing SMEs is harder than it sounds. Standard enrichment providers often return the Companies House government page, rather than the actual company website. To get around this, I built a three-layer approach:

• Clay’s Company Domain enrichment - fast and cheap, works for well-indexed companies

• Domain validity check - a formula column that flags service.gov.uk, Yell, Checkatrade, or other directories as “Invalid”

• Claygent web research fallback - for Invalid or blank rows, Claygent searches Google with an explicit exclusion list for directories

The Claygent prompt instructs it to visit the result and return NULL if no direct company website exists - rather than a false positive. A Preferred Domain formula column then merges the best result from both sources into a single clean field, which gets passed into the email finding waterfall in the next step.

Claygent response:

Domain name results:

Step 6 - B2B/B2C Classification

Once I have a verified domain, Claygent visits each company website in web browsing mode and classifies the business as B2B, B2C, or Hybrid. This only runs on rows where Preferred Domain is not blank.

The classification matters for acquisition. B2B and Hybrid businesses can have more predictable revenue and longer-term contracts, where consumer-focused businesses may live on word of mouth - much harder to scale or hold together through an ownership transition.

The Claygent prompt, containing: Context, Objective, Instructions, Examples.

Claygent response:

Step 7 - Officers Enrichment

For qualified companies only, I pull director data from the Officers endpoint. The data pulled so far in the process tells me whether a business is worth buying. Officer data tells me whether the owner is likely to sell.

The profile I’m looking for: a single long-serving director who founded the business, is approaching retirement age, and has no obvious succession plan. A 62-year-old sole director who incorporated the company in 1998 and has been running it ever since. They built something valuable, they have no one to hand it to, and they haven’t seriously thought about an exit. That’s the opportunity.

The fields I pulled:

• Name - the director’s full name

• Officer Role - confirms director rather than secretary or other officer type

• Appointed On - how long they’ve been running the business

• DOB_Year - used to calculate estimated age

• Age - derived from DOB_Year; the primary acquisition signal

• Person Number - unique identifier for each officer

The raw API response:

The resulting data:

Step 8 - Lead Tiering / Scoring

A formula column combines B2B/B2C classification and director age into a simple tiering:

• HIGH: B2B or Hybrid + director aged 60+

• MEDIUM: B2B or Hybrid, director aged 50–59

• LOW: B2C, regardless of age

The logic is simple. The temptation is to build a weighted scoring model - points for accounts type, company age, director tenure. In practice, the two signals that matter most for acquisition fit are “company type” and “director age”. Everything else is noise.

Keeping it deterministic also means it’s fully auditable. For any row I can explain exactly why it scored HIGH - no black box or model drift. That matters when outreach decisions flow directly from the output.

HIGH tier contacts go into a priority sequence. MEDIUM tier go into a standard sequence. LOW tier will receive nothing - no copy generated, no credits spent, no sequencer push.

Step 9 - Contact Finding

This is the first step that real money. Everything up to this point - API calls, formula columns, Claygent classification - has been free or low cost. Contact finding is where third-party credits are consumed at scale, which is exactly why it sits here rather than earlier in the pipeline.

Run condition: only fires on directors linked to HIGH or MEDIUM tier companies. LOW tier directors are never enriched.

On the directors table, I run a work email waterfall. Clay tries each provider in sequence and stops on the first hit. No single provider has complete coverage of the UK SME market - chaining multiple sources increases the overall hit rate while only charging for providers actually queried.

The domain from Step 6 is the key input. Passing the director’s name alongside the company domain gives each provider a much tighter lookup target than the name alone - this is why website finding had to run first.

For directors with no work email found, personal email enrichment runs as a fallback. Many 60-year-old plumbing company owners use a Gmail address for business rather than a domain email - particularly those filing total-exemption-small accounts. Ignoring personal emails in this market means missing a meaningful proportion of the target list.

Step 10 - Email Verification

MillionVerifier runs only on rows where an email was actually found - no email, no verification credit spent.

The reason this step exists is deliverability. Sending cold email to unverified addresses damages sender reputation - bounces above a certain threshold trigger spam filters, which kills deliverability for every email sent from that domain.

MillionVerifier checks whether the address is real, active, and likely to accept mail.

This is the final gate before outreach. Everything downstream - copy generation, sequence push - only runs on contacts that have cleared every previous filter: qualified company, passing tier, found email, verified address.

Step 11 - Personalised Outreach Copy

[coming soon]

Step 12 - Outreach Sending Tool

[coming soon]

Part 3 covers how I’d improve what I built.

Click here for “part 3”.

Over the past year, I’ve met several individuals pursuing search funds – individuals raising capital to find, acquire, and operate a single owner-operated SME, below the radar of institutional buyers.

The most interesting part of these conversations was often how these individuals were going about finding a business that they want to buy.

Most were using a combination of outbound (telephone, email, direct mail) and inbound (press, online communities). But none had a systematic process behind it.

The Project

Inspired by the above conversations – and as part of a new project – I set myself a challenge: build a full acquisition pipeline from scratch. One that could:

• Pull the addressable market.

• Score businesses against acquisition criteria.

• Enrich the dataset with contact details.

• Write personalised outreach messages.

• Push everything to an outreach sending tool.

I’ve documented this project across three articles:

• Part 1 - what I planned to build + why.

• Part 2 - how I built it.

• Part 3 - how I’d improve it.

The Target

My target company should meet the following criteria:

• UK plumbing business.

• Generating £2M–£8M in revenue (£1M+ EBITDA).

• Run by a founding director in their late fifties or sixties.

• Debt-free, no insolvency, no broker involved.

The logic behind the above is straightforward. A business like this sits below the threshold a typical PE firm would touch - but generates enough EBITDA to make a search fund transaction viable.

At that age, the owner is approaching retirement with limited options: shut it down, hand it to a family member who doesn’t want it, or sell. That dynamic creates more receptivity to creative structures like seller financing.

Traditional deal sourcing relies on intermediaries who take a cut and only surface businesses already committed to selling. The goal here is to get upstream - reach owners before they’ve listed, before the valuation conversation has started, while seller financing is still on the table.

Technology Stack

The technology stack I use:

• Claude Code - increasingly where I spend my time. The reason: context engineering. Using CLAUDE.md project files and persistent skills means Claude retains full project context across sessions.

• WisprFlow - voice-to-text, speeds up everything.

• Clay.com - data orchestration, enrichment, outreach logic.

• Companies House API - the primary data source.

• Email finding - Prospeo, Apollo (waterfall: try each provider in sequence, stop on first hit). No single provider has complete coverage of the UK SME market, so chaining multiple sources meaningfully increases hit rate while only charging for providers actually queried.

• Phone finding - Surfe, Phantombuster (same waterfall logic).

• Email verification - MillionVerifier.

• Sending infrastructure - Smartlead / Instantly.

The Constraints

Two self-imposed constraints shaped the entire build.

• Timeframe - three hours. Hard stop. This prevents scope creep.

• Cost - no subscriptions beyond Claude. Everything had to run on free tiers.

The free Clay tier is the largest binding constraint. Three limitations worth flagging:

• 50 rows - the HTTP API import is capped. The full Companies House dataset for SIC 43220 runs to ~52,000 companies.

• Limited Claygent credits - web research, email finding waterfalls, and phone enrichment all consume credits. This restricted the creativity that I could apply in my build.

• No mobile phone enrichment - the mobile waterfall requires a paid plan. For a 60-year-old owner-operator, mobile is the right channel. This is a meaningful gap in practice.

Despite the above, a working end-to-end pipeline is possible.

Design Principles

Three principles governed every decision in the build:

• Data dependency - never run a step before its required inputs exist.

• Eliminate before enriching - filter bad rows before spending credits.

• Cheap before expensive - run free steps first; gate expensive steps behind them.

Part 2 covers how I actually built it - step by step.

Click here for “part 2”.

1. AI agents

2. Data

3. Automation

4. Sending

There’s also a dotted line from “pipeline generated” to “agent layer”, in order to complete the feedback loop, so that the end-to-end workflow gets better over time!

The below list is not exhaustive, and many of the tools mentioned can be substituted with alternatives that offer similar functionality, but I believe it reflects the core GTM stack today.

Rather than keeping these notes to myself, sharing in case helpful for others currently in the process of building out their infrastructure.

Automation

• Claude code - https://code.claude.com/

• Cursor - https://cursor.com/

Hosting

• Github - https://github.com/

• Trigger dev - https://trigger.dev/

• Railway - https://railway.com/

Productivity tools

• Wispr Flow - https://wisprflow.ai/

List building - databases

• Prospeo - https://prospeo.io/

• Apollo - http://apollo.io/

• Icypeas - https://www.icypeas.com/

• Blitz API - https://blitz-api.ai/

• Predict Leads - https://predictleads.com/

List building - scraping

• Zenrows - https://www.zenrows.com/

• Apify - https://apify.com/

List building - agents

• Exa - https://exa.ai/

• Parallel - https://parallel.ai/

List building - lookalikes

• Ocean - http://ocean.io/

• Discolike - https://discolike.com/

Alternative data sources

• Trigify - https://www.trigify.io/

• RB2B - https://www.rb2b.com/

Data validation

• Millionverifier - https://www.millionverifier.com/

• Debounce - https://debounce.com/

Databases

• Supabase - https://supabase.com/

Data enrichment / orchestration + more!

• Clay - https://www.clay.com/

E-mail automation tools

• Instantly - https://instantly.ai/

• Smartlead - https://www.smartlead.ai/

Pitch decks

• Gamma - https://gamma.app/

A SIMPLE EXAMPLE

Imagine a business that sources deals for private equity firms. The offer is straightforward: identify businesses that may be looking to sell, gauge interest, and book meetings.

Perhaps the fee for this service is $10,000 per month, and clients stay for 12 months.

-> LTV = $120,000

Now imagine that acquiring this customer relationship costs $30,000 - in conferences, meetings, travel, and sales calls.

-> LTV:CAC = 120,000 / 30,000 = 4

A DECISION POINT

That’s a healthy ratio, but it's interesting to consider the impact that business model design can have on the same ratio.

Instead of focusing only on acquiring more customers for the same core service, the business could layer on complementary offerings:

1. Due diligence support - evaluate whether a target company is worth acquiring

2. Value creation planning - design a roadmap for post-acquisition growth

3. Recruitment support - fill key roles to execute the value creation plan

With each additional service sold into the same account, LTV increases - without necessarily increasing CAC.

The strategic question becomes:

• Does one acquire more customers for a single offering?

• Or is it better to identify additional problems to solve for existing customers?

THE TAKEAWAY

Many businesses have this choice. Once a business has solved one problem for a customer, it’s interesting to ask the question:

“What’s the next problem they’ll need solved?”

Answering that question well can be the difference between a business with a good CAC:LTV ratio, and one with an exceptional one.

Every few months, someone declares that “cold outreach is dead.” I think the opposite is true - it can remain very effective as a channel, but as inboxes have become more saturated, with the abundance of cheap data, and automated sending infrastructure, the playbook has evolved.

Outbound still works incredibly well when executed thoughtfully.

The ideal scenario is simple to describe, but hard to execute: reach the right person, at the right time, with the right offer, using the right message, and through the right channel.

When you get those variables right, you have a predictable pipeline engine. When you don’t, you're spamming, with no surprise that reply rates approach 0...not to mention you risk a fine under the US CAN-SPAM Act!

Most outbound campaigns start with a basic firmographic filter - for example, a private equity firm, looking to acquire a new business, might pull a list of all businesses with $10M+ EBITDA in a particular sector, and then enrich the dataset with decision-makers.

But how receptive would a generic list of accounts like this, be to a cold offer to enter buyout discussions? Possible, but not likely.

ENTER “SIGNAL DATA”

The real leverage comes from layering firmographic data with signal data - information that helps you predict which prospects are most likely to be receptive right now.

Imagine enriching that same private equity target list with signals like:

• Owner age approaching retirement age

• Owner posted on social media about fatigue

• Company has seen recent high employee turnover

• Owner's children in corporate careers (less likelihood of succession?)

• Owner listed business on brokerage platforms for >6 months

Thanks to tools like Clay, Trigify.io, and Momentum.io, obtaining the above data is no longer a manual exercise. You can systematically capture, enrich, and score prospects against these signals.

THE IMPACT

Suddenly, the private equity firm is not just sending cold emails into the dark, but having perfectly timed conversations with tired business owners, already exploring exit options.

And because this is top-of-funnel work, small improvements like doubling message reply rates from 1% to 2%, assuming downstream conversion rates hold, literally doubles the number of closed deals.

...that's certainly enough to pay for the extra cost of "signal data"!

Teams embracing this new approach can scale their impact without increasing headcount, whilst also reducing the risk of brand damage that can come from blasting irrelevant messages to a unrefined prospect lists!

This morning, a friend shared an article about a Y-Combinator company that raised a large seed round. The piece highlighted “impressive traction” on little headcount. I don’t know much about the company itself, but it prompted reflections on how the tech press reports early stage funding rounds.

Every few weeks, the same story appears: headlines announce millions raised, valuations are presented with excitement, and inevitably there’s mention of ARR.

At first glance, the numbers are impressive, designed to give the impression of a company on a rocket ship. But step back, and it’s worth asking who benefits from these narratives, and how much reflects business reality. The company, investors, and media all have “skin in the game” when it comes to sustaining a valuation story - and behind the scenes, things often look very different.

HOW IS ARR CALCULATED

Having been involved in multiple early stage companies myself, and knowing many entrepreneurs who have raised funding, I know that ARR figures often tell you far less than they suggest. The real questions - almost always missing from press releases - are the ones that determine whether a business is strong:

1. How is monthly recurring revenue calculated? Is it based on signed contracts, invoices paid, or verbal commitments?

2. What does the gross margin look like? $1M revenue is meaningless if it costs $2M to acquire customers and $3M to deliver.

3. What about retention? If 80% percent of customers churn after month 1, ARR is misleading.

4. What is LTV:CAC? That ratio tells you about sustainability.

5. Are refunds, discounts, or chargebacks included in the revenue figure?

Strip away the hype, and reported ARR can prove fragile.

ANOTHER TYPE OF COMPANY

Some of the most successful entrepreneurs I know, operate very differently. They run profitable businesses, generating consistent cash flow, but you won’t find them on TechCrunch. They don’t hype themselves on social media, and in some cases, their businesses don’t even have websites!

The irony is that for most customers, knowing their vendors' ARR is not a selling point. What matters is the delivery of a customer outcome, not fundraising milestones.

This all makes sense once you realise who ARR announcements are really for. Customers are not the target. The audience is investors. The risk, however, is that founders shape their companies around what appeals to investors, rather than what delights users.

HYPE IS NOT ALL BAD

I’m not suggesting that all hype is bad. If it attracts investment and talent, it can become a self-fulfilling prophecy. But it’s worth maintaining a scepticism when reading about a company’s ARR.

Ask why the numbers are being shared, who stands to benefit, and what might be missing. Until you have visibility on the actual P&L (ideally company bank statements!), it’s wise to treat such claims with caution.

At the end of the day, the best businesses don’t always shout the loudest. They work - whether or not TechCrunch writes about them.

• Credibility / track record

• Proven standards of service delivery

• Access to internal networks, resources, and frameworks

• Third-party validation (brand acts a “seal of approval”)

• Reputation at stake (“more to lose”)

When someone operates under a powerful brand, they borrow accumulated equity.

It is rational, from a buyer’s perspective, to place higher value on goods or services bought under a brand, than without, as the former carries less "risk" than the latter.

When, as consumers, we purchase a product or service, we calculate our willingness to pay on the basis of "expected value"; a function of the perceived outcome that the purchase will drive, and the probability of said outcome being realised.

In this case, the executive had not operated outside of a "strong brand" environment - and was surprised by the contrast.

The absence of a strong brand has significant adverse commercial implications:

• Lower margins (discounts required to overcome doubt)

• Longer sales cycles (trust-building needed before closing)

• Higher friction for the sale (value must be delivered upfront)
  

What can an individual, operating without a strong brand, do to offset the above challenges?

Some strategies include:

1. Lean into networks that already know, like, and trust the individual

2. Partner with established organisations to “inherit” brand equity

3. Reduce the trust required by offering outcome-based guarantees

4. Leading with value to reduce the trust gap

5. Long term, build brand equity!

Another interesting takeaway to the above, relates to "target hiring pools" for early stage companies, without yet a strong brand, recruiting into revenue-generating functions, but that's a story for another post. :)

Across performance channels, whether that be paid, organic, owned or SEO etc, many of the legacy workflows can be re-designed today, through taking a more data-driven, and less labour intensive, approach.

Historically, these companies run at 20-40% net margins, but through taking a tech-driven approach, the profitability could be driven higher.

My notes on how I would approach such a venture, whether as a standalone business, or as a new business line, within an established organisation:

-> Picking the industry

I'd start by selecting a vertical where there is an already well-established revenue pool for performance marketing solutions (e.g. finance, home services, legal, education).

This way, I would reduce the risk around whether there's market demand for the value proposition, and instead seek to build a better solution than incumbents.

Over time, I would identify industries which are underserved today by performance marketing solutions (i.e. no or little existing revenue pool), but I wouldn’t start there.

-> Pricing

To create a compelling value proposition, I’d price on “outcomes generated”, whether that be leads, booked appointments, live transfers, signed contracts or even sales.

If I wasn’t initially confident in committing to the delivery of an outcome very close to the end sale (to offer “risk free acquisition”), I would start by committing to outcomes further up the funnel, with the intention over time, of moving closer to the sale.

-> Channels

To build momentum, and to deliver outcomes fast, I would start with paid traffic, and choose to work with only one channel between social, search, native, and display etc.

The intention would be to fully automate and dominate this channel, before over time, adopting other marketing channels.

-> Technology / data

When building out the technology, I would start with the “quick wins”, and then automate the entire end-to-end workflow over time.

For example, with “social”, I'd start by automating the research process, before automating the creative brief itself, before moving onto the creative production, and then finally the media buying itself.

With more clients, and experience, would come more data. I would leverage this data to complete the loop in evaluating the success of the technology being built, and over time strive to drive more volume (“results”) of a higher quality, at lower CPAs.

-> Branding

In whatever vertical is chosen, I would build a trusted consumer-facing brand in the marketplace. This would develop brand equity, and overtime lower the overall cost of acquisition, whilst acting as a competitive moat.

If anyone is building solutions around the above, always enjoy discussing these types of topics...feel free to reach out! :)

A while back, I was hired as a GTM consultant for a series A start-up.

On paper, things looked promising.

But within several days of working together, I arrived at 2 conclusions:

1. There was no real demand for their product.

2. Engineering was building features based on intuition (not customer demand).

Meanwhile, the internally held view was that the start-up's challenges were down to a "sales and marketing" issue..

....rather than something structural related to the business model itself.

Hence, the reason for my hiring. 🤦

At the time, when I suggested product changes (rooted in customer discovery), and tweaks to website messaging, to better align with market demand, I was advised that:

"We can’t make these changes. Our investors have given us a mandate to build what’s on the site."

Over the years, I've noticed this sequence of events occurs more often than I would have originally expected.

A new start-up raises capital after some early traction.

But as the team dive deeper into customer discovery, and product development, it transpires that something in the model doesn't work.

This could be that the customer segment is wrong, or that the product isn’t solving a real pain point...or that there's no real demand.

But rather than pivot, the team sticks to the original plan, as the narrative that was "sold" to raise that initial round of capital, becomes a binding constraint.

Sometimes, this constraint is explicit - the capital came with strings.

But other times, the constraint could be internalised - the team mistakes investor belief for product-market fit.

But notably - markets, not fundraising, defines PMF.

So in this case, the start-up continues, even potentially raising another round of funding.

But eventually the business is not able to escape its structural challenges, and breaks down.

WHAT ARE SOME IDEAS TO MITIGATE THE ABOVE?

→ Independent thought

Establishing independent advisors who can provide fresh perspective to the internal narrative.

Creating space for a diverse range of opinions.

Trying to avoid "groupthink".

→ Run small, parallel tests.

Rather than pivoting abruptly, layering in controlled experiments can be helpful.

If traction then appears in any of those parallel tests, this can be used to secure investor buy-in on any new direction.

Investors are ultimately aligned with value creation.

→ Delay raising, if possible.

Early capital can bring pressure, and another stakeholder to manage.

With modern tools and channels, it’s often possible to validate demand without external funding - or with far less.

→ Define what counts as real traction.

MOUs, soft interest, or friends supporting?

It can be dangerous to count these as market demand.

Paying customers, engagement, usage?

These are more real.

1. STARTING WITH BUSINESS FUNDAMENTALS

Before writing a single ad, we clarified a few basics:

→ Who are we targeting?
→ What pain point are we solving?
→ What product or offer are we promoting?
→ What specific action do we want the audience to take?

2. STUDYING THE COMPETITION

Rather than starting from scratch, we looked at what others in the space were already doing.

→ We built a list of competitors using Google, LLMs, and business databases.

→ We reviewed their live ads using public ad libraries (e.g. Facebook Ads Library, Google Ad Transparency etc) and ad spy tools.

→ We looked for signs of success—it's rare for platforms to give performance data directly, but proxies exist:

A. Are they running lots of ads?
B. Are they testing multiple variations of the same ad?
C. Have the ads been running for a long time?

3. REVERSE ENGINEERING THE WINNERS

Once we’d identified “working” ads, we broke them down to understand what might be driving performance.

→ What audience are they speaking to?
→ What’s the copy style?
→ What kind of creative are they using (video, image etc)?
→ What’s the CTA?
→ Any patterns in headline structure?

The goal wasn’t to copy, but to understand the constituent drivers of success so that we could model.

4. BUILDING AND LAUNCHING

Using what we’d learned, we built several variations of our own.

From there, we launched with a small budget, and will be paying attention to early data signals, so that we can then iterate based on performance.

I hope this short (and somewhat oversimplified!) guide helps another business exploring paid media.

Perplexity can now draw direct insights from SEC data / filings.

This is a big time-saver for anyone who relies on company intelligence:

• Investors conducting due diligence.

• Companies understanding their competitors.

• Consultants completing a market landscape.

• Entrepreneurs launching a new company.

How about UK datasets?

Companies House is a rich source of information on private companies.

It's only a matter of time before such datasets are also leveraged.

I know a number of UK-based search fund entrepreneurs:

Companies House forms a core part of their "prospect building".

Right now, that process is manual and slow.

I'm sure many will be excited by the workflow improvements as the technology evolves!

Here’s a link to the full HBR article:

https://hbr.org/2025/04/how-people-are-really-using-gen-ai-in-2025

→ Criminal innovation

Multimodal models make it easier than ever to generate realistic synthetic content — voices, faces, full conversations.

→ Upgraded old tactics

Phishing, social engineering, payment fraud — now layered with AI that adds nuance, believability, and scale.

→ Exploitation of trust

Familiar voices. Known faces. AI weaponises social signals we instinctively trust — urgency, authority, emotional cues.

So what can be done?

The report outlines the potential to use "AI to fight AI".

For example, using AI to flag deepfakes, detect phishing emails, trace anomalies in real time, and strengthen threat detection across large datasets.

Here’s a link to the full article for anyone interested to read:

https://cetas.turing.ac.uk/publications/ai-and-serious-online-crime

1. Capital

Some markets still require significant investment to enter—whether that's building physical infrastructure, acquiring licenses, or funding deep research. Not everyone can raise the capital to play.

2. Network effects

A product that becomes more valuable as more people use it is harder to replicate. Communities, marketplaces, platforms—they all benefit from this dynamic.

3. Proprietary data

As LLM-powered applications and agents become easier to build, having one's own unique dataset offers an advantage.

4. Distribution

Having efficient mechanisms to distribute products and services to customer segments can offer some moat. This can include having a pre-existing customer base with high switching costs.

5. Brand

A strong brand builds trust, loyalty, and differentiation—especially in commoditised markets where everything looks similar.

6. Regulatory

If operating in an industry where licensing, compliance, or relationships are non-trivial (e.g. fintech, healthcare, defense), that can serve as a durable barrier to entry.

The question of durable competitive advantage is interesting to ask from the different viewpoints of investors, operating executives and entrepreneurs.

→ Outbound.

The days of blasting cold emails to a list, and expecting results, are long gone. The channel works best with a differentiated offer, targeted data, strong unit economics, sales assets like case studies — and ideally, some brand awareness. Early-stage teams are still in the process of building all of that.

→ SEO.

Often positioned as a “free” growth channel — in reality, it rarely is. Ranking well in search takes time, content resources, backlinks, and a clear strategy. And SEO is slow to compound. If you need customers today, SEO won’t help. Worse still, if there’s uncertainty around one's target customer (which there often is early on), there's a risk of driving traffic that turns out later to be entirely the wrong audience.

→ Social media marketing.

Building an audience takes time — sometimes a lot of it. And just like SEO, there’s always the risk of investing months into building the wrong audience. Early-stage companies are still figuring out their ICP, offer, and positioning. If those things change — and they usually do — the audience might not move with any re-positioning. Inbound content works best when it speaks directly to clear customer pain points. But if one doesn't know exactly who the customer is yet, knowing what to post becomes a challenge.

→ Paid acquistion.

This might seem like the quickest route to traction — and technically, it can generate traffic fast. But doing paid well is expensive. Ad platforms operate as auction mechanisms, and competition is intense. Creative assets fatigue quickly. And unless one's unit economics are already in a good place, paid channels can end up burning budget fast.

-------------

Challenges are compounded by early stage companies having sub-optimal unit economics. There’s uncertainty around the ideal customer, positioning is still being tested, the sales process is rough, pricing models are evolving, and the product itself might still have major gaps.

As the saying goes: the company that can afford to pay the most to acquire a customer usually wins. But in the early days, most startups are a long way from being that company.

So given the above, what are some ideas on how to best approach early stage distribution?