Charley McDonnell

A brand can scale revenue while losing money on every customer, and no amount of campaign optimisation fixes that.

• McDonnell distinguishes media buying, which optimises within the numbers it is given, from growth strategy, which can conclude that optimisation is not the answer.
• The proposed fixes are to raise lifetime value through retention, bundles or upsells, or lower acquisition cost through better conversion rates and more efficient creative.

A brand doing £180,000 a month in revenue looked healthy on paper. Its true acquisition cost was £67 against a lifetime value of £61, so it lost £6 on every customer it acquired, at scale, while its Meta campaigns kept hitting their ROAS targets.

Read original

Brands may blame algorithm changes, competition or market saturation, but McDonnell points to problems inside the business: the scrappy approach, manual daily optimisation and gut-feel decisions that built the brand cannot support a larger one.

• Breaking through means rebuilding, not doing more of the same: invest in creative production capacity before it is needed, simplify campaign structures instead of adding complexity, and measure what matters rather than what is easy to measure.
• The constraint shifts from execution to systems.

The proposed changes are concrete: roughly triple creative production velocity, systematise manual optimisation and replace gut-feel calls with data-driven frameworks. McDonnell argues that brands often keep optimising campaigns when they need to rebuild the operating model.

Read original

Statistical modelled data, view-through attribution and delayed reporting make platform figures unreliable on their own.

• The method is to compare several sources — a third-party tool’s triple attribution, Shopify revenue and Meta’s own data — and make decisions from the comparison.
• The rule is blunt: never look at data in a silo or rely solely on Meta’s data.

For budget decisions, compare Triple Whale with Shopify revenue and Meta’s reporting; use all three rather than trusting any single dashboard.

Read original

An agent, in this account, is not a chatbot that answers questions but a system that executes tasks without supervision.

• Brands still running manual checks, spreadsheet updates and business-hours monitoring risk falling behind a smaller group deploying agents that never sleep or miss a threshold.
• This is a forward-looking claim, not a reported result: McDonnell describes their own early experimentation and says the technology already makes the scenario possible, but provides no reliability data, failure rates or named client.
• Ask for those before assessing deployment risk.

A best-selling product hits low inventory at 2am.

• The agent notices, cross-references supplier lead times, calculates from sales velocity that the brand will stock out in four days, generates a purchase order and sends it to the supplier.
• The owner wakes to a single message saying the reorder is placed and arriving Tuesday.

Here McDonnell moves beyond advertising and its surrounding business model into back-office operations, presenting automation as the next competitive divide. The claim is illustrated rather than tested; ask for evidence from an engagement.

Read original

The stated position is that these tools are not usually recommended because they have no machine learning and cannot act on the data themselves, but the integration would be worth examining if it proves real. The post is a heads-up, not confirmation; ask for the announcement and technical details about what the integration would expose.

The same brand’s measurement stack could use NorthBeam, Triple Whale or GA once the API integration lands, but McDonnell still expects a human to interpret the output rather than the tool making its own decisions.

McDonnell is willing to revisit a stated view when new information emerges: a category they usually do not recommend becomes worth examining once the access route changes, without any claim that the tools themselves produce insight.

Read original

Read it first: it sets the threshold McDonnell applies before discussing ads, and the revenue figure makes the point checkable.

McDonnell opens client conversations with one question: what is your customer acquisition cost and customer lifetime value?

• They write that about 60% of brands cannot answer it, even though they know their ROAS, conversion rate, and daily spend.
• To make the point concrete, they describe a brand doing £180,000 a month in revenue whose true CAC was £67 against an LTV of £61, losing £6 on every customer acquired.
• Meta campaigns were hitting ROAS targets throughout; the problem was the business model, not the ad account.
• McDonnell says the fix must either raise LTV through retention, bundles, or upsells, or lower CAC through conversion rates and creative efficiency.
• Media buying optimises what it is given; growth strategy determines when optimisation is not the answer.

For brands planning a campaign

For a DTC brand scaling paid social, expect unit-economics scrutiny before a media plan.

• Have recent CAC, LTV, retention, and margin figures ready.
• If LTV is below CAC, more creative testing will not fix it; brief retention, bundling, or conversion work first.

Start here if the focus is scale: it reframes the problem from tactics to operating model.

McDonnell describes a wall most ecommerce brands hit at roughly £80,000–£100,000 a month in revenue: performance gets harder to sustain, CPA creeps up, and what worked last quarter stops working.

• The common instinct is to blame algorithm changes, competition, or saturation, but they argue the ceiling is internal: the scrappy systems that built an £80K-per-month brand cannot support a £200K-per-month one.
• The changes they name are unglamorous: creative production velocity has to roughly triple, daily manual optimisation has to become systematic, gut-feel decisions have to become data-driven frameworks, campaign structures should be simplified rather than expanded, and measurement should shift to what matters rather than what is easy.
• Most brands, they conclude, keep optimising campaigns when they should be rebuilding operations.

For brands planning a campaign

A scaling brief should ask about creative production capacity, measurement design, and who owns daily decisions—not just account structure. Use it as a diagnostic checklist before proposing a rebuild.

It is a compact, quotable statement of method for anyone deciding which numbers to trust when three dashboards disagree.

McDonnell identifies attribution as the under-discussed problem on a large Meta account, with three moving parts: statistical modelled data, view-through attribution, and delayed reporting.

• They recommend using a tool such as Triple Whale and comparing triple attribution, Shopify revenue, and Meta's own numbers side by side.
• Their rule: never look at data in a silo or rely solely on Meta's data.

For brands planning a campaign

For high-spend accounts, agree in advance which revenue source will break a tie and which third-party tool to use, rather than reconciling numbers after the fact. Reporting should compare platform data with an independent revenue figure.

Read it if the focus is back-office automation rather than advertising, and treat the scenario as an illustration, not a tested case.

McDonnell compares the current state of business operations to 2007, when the iPhone changed what "having a phone" meant, and argues that autonomous AI agents have just moved from experimental to functional.

• They report experimenting with OpenClaw and describe a scenario: a bestselling product hits low inventory at 2am; an agent notices, cross-references supplier lead times, calculates a four-day stockout from sales velocity, generates a purchase order, and sends it to the supplier, so the owner wakes to "reorder placed, arriving Tuesday."
• They contrast this with manual checks, spreadsheet updates, and business-hours monitoring, and predict the efficiency gap will become near-impossible to close within 18 months.
• Before relying on the scenario, ask for reliability data and failure rates; no named client is given.

For brands planning a campaign

Use this as a starting point for mapping repetitive tasks such as reorder triggers and monitoring. Before automating anything that commits spend, ask what evidence supports its reliability; treat the 18-month prediction as a forecast, not a fact.

McDonnell flags an expected Meta rollout of API integrations with third-party analytics tools, naming NorthBeam, Triple Whale, and Google Analytics.

• Their position is notable: they usually do not recommend these tools because they "can't actually learn anything from the data" in the absence of machine learning, but call the integration worth a look if it happens.
• The post is a heads-up, not a confirmed change.

For brands planning a campaign

If a brand's measurement stack depends on these tools, expect skepticism that they generate insight on their own and a preference for human interpretation. Verify current platform status before planning around the integration.

The post names statistical modelled data, view-through attribution, and delayed reporting as moving parts that make platform numbers unreliable on their own. It recommends using a tool such as Triple Whale and comparing Triple Attribution, Shopify revenue, and Meta data, rather than looking at any one source in isolation.

Triangulate platform, store, and third-party attribution before making budget decisions; avoid siloed reporting.

For brands planning a campaign

For a DTC brand scaling paid social, ask for a measurement plan that reconciles Meta, Shopify, and a third-party attribution source. Treat single-platform ROAS as insufficient for a high-spend account.

Examples: Source 1

A worked example describes a brand doing £180,000 per month with a true CAC of £67 and an LTV of £61, meaning it lost £6 on every customer while Meta ROAS still looked acceptable. The post distinguishes media buying, which optimises within the numbers given, from growth strategy, which can say that optimisation is not the answer.

Ask for CAC and LTV before discussing channel tactics; separate business-model problems from campaign problems.

For brands planning a campaign

Make unit economics a gating question for a campaign aimed at scaling media spend. A media plan should not scale volume unless the brand can show whether acquisition is profitable after retention and repeat purchase are included.

Examples: Source 1

The cause is described as internal: the scrappy systems, manual optimisation, and gut-feel decisions that built the brand cannot support a larger one. The post says breakthrough brands rebuild operations, invest in creative production infrastructure before they need it, simplify campaign structures, and measure what matters rather than what is easy to measure.

The ceiling is presented as common, not universal; the exact revenue range is a general pattern rather than a fixed rule.

For brands planning a campaign

For a brand near that revenue range, assess whether the real need is media buying or an operating-model rebuild. The campaign scope should cover creative production capacity, measurement, and decision-making systems, not only ad account tactics.

Examples: Source 1

The post says these tools are not usually recommended because they cannot learn from the data and have no machine learning, but adds that the integration would be worth a look if it is real. The observation separates tool access from analytical capability.

Ask whether the integration is live and what it enables; Charley McDonnell's posts do not confirm that it is functioning.

For brands planning a campaign

When a brand or agency evaluates an attribution stack, ask what decisions the data changes and whether the tool can act on or learn from it, not only whether it connects to Meta. Don't present an announced integration as an available control.

Examples: Source 1

The post describes experimenting with OpenClaw and gives a concrete operations example: a best-selling product hits low inventory at 2am, an agent checks supplier lead times, calculates a stockout in four days, generates a purchase order, and sends it to the supplier. The claim is that most ecommerce brands still rely on manual checks, spreadsheets, and business-hours monitoring, while a smaller group deploys agents that never sleep or miss a threshold.

Ask for production results and verified savings; the post describes an early experiment with a named tool, not production results or guaranteed savings.

For brands planning a campaign

For operations-heavy DTC campaigns, test whether automation is a credible part of the scope. Ask for the specific workflow, thresholds, approval steps, and supplier integrations rather than accepting generic AI claims.

Examples: Source 1

Their writing uses specific figures, numbered operational steps, and direct comparisons: media buyer versus growth strategist, or BlackBerry versus iPhone.

• Each post closes with a direct question to the reader.
• The recurring subjects are paid-social measurement, unit economics, operating systems, and automation.

Diagnose the business model, measurement, or operating system before optimising campaigns; use concrete numbers and choices rather than general channel advice.

For brands planning a campaign

A brand evaluating Charley McDonnell should expect a diagnostic conversation about unit economics, systems, and measurement before tactics. The useful test is whether that diagnosis produces a specific next step for the brand's own numbers.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5

Their posts discuss attribution problems on accounts spending $300,000 per month on Meta, including statistical modelled data, view-through attribution and delayed reporting. They also note that Meta was preparing to roll out API integrations with third-party analytics tools.

View example

View example

They also say they usually avoid recommending such tools because they cannot learn from the data.

View example

View example

They illustrate an inventory reorder process it could run.

View example

Lagace's two-part walkthrough of an 11-step bottom-up forecast for M&Ms, Purdy & Figg, and WildBird covers the same ground McDonnell does when asking for acquisition cost, lifetime value, and margin targets before any campaign decision.

Lagace teaches a declared system, laid out step by step with inputs, targets, and attribution windows spelled out for people building the model. McDonnell writes diagnostic posts that challenge one assumption at a time and leave readers to act on it, reaching beyond the ad account into operations and automation.

Read original · Original 2

Reeto's posts on blended new-customer CAC, predictive CAPI, and incremental attribution testing address the same problem McDonnell identifies in a $300,000-a-month account: modelled data, view-through attribution, and delayed reporting make single-dashboard numbers unreliable.

Reeto writes from a measurement vendor's position, reporting A/B test results and product capabilities, often to start a conversation about the tool. McDonnell writes as an outside strategist, recommends comparing third-party tools with store revenue, and makes clear that a tool connecting to Meta is not the same as one that can learn from the data.

Read original · Original 2

Shakoor's description of an account built around one winning creative, and the need to keep a pipeline of tests running, echoes McDonnell's point that the operating model—not the campaign—is what breaks as spend grows.

Shakoor sticks to creative testing and account structure, writing short, conversational posts that often read as personal career reflections. McDonnell's posts range from unit economics and measurement to business operations and AI agents, each built around a worked example or set of figures.

Read original · Original 2

Lagace sets out an 11-step bottom-up forecast built from the brand's own history, not a desired target: goals first (exit plans, topline revenue, profitability), then a model assembled from aMER target, first-time and returning AOV, customer file size, recently acquired repeat rate, active non-recent repeat rates, churn, gross margin, cost of delivery and fixed costs.

• Comparing the output with what the founder wants exposes the real gap—in the worked example, topline is ahead of goal but profitability is well behind.
• The model then converts the business aMER target into a prospecting target across Meta, Google, TikTok and Snapchat, translates it into a 7-day click target bridged by a delayed-attribution multiplier, and sets SKU-level targets so each product delivers the same first-purchase contribution margin instead of letting one high-margin product hide a weak one.
• The posts name M&Ms, Purdy & Figg and WildBird as brands that have used the system.

This is a modelling and expectation-setting engagement that needs live access to Shopify, Klaviyo and the ad platforms, so it suits a brand with clean data and a planning decision ahead.

• The two posts cover steps 1–4 and 6–9 only.
• Because the work focuses on forecasts and targets, a brand whose actual problem is that acquisition cost exceeds customer value, or whose creative output cannot support the plan, still needs the unit-economics and operating-model work McDonnell describes.
• Ask whether either person offers a sponsorship or booking format before making a paid endorsement decision.

Examples: Source 1 · Source 2

The reviewed work at Angler AI treats blended new customer CAC as the north star instead of platform-reported figures.

• One post describes a brand spending seven figures a month across Meta, Google and affiliate, with more than half of daily orders from returning customers.
• Platform-reported and blended new customer CAC kept rising even after customer and CRM lists were synced to Meta hourly and excluded from prospecting.
• The test switched from an advanced CAPI provider to Angler's predictive CAPI for Meta while holding Google and affiliate spend within roughly five percent; the reported early result was blended new customer CAC falling as Meta spend rose.
• A second post explains the testing method—a two-cell A/B test against the client's existing setup (Elevar, Blotout, Triple Whale Sonar, Shopify, or CDPs such as Segment)—and uses Meta's incremental attribution reporting to check whether view-through lift is genuinely incremental.
• If a brand has reconciled its economics but cannot trust what its dashboards say is working, this is the comparison to run.

The approach depends on a predictive CAPI product, conversion-API or CDP infrastructure, and a testing budget.

• That puts it out of reach for smaller accounts and adds a vendor and experimental layer to the stack.
• It optimises acquisition efficiency within a working business model; it does not examine whether the model works, rebuild creative production capacity or answer the CAC-versus-LTV question McDonnell starts with.
• If the constraint is the operating model rather than attribution, this measurement work will not fix the underlying problem.
• Ask whether sponsorship or commercial availability is offered.

Examples: Source 1 · Source 2

The post describes a familiar pattern: one ad works, spend scales from $1K to $10K, then CPA climbs, ROAS drops and sales slow.

• The immediate explanation is that Meta is getting expensive.
• Shakoor argues the problem started earlier: the account was built around one hook, one angle and one message.
• The discipline is to keep testing different hooks, angles, customer problems and formats so a replacement is already in progress when the current winner tires.
• For a campaign where creative fatigue and testing throughput are the named issues, this is a more direct match than a diagnostic interview.

The posts cover creative-testing discipline and a career reflection after a pause; they do not set out CAC and LTV, contribution margin, forecast targets, attribution reconciliation or operating-system changes.

• Choosing Shakoor narrows the engagement to creative volume and testing, so the financial and measurement analysis McDonnell or Lagace brings would still be needed to assess whether scaling is profitable.
• The comparison draws on two short posts.
• Ask for client results, the scale of accounts worked and whether a sponsorship format is available.

Examples: Source 1 · Source 2