The $3 Million Warning Light

1The business and its two very different faces

The business was a global managed and cloud services provider with approximately $300M CAD in annual revenue, operating as a subsidiary of a larger parent organisation. Its customer base of approximately 13,500 accounts was, on the surface, a single business. On examination, it was two entirely different businesses that happened to share infrastructure, systems, and a brand.

Exhibit · Case outcomeAnonymised

PPC on the unprofitable segment$3.0M / yr

Attributable revenue it generated< $0.4M

New pipeline after reallocation$1.05M / mo

The small customer segment, approximately 12,500 accounts

The majority of the customer base consisted of individuals and very small businesses purchasing low-cost hosted server capacity. Monthly recurring revenue for this segment was typically a few dollars, transacted by credit card with no sales involvement. The purchase journey was short: an ad click, a landing page, a free trial offer, an automated sign-up. Conversion happened entirely online.

The typical customer in this segment was a technically minded individual, often a young person working on a personal project, a small website, or an online game, who needed hosting capacity and basic support. They used the telephone support service extensively, frequently for questions that were essentially educational: how to configure a server, how to build certain website features, how to resolve basic technical problems. The support cost per customer was high relative to their revenue contribution.

Inside the business, the belief was that these customers represented the engine of future growth. The logic was appealing: acquire customers cheaply at low MRR, support them as they develop, and as their projects and businesses grow, they grow with us. Low CAC today, compounding CLV over time. The bedroom developer becomes a successful business. The personal project becomes a commercial platform. The small account becomes a significant one. The problem was that nobody had tested this hypothesis with data. It had been elevated, over time, from an assumption to a conviction to an article of faith, and because it was dressed in plausible logic, it was never subjected to the scrutiny that any capital allocation decision of $3M annually deserved.

The growth segment, approximately 1,000 account

At the other end of the customer base were approximately 1,000 accounts with a fundamentally different commercial profile. These were typically retail e-commerce businesses, predominantly online retailers, with a particular concentration in the fashion sector, whose entry MRR was around $1,100 per month, and who demonstrated a consistent pattern of expanding their spend as their own businesses grew. Their hosting and managed services needs scaled directly with their commercial success.

The most valuable accounts in this segment were spending many thousands of dollars per month on complex managed services, including one major consumer technology platform operating at approximately $300K in annual ARR. These customers had not arrived via a PPC click. They had been won through direct sales effort over sales cycles measured in months, supported by technical consultancy, commercial negotiation, and the kind of sustained relationship-building that no automated ad campaign could replicate.

This segment was profitable. It expanded. It referred to other businesses. Its support interactions were substantive rather than educational. And it was almost entirely invisible to the marketing function, because marketing's metrics were built around the activity that generated the small customer segment, the clicks, the form fills, the inbound calls from the lower end of the market.

Table 1The two businesses compared. The 13,500-account base contained two fundamentally different customer models with almost nothing in common except the infrastructure they sat on.

Metric	Small customer segment (~12,500 accounts)	Growth segment (~1,000 accounts)
Number of accounts	~12,500	~1,000
Typical customer	Individual or micro-business. Personal project, small website, bedroom developer.	Retail e-commerce business. Online retailer, fashion, fast-growing digital commerce.
Entry MRR	$5 to $10 per month, by credit card	$1,100+ per month
Acquisition route	PPC click → landing page → online sign-up. No sales involvement.	Direct sales team. Multi-month sales cycle. Technical consultation required.
Support usage	High. Frequently educational, how to configure, how to build, how to fix.	Substantive. Technical and commercial. Lower frequency, higher complexity.
Expansion pattern	Rare. Below $1,000 MRR, customers did not grow.	Consistent. E-commerce growth drove hosting and services growth directly.
Average tenure	6 to 9 months before going inactive.	Effectively indefinite. Migrating an e-commerce platform is operationally prohibitive.
Profitable?	No. High support cost, no growth, low MRR.	Yes. Expanding revenue, manageable support, high lifetime margin.
CLV:CAC (illustrative)	≈0.012:1	≈4:1 to 6:1

The business had 13,500 customers and believed it was running one business. It was running two, with almost nothing in common except the infrastructure they sat on. And it was spending $3M per year of its marketing budget almost entirely on the one that did not work.

2The warning light

The first signal was not analytical. It was experiential. Three million dollars on a single pay-per-click channel was, in the incoming marketing director's experience across multiple businesses and markets, an extraordinary sum. Not wrong in principle, large businesses spend large amounts on paid media, but sufficiently large to prompt a simple question: how do we know this is working?

The answer provided was the standard digital marketing dashboard: ad impressions, click-through rates, cost-per-click, landing page conversion rates. The programme was generating activity. But activity and commercial return are not the same thing. The incoming director began asking different questions, not 'how many leads did it generate?' but 'which of those leads became customers, what did those customers pay, and what was the fully loaded cost of acquiring each one?'

The data required to answer these questions did not exist in the agency's reporting. The agency's campaign system captured the front-end data: clicks, conversions on landing pages, inbound calls. The CRM held the customer and opportunity data. The two had never been properly connected. When they were, what emerged was a picture that nobody inside the business had seen before, and that directly contradicted the assumptions the business had been operating on.

3The diagnosis: two data problems in one

The diagnostic work uncovered two distinct problems. The first was an attribution problem. The second, and more fundamental, was a unit economics problem. Each made the other worse.

The attribution problem: a broken feedback loop

The agency had been using CRM opportunity data, including late-stage and closed-won records, to optimise the ad campaigns. The logic was sound in principle: use signals from successful customers to improve targeting. The execution was structurally flawed in a way that nobody had thought to examine.

The question that stopped the room: 'You are using closed-won opportunity data to optimise today's campaigns. Given that our sales cycle runs to many months, how can signals from a deal that closed six months ago be relevant to an ad running today?' Nobody had an answer. Nobody had thought about the length of the sales cycle.

The agency was running an e-commerce optimisation model, one built for markets where the journey from click to purchase is measured in hours, inside a B2B managed services market where that journey was measured in quarters. The feedback loop was broken by the nature of the product. The model was optimising confidently on the basis of information that was structurally irrelevant to the decisions being made.

The unit economics problem: the wrong customers at the wrong cost

The broader problem was what the campaign was actually buying. The search terms being bid on, broad terms like 'managed hosting' at approximately $85 per click, were structurally misaligned with the customer the business needed to grow.

The clicks were coming from individuals and small businesses seeking low-cost hosting. These were exactly the customers already represented in the majority of the 13,500-account base. They would sign up via credit card for a few dollars per month. They would use the support service heavily. They would not grow. And on any honest unit economics analysis, they were customers the business could not afford to acquire at $85 a click plus the fully loaded cost of serving them.

The customers the business genuinely needed, the retail e-commerce platforms, the fast-growing online businesses, the enterprise accounts that would expand their spend over time, were not buying after clicking on an ad. They were being won through months of sales effort, technical consultation, and relationship development. The channel was structurally incapable of reaching them. The $3M was not just inefficient. It was pointed in the wrong direction entirely.

3a. The unit economics in retrospect

The customer base research was not, at the time, framed explicitly in CLV:CAC terms. The analysis was guided by commercial instinct and the data available, primarily revenue by customer, tenure patterns, expansion behaviour, and support consumption by cohort. But the findings, when expressed in unit economics terms, make the argument with a precision that no amount of impression data or pipeline reporting could match.

Before presenting the numbers, it is worth understanding the belief they dismantled. The leadership team's conviction that the PPC programme was working was not irrational in structure. It was a version of a legitimate argument: acquire customers cheaply at low initial value, support them as they develop, and as their projects and businesses grow, they grow with us. This is, in fact, a recognisable growth model in technology markets, the idea that early, low-cost users represent the seeds of future enterprise customers. The problem was not the logic of the hypothesis. It was that nobody had ever tested it with data.

Below $1,000 MRR, customers did not grow. The bedroom developers did not become enterprise accounts. The small projects did not become commercial platforms. The engine of future growth was not an engine. It was a cost centre with a compelling narrative.

The leadership team were not being wilfully ignorant or commercially reckless. They were being human, holding onto a story that was internally consistent, had never been disproved, and happened to be wrong. The data did not reveal negligence. It revealed an untested assumption that had been running at $3M per year.

The figures below are reconstructed estimates, not audited calculations. They are grounded in the known inputs at the time: the blended average cost-per-click across all keywords in the campaign (approximately $40, with individual keywords such as 'managed hosting' costing up to $85), the approximate conversion rates, the MRR ranges of each segment, tenure patterns from the research, and the approximate cost of the sales function. They are directionally honest, not forensically precise. The order of magnitude is what matters.

Table 2Unit economics by segment. All figures are retrospective estimates based on known inputs. The directional conclusion, that the small segment CLV:CAC ratio was deeply negative while the growth segment ratio was comfortably positive, is robust regardless of the precise assumptions used.

Metric	Small customer segment	Growth segment
Primary acquisition channel	PPC, short and long-tail search terms	Direct sales + marketing support
Blended average CPC	~$40 (range: $10 to $85 depending on keyword)	Not applicable
Conversion rate (click to customer)	~2% (broad search terms, low purchase intent)	Not applicable, won via multi-month sales cycle
Media cost per acquired customer	~$2,000 ($40 ÷ 2%)	—
Fully-loaded CAC	~$3,000 to $4,000	~$10,000 to $15,000 (sales team cost allocated per close)
Typical entry MRR	$5 to $10 per month	$1,100+ per month
Expansion pattern	Rare. Below $1,000 MRR, customers did not grow. The hypothesis that they would was central to the business's belief in the PPC programme.	Consistent. E-commerce business growth drove hosting growth directly.
Average tenure	6 to 9 months before going inactive	Effectively indefinite, migrating an e-commerce platform is operationally prohibitive for the customer
CLV (illustrative estimate)	~$25 to $55 ($8 avg MRR × 7 months × 50% margin)	~$57,000 to $65,000 ($2,500 avg MRR × 60 months × 50% margin, discounted at 10%)
CLV:CAC ratio	≈0.012:1	≈4:1 to 6:1
Verdict	Destroying capital. Each PPC-acquired customer cost ~$3,000 to $4,000 to win and returned ~$40 in lifetime value.	Creating capital. At 4:1 to 6:1, comfortably above the 3:1 benchmark that defines a healthy B2B acquisition model.

The ratio in the small segment, approximately 0.012:1, means that for every dollar of acquisition cost deployed against this segment, the business recovered approximately one cent in lifetime customer value. The $3M programme was not generating $3M of value. It was generating a fraction of that, from the wrong customers, based on a growth hypothesis that the data showed to be false.

The ratio in the growth segment, 4:1 to 6:1, means that every dollar invested in winning an e-commerce customer returned four to six dollars in lifetime margin. These customers were won through sustained sales effort over months, without meaningful marketing support. The acquisition cost was higher in absolute terms. The return was an order of magnitude larger.

The unit economics framing settles, in two numbers, an argument that nine months of qualitative advocacy could not. A CLV:CAC ratio of 0.012:1 in one segment and 4:1 to 6:1 in another is not a marketing problem. It is a capital allocation problem. And capital allocation is a language every CFO speaks.

4The three-way disconnect

What made this situation instructive was not simply that $3M of spend was misallocated. It was that the misallocation was invisible to every function reporting on its own metrics. Each function had a coherent picture. None of the pictures were connected.

Table 3The three-way disconnect. Each function was operating rationally within its own frame. No mechanism existed to connect the data across functions and ask the unit economics question.

Function	What they were reporting	What they were not seeing
Marketing	Ad impressions, landing page conversion rates, inbound lead volume. Metrics performing within expected ranges. PPC generating activity.	The commercial quality of leads generated. That almost all PPC-sourced customers sat in the unprofitable small segment. No connection between campaign spend and customer lifetime value.
Sales	Pipeline volume, forecast accuracy, revenue target attainment. Growth segment won via direct sales effort and hitting target.	That the growth segment was being won despite marketing, not because of it. The cost embedded in long sales cycles for high-value accounts. The absence of marketing support for the segment that mattered most.
Finance	Revenue, EBITDA, budget compliance. The $3M was within its authorised envelope. Business financials on plan.	What the $3M was actually returning by segment. The structural unprofitability of the majority of the customer base. The mismatch between where the budget was deployed and where the commercial value was generated.
Leadership	Three functions reporting positively. The belief, held firmly and for years, that PPC was the engine of new customer acquisition and future growth.	That this belief had never been tested with data. That the customers PPC was winning were the ones the business could least afford to keep. That the customers driving real commercial value had nothing to do with the $3M.

This is not a story of incompetence. Every function was operating rationally within its own frame. The problem was structural: there was no mechanism, and no person, responsible for connecting the data across functions and asking the unit economics question. What does it cost to acquire a customer in each segment, and what does that customer return?

5The resistance

Making the case to redirect the budget took close to nine months from initial diagnosis. It was refused on multiple occasions. The CEO held a firm conviction, reinforced by years of reporting that appeared to confirm PPC's effectiveness, that the channel was fundamental to the business. When presented with the analysis, the offer made was not to redirect the budget but to increase it. The incoming director refused.

The eventual approval was not enthusiastic. It was given in the manner of a test, with the implicit understanding that if the alternative did not produce results, the decision would be reversed and the consequences for the marketing director would follow. The approval came, as it often does in these situations, with the rope attached. The leadership expected to be proven right.

This is the reality of making a counter-intuitive commercial argument inside an organisation whose tribal knowledge has never been tested. The data can be correct and the resistance can still be substantial. Nine months from diagnosis to approval is not unusual. It may be optimistic.

The resistance also reflected the absence of a shared measurement framework. Because no one had calculated CAC or CLV by customer segment, the debate was necessarily one of assertion against assertion, the director's analysis against the leadership's experience. The data existed to settle the argument. But the habit of connecting it had not previously existed, which meant the argument had to be won on credibility before it could be won on evidence.

6The intervention: separating the two businesses

The response to the diagnosis was not to abandon PPC entirely. That would have been commercially naive, the small customer segment was real, it generated some revenue, and it was what the existing infrastructure and systems were partly built to serve. The intervention was more precise: to acknowledge that the business was running two fundamentally different customer models and to treat them as such.

The small customer segment: restructure, reduce, contain

PPC was retained as the acquisition channel for the small customer segment, but at a fraction of the previous spend. The targeting was tightened to reduce cost-per-click and improve the quality of traffic being generated. A self-service purchase platform was developed, a more sophisticated online experience that allowed customers in this segment to sign up, configure, and manage their accounts without sales involvement. This reduced both the cost of acquisition and the cost of ongoing service.

The support model for this segment was fundamentally changed. Access to staffed telephone support was removed. Support was restructured around a comprehensive FAQ and knowledge base, a resource that served the educational needs of the segment's typical customer without requiring expensive human intervention at every touchpoint. This change alone substantially reduced the support cost per customer.

The keyword strategy was also revised. The campaign moved away from expensive short-tail terms and toward longer-tail alternatives, phrases such as 'on demand hosted services for websites', that attracted visitors with more specific intent and lower expectations of a managed service relationship. Significantly, the word 'managed' was removed from the keyword set entirely. It had been implicitly promising a level of hands-on support that the segment's economics could not sustain.

The economic logic was to make the small customer segment self-sustaining at a lower cost base, rather than subsidising it from a budget better deployed elsewhere.

The growth segment: invest, position, and build

The majority of the redirected budget was deployed against the growth segment, the retail e-commerce businesses and similar accounts that had demonstrated genuine expansion potential and profitability.

The first investment was in understanding the ICP properly. The customer base analysis that had identified the growth segment also revealed its specific characteristics: the sectors most likely to produce high-value, expanding accounts; the trigger events that preceded a buying decision; the commercial concerns, around availability, security, and performance, that were most important to these buyers. This analysis became the foundation for everything that followed.

The product proposition was developed to lead with managed services rather than commodity hosting. Security, availability, and performance guarantees were brought to the front of the commercial offer. This was not simply a repositioning exercise, it reflected what the best existing customers in the growth segment were actually buying and why they had chosen to stay and expand. Making this explicit in the market created a clearer value proposition that was harder to commoditise and easier for the sales team to defend on price.

Paid search was retained for the growth segment, but with a completely different strategy. Rather than competing on expensive broad terms, the campaign moved to long-tail keywords that self-qualified the searcher by specificity, phrases such as 'high availability e-commerce hosting and managed services' that attracted only those who already understood what high availability meant, who were running an e-commerce operation at scale, and who were actively evaluating managed services providers. Click volume was lower. Purchase intent was substantially higher.

Content and ABM activity was developed specifically for the growth segment, designed to support the sales cycle rather than generate top-of-funnel volume. The sales team was brought into the process, their knowledge of what mattered to prospects in the growth segment shaped the content, the campaigns, and the conversations. The relationship between marketing and sales in this segment changed from transactional to genuinely collaborative.

Support services for growth segment customers were upgraded rather than reduced. This was the inverse of the decision made for the small customer segment, and it was deliberate. The commercial value of these customers justified the investment.

7The outcome

The monthly new business pipeline generated by the reoriented approach reached $1.05M, measured as first-touch attributed marketing-sourced pipeline. This was a fundamentally different number from the impression and conversion data that had previously constituted the marketing report. It was pipeline: identifiable opportunities, from identifiable sources, at identifiable value, attributable to specific marketing activities directed at the growth segment.

Beyond the pipeline figure, the results were visible across the business. Margin on new business improved as the higher-value managed services proposition commanded better pricing and required less reactive support per pound of revenue generated. The sales team had a coherent commercial story, a genuine value proposition rather than a price-led pitch, and they used it with confidence. The product development direction shifted toward the needs of the growth segment, producing better offerings in security and availability that further differentiated the business in a competitive market.

The business was subsequently acquired by a Canadian communications group. The acquisition reflected both its financial performance and its strategic position in the managed services market. A business with a clearer commercial proposition, a better-defined target customer, and an improving quality of new customer acquisition is a more credible acquisition target than one without these things.

8. What was left undone, and why it matters

The intervention improved the business's unit economics materially. It did not fully resolve the structural problem that the diagnosis had identified.

The customer base analysis had been unambiguous: approximately 10,000 of the 13,500 accounts were customers the business could not serve profitably, who had shown no pattern of growth, and who were consuming support resource and infrastructure capacity that would have been better deployed against the growth segment. The commercially rational course of action would have been to exit these contracts, either by migrating these customers to a platform better suited to their needs, or by terminating the arrangements and recovering the resource.

This did not happen. The accounts were retained. The argument for exiting them could not be won inside the organisation, partly because reducing the headline customer count was culturally unacceptable, and partly because the analysis, however clear in unit economics terms, was still relatively new and its authority was not yet fully established.

This is a common outcome in unit economics work. The analysis identifies what should be done. The organisation accepts the findings but not their full implications. The intervention is partial. Results improve, but not by as much as they could. The warning light is acknowledged, but not fully acted on.

It is also, as noted in the third working paper in this series, a function of timing. The intervention came after nine months of resistance. By that point, the cost of a full correction had grown. An earlier diagnosis, conducted before the small customer base had reached 12,500 accounts, before the support infrastructure had been scaled to serve them, before the tribal belief in PPC's effectiveness had become deeply embedded, would have produced more options and a more complete response.

The lesson is not that the intervention failed. It succeeded, measurably, and the business was the better for it. The lesson is that the same analysis, conducted two or three years earlier, would have produced a wider set of choices and a lower cost of acting on them. The warning light had been on for longer than anyone had been willing to acknowledge.

9What this case illustrates

The value of this case is in the pattern it represents, one that, in the author's experience across multiple businesses and markets, is more common than the exception.

The business was not managed incompetently. Every function was operating rationally within its own frame. The digital team and their agency were optimising for the metrics they were responsible for. Sales was hitting targets. Finance was managed within budget. Leadership was operating on a reasonable but untested assumption about how new customers were acquired. Nobody was acting in bad faith.

What was missing was a shared unit economics framework, a set of numbers that all three functions could look at together and that answered the question nobody had thought to ask: what does it cost to acquire a customer in each segment, and what does that customer return? When those numbers were finally produced, they settled, in a matter of weeks, an argument that the business had been unable to have for years.

The diagnosis was not the hard part. The hard part was the nine months required to turn a correct analysis into an approved intervention. And the harder part still was the recognition, in retrospect, that a full correction was available and was not taken, because the organisation was not ready to accept the commercial logic that pointed toward it.

The question this case poses for any business operating with similar comfortable metrics and similar unexamined assumptions is not whether this pattern exists in their organisation. It almost certainly does. The question is how long they are prepared to wait before someone asks the unit economics question, and how much that wait will cost.

As Mark Jeffery of the Kellogg School of Management puts it: ‘if you can’t tie marketing to the P&L, you’ll always be fighting for budget.’¹ The case documented here is a detailed illustration of what that connection looks like when it is finally made, and what it costs when it is not made sooner.

A note on anonymisation

This case study is based on real events at a real business. All identifying details, the name of the business, its parent company, the agency involved, the customers referenced, and the individuals concerned, have been removed or altered. The commercial figures, the timeline, the decisions made, and the outcomes described are accurate.