The Metrics That Feel Rigorous (And What They're Actually Measuring)

TLDR;-) CAC, CLV, ROAS, retention rates. These metrics feel like rigour. They appear in every board pack, every due diligence process, every growth review. The problem isn't that they're wrong. It's that they're downstream. They measure the output of a commercial engine that hasn't been built on verified foundations. When the upstream conditions aren't established, these numbers aren't performance data. They're noise with a decimal point.

The conversation usually starts with the metrics.

CAC is too high. CLV doesn't justify it. Retention is declining. ROAS has dropped. The board wants to know why growth has stalled, and the team has arrived with a slide deck full of numbers to explain it.

The metrics are real. The diagnosis they produce is wrong.

These are all downstream measures. They capture what happens after a buyer enters the commercial engine: acquisition cost, lifetime value, return on spend, whether they stay. What they cannot tell you is whether the commercial engine is built on the right foundations. And in the majority of cases where growth has stalled, it isn't.

What the metrics actually require

Every one of these measures depends on upstream conditions that most organisations have never verified.

• CAC is only a meaningful number if you're acquiring the right customers. If the ICP isn't defined with genuine precision, CAC is the cost of a distribution experiment disguised as commercial performance. The denominator is real. The numerator includes people who will never convert to long-term value.

• CLV requires that "lifetime" be long enough to matter. The Ehrenberg-Bass Institute at the University of South Australia has documented across hundreds of categories that most brands have much weaker loyalty dynamics than their internal data suggests. The customers who look loyal in the cohort analysis are often heavy category buyers who buy many brands frequently, not brand loyalists. CLV models that treat these buyers as locked-in are measuring a projection, not a reality.

• ROAS measures the relationship between spend and revenue. It says nothing about whether the revenue came from buyers who had genuine mental availability for the product before the ad fired, or whether the ad was doing the upstream work of building recognition that should have been done earlier and more cheaply.

• Retention tracks who stays. It does not tell you whether the ones who left left because of execution failures, because the product was never the right fit for them, or because the brand has no mental availability to return to when the need reactivates.

Every one of these measures is a trailing indicator. Useful, once the underlying structure is right. Misleading when it isn't.

The sequencing error

The reason organisations reach for these metrics before the underlying structure is established is downstream gravity.

I use this term in my diagnostic work to describe the pull toward later-stage activity before earlier-stage dependencies are resolved. The pull is strong because downstream work is visible. Metrics dashboards, pipeline reviews, agency briefings, conversion optimisation. It generates data constantly. It creates the appearance of rigour.

The upstream work that makes downstream metrics meaningful is less visible. Who, precisely, is the buyer? Not "CMOs at mid-market B2B companies" but the specific role, the specific trigger, the specific problem they're trying to solve in the language they'd use to describe it. What's the competitive context from the buyer's perspective, not from the founder's? What will they pay, and does that number make the economics work across the full customer acquisition cycle?

These questions sit in what I call Layer 1 of the Critical Path Layers: market clarity. They have a specific structure. ICP, positioning, and pricing form an interdependent triangle. Movement on any one forces reassessment of the other two. A buyer persona without a clear competitive position isn't an ICP. A competitive position without a validated price point is a hypothesis, not a strategy. And none of these is resolved by campaign performance data, because campaign data measures response to a message aimed at people who may or may not be the right buyers.

The Ehrenberg-Bass research makes the dependency structure explicit. Byron Sharp's How Brands Grow (2010) identified mental availability as the primary upstream driver of commercial performance: buyers recall the brand when a category need arises. Not "awareness" in the abstract. Recall at the specific moment when the need is triggered. That is a Layer 1 problem. It requires knowing which category entry points your buyers use, and whether your brand is linked to them in memory. No amount of ROAS optimisation builds this. It's built upstream, before the commercial engine runs.

What the Double Jeopardy Law tells you

Andrew Ehrenberg's Double Jeopardy Law, first formalised in the 1960s and validated across virtually every category since, shows that smaller brands have two disadvantages simultaneously: fewer customers and lower loyalty from those customers. The loyalty gap is a function of market share, not a cause of smallness.

This matters for how you read your own metrics. If your retention is below benchmark, there are two possible diagnoses. The first is an execution problem: the product isn't delivering, the onboarding is broken, the support is inadequate. That's fixable with Layer 3 and Layer 4 work. The second is a penetration problem: you have a small customer base with normal loyalty dynamics for your size, and the metric looks bad because the penetration isn't there yet. Investing in retention programmes to fix a penetration problem doesn't fix anything. It manages a small base more carefully while the underlying constraint goes unaddressed.

Most retention conversations I observe are trying to solve a penetration problem with retention tactics. The diagnosis requires going upstream.

Where most growth reviews go wrong

A growth review that starts with metrics has already made a structural error. It has accepted the metrics as the problem definition, which means the diagnosis will be conducted entirely in the downstream layer.

"CAC is too high" becomes "we need to reduce acquisition costs," which becomes a brief to the media agency. The agency optimises the channels it controls. CAC may improve marginally. The underlying issue, which is that the ICP was never sharp enough to generate efficient targeting, doesn't move.

"Retention is declining" becomes "we need a loyalty programme," which becomes a product investment in features designed to increase stickiness. Some users engage with the features. The cohort curves don't change, because the users who are churning weren't going to be retained by product features. They were never the right buyers.

"ROAS has dropped" becomes "we need better creative," which generates a round of creative testing. The creative improves. ROAS recovers partially. Nobody examines whether the spend was building mental availability in the right segments or efficiently converting buyers who were already going to purchase.

Each of these responses is reasonable within its own layer. The error is treating a downstream metric as a root cause rather than as a signal that something upstream hasn't been established.

The diagnostic before the dashboard

The questions that reorient a growth conversation aren't about the metrics. They're about the upstream conditions the metrics depend on.

Do you know which category entry points trigger purchase in your segment? Not what buyers say when you ask them. What actually fires the consideration process. Sharp and Romaniuk's research on category entry points shows that brands with strong mental availability are linked to multiple specific triggers, not to a generalised brand positioning. If you can only name one or two triggers, your mental availability mapping is incomplete. Your acquisition spend is probably hitting buyers who weren't activated.

Is your ICP defined by observed buying behaviour, or by demographic profile? A buyer persona built from demographic data describes who the buyer is. It says nothing about why they buy, what they're replacing, what their decision process looks like, or whether they'll be advocates. Ehrenberg-Bass research shows that buyers across most categories are light buyers most of the time. The "ideal customer" who buys heavily and loves the brand is a small fraction of most companies' revenue, even when CLV models suggest otherwise. Growth comes from reaching the broad population who buy occasionally, not from intensifying the relationship with the few.

What is your penetration rate, and what does the Double Jeopardy benchmark for your category suggest it should be? If penetration is the driver of loyalty, not the other way around, then your loyalty metrics are readable only in the context of your penetration. Without the benchmark comparison, you're reading one instrument and ignoring the one that explains it.

If none of these questions can be answered with evidence, the metrics review isn't the right conversation. The right conversation is upstream.

Further Reading

• Byron Sharp, How Brands Grow (2010) — The empirical case for penetration over loyalty, and the structural conditions that enable it.

• Jenni Romaniuk and Byron Sharp, How Brands Grow Part 2 (2016) — Category entry points and distinctiveness, with application to specific market types.

• Andrew Ehrenberg, Gerald Goodhardt, and Patrick Barwise, "Double Jeopardy Revisited" (1990) — The foundational paper on penetration and loyalty dynamics.