The unusual position of this stage
Define and Build are dominated by practitioner methods with thin independent evidence. Drive is different, and the difference is uncomfortable.
Marketing has one of the strongest empirical bodies in commercial research — decades of replicated findings across categories, countries, and eras. It also has an industry that largely operates as though those findings do not exist.
The gap is not subtle. The best-evidenced conclusions in marketing science contradict the loyalty programmes, targeting strategies, and segmentation exercises that occupy most marketing budgets.
That makes this cluster different in character. In Define, the work was qualifying overclaimed statistics. Here, the work is mostly reporting what the evidence says and noting how far standard practice sits from it.
This article covers the Drive stage only: generating demand, converting it, and measuring whether any of it worked. It is the third of four stage guides.
The same evidence test
Three grades, carried across all four clusters.
Empirically grounded. Independent research, replicated across contexts, published where it can be challenged.
Documented practitioner method. Field-developed, internally consistent, with evidence largely produced by its advocates.
Popular but thinly evidenced. Widely taught, intuitively appealing, resting on selected cases.
Drive has the best ratio of the four stages. Four of the six frameworks below are empirically grounded.
Six frameworks that hold at Drive
1. The Ehrenberg-Bass laws
Empirically grounded. The strongest body of evidence in this series.
A set of replicated regularities in buying behaviour, developed at the Ehrenberg-Bass Institute and set out in How Brands Grow by Byron Sharp and its second volume by Jenni Romaniuk and Sharp.
The central law is double jeopardy, first observed by McPhee in 1963 and formalised by Ehrenberg, Goodhardt, and Barwise in 1990. Smaller brands are punished twice: they have fewer buyers, and those buyers are slightly less loyal. Loyalty is not an independent lever — it moves with penetration.
The replication record is the reason to take it seriously. Double jeopardy has been found in packaged goods, retail banking, insurance, luxury goods, political voting, car buying, concrete supply, airlines, and newer categories including music streaming and ride-hailing. Romaniuk and Wight tested it across 60 brands, 10 categories, and 7 countries in Marketing Letters. Trinh, Dawes, and Sharp extended the analysis to buyer groups, finding that incremental growth comes from light and non-buyers rather than from intensifying heavy buyers.
What follows practically: growth comes from reaching more category buyers, most of whom buy your brand rarely or never. Loyalty campaigns aimed at existing heavy buyers target the group with the least remaining headroom.
Where it strains: the institute is funded by corporate sponsors, and Sharp's responses to critics are notably dismissive of them. Neither undermines the replication record, and both are worth knowing.
2. The 60/40 brand and activation ratio
Empirically grounded.
Les Binet and Peter Field's analysis of the IPA Databank found that campaigns allocating roughly 60% of budget to long-term brand building and 40% to short-term activation produced the strongest business effects over time.
The mechanism is that the two do different jobs on different timescales. Activation converts demand that already exists. Brand building creates the demand that activation later converts. Measured over a quarter, activation always looks more efficient, which is precisely why budgets drift toward it.
What follows practically: short-term efficiency metrics systematically under-reward the activity that produces most of the long-term effect.
Where it strains: 60/40 is an average across a case database, with real variation by category, brand size, and purchase cycle. Treated as a universal constant it becomes the same kind of borrowed rule this series keeps finding.
3. Behavioural science that has been tested
Empirically grounded, when the right frameworks are used.
COM-B — capability, opportunity, motivation, behaviour — and the Behavioural Insights Team's EAST framework (easy, attractive, social, timely) come from applied behavioural research with published trials behind them.
This matters because behavioural science in marketing is dominated by lists of cognitive biases with weak replication records. COM-B and EAST are the versions with methodology attached.
Where it strains: behavioural interventions produce modest effects reliably, not large effects occasionally. Anyone promising transformation from a nudge is selling something.
4. Incrementality measurement
Empirically grounded.
Geo-based experiments, holdout tests, and marketing mix modelling answer a question attribution cannot: what would have happened anyway.
Platform-reported conversions credit outcomes that would have occurred without the ad. Incrementality testing measures the difference between treated and untreated populations, which is the only figure that supports a spend decision.
Where it strains: it requires scale, patience, and tolerance for a number lower than the platform reports.
5. Growth loops
Documented practitioner method.
The argument, developed by Brian Balfour and others, is that funnels describe a linear path while real compounding growth comes from loops where the output of one cycle becomes the input of the next — referral, content, marketplace supply.
Where it applies: any business where an existing customer's activity can produce a new one.
Where it breaks: loops are asserted rather than measured. A loop with a cycle rate below one is a funnel with extra diagram.
6. Structured channel selection
Documented practitioner method.
The Bullseye framework and its relatives address a real problem: businesses default to the channels their team already knows rather than testing systematically.
Where it applies: early channel discovery, where the honest position is that nobody knows which channel will work.
Where it breaks: tests run without enough spend or duration to detect a real effect, producing confident conclusions from noise.
Two frameworks that need handling
The funnel as a model of behaviour
Documented practitioner method, over-applied.
AARRR and its variants are useful for organising measurement. They are poor descriptions of how people actually buy, because real purchasing is non-linear, involves many people, and mostly happens among buyers who were not in-market when they first encountered the brand.
Use the funnel as an accounting structure. Do not use it as a theory of the customer.
The Hook Model
Popular, thinly evidenced.
Trigger, action, variable reward, investment. Intuitively compelling, built from selected examples of habit-forming products, with no controlled research showing that following it produces habits.
It also carries an ethical problem its own author has since acknowledged publicly. Engineering compulsive use is a design choice with consequences, and a framework whose core mechanism is variable reward deserves that scrutiny rather than a slide.
These are one sequence, not a menu
Understand how the category actually buys. The Ehrenberg-Bass laws describe the buyer distribution you are working within. Most growth plans assume a shape the evidence says does not exist.
Split the budget by job before by channel. Brand building and activation do different work on different timescales, and the split decision precedes every channel decision.
Design the behaviour, not just the message. COM-B and EAST address whether the action is possible and easy, which frequently matters more than persuasion.
Choose channels by test, not by familiarity. With enough spend and duration to detect an effect.
Measure incrementally. Attribution reports what happened near an ad. Incrementality reports what happened because of it.
Reversing this order produces the common pattern: channels chosen first, budget split by whatever the channels consume, measurement supplied by the platforms selling the media.
Diagnostic: is the Drive stage working?
Seven tests. Two or more failures means the growth plan is running on assumptions.
The plan targets category buyers broadly, not only existing customers or a narrow segment.
Brand and activation budgets are split deliberately, and the split is defensible.
At least one incrementality test has been run, and its result is known.
Reported returns are not taken from the platforms selling the media.
Any claimed growth loop has a measured cycle rate.
Channel decisions trace to tests rather than to team familiarity.
Someone can name a channel that was tested and stopped.
Test three is the one that changes budgets. Most businesses have never run it.
What this stage produces
Demand that can be attributed to something the business did, at a cost it can sustain.
That is the honest description. Not awareness, not engagement, not reach — those are inputs. The Drive stage closes when a business knows which activity produced incremental customers, what each one cost, and whether that cost is inside the economics the model requires.
Most businesses never reach that state, and the reason is rarely capability. It is that the measurement they rely on is supplied by the parties being measured.
Frequently asked questions
What is the double jeopardy law?
Smaller brands are penalised twice: they have fewer buyers, and those buyers are slightly less loyal. First observed in 1963 and formalised in 1990, it has been replicated across categories from packaged goods and banking to airlines, concrete supply, and ride-hailing.
Does loyalty marketing work?
Less than most plans assume. Loyalty metrics move with penetration rather than independently, and research on buyer groups finds incremental growth comes from light and non-buyers rather than from intensifying heavy buyers, who have the least headroom left.
What is the 60/40 rule?
Binet and Field's finding from the IPA Databank that roughly 60% of budget to long-term brand building and 40% to short-term activation produces the strongest business effects. It is an average with real variation by category and purchase cycle, not a constant.
Why does short-term measurement favour activation?
Because activation converts demand that already exists, while brand building creates demand converted later. Over a quarter, activation always looks more efficient, which is why budgets drift toward it regardless of the longer-term effect.
What is incrementality testing?
Measuring the difference in outcomes between a treated and an untreated population, which establishes what would have happened anyway. It answers a question platform-reported attribution cannot.
Is the Hook Model worth using?
It is thinly evidenced — built from selected examples with no controlled research showing that following it creates habits — and it carries ethical concerns its own author has publicly acknowledged. Treat it as vocabulary rather than as method.
Where should a business start at the Drive stage?
With the buyer distribution in its category, then the brand-activation split, then channels, then measurement. Reversing that order produces a plan built around whichever channels the team already knows.
Sources
Sharp, B., How Brands Grow (2010); Romaniuk, J. and Sharp, B., How Brands Grow Part 2; Ehrenberg-Bass Institute research
Ehrenberg, A., Goodhardt, G. and Barwise, P., on double jeopardy (1990); McPhee, W. (1963)
Romaniuk, J. and Wight, S., Marketing Letters; Trinh, G., Dawes, J. and Sharp, B., Marketing Letters (2023)
Binet, L. and Field, P., The Long and the Short of It, IPA Databank
Michie, S. et al., COM-B; Behavioural Insights Team, EAST framework
Balfour, B., published work on growth loops; Weinberg, G. and Mares, J., Traction
Eyal, N., Hooked, and subsequent published reconsideration
Structure your next phase
Zerologic builds acquisition and behavioural campaign systems on the evidence rather than on category convention — and measures them against what would have happened anyway.
Talk to us: partners@zerologic.io · zerologic.io
