Why Most Growth Strategies Fail

The first discipline any executive team needs is a shared vocabulary. Five things get called growth in most companies, and only one of them is.

• Activity: the work performed. Campaigns run, calls made, releases shipped.
• Output: what the work produced. Impressions, sign-ups, feature usage, meetings booked.
• Customer outcome: whether the customer's situation actually improved. A problem solved, a decision made easier, a cost removed.
• Incremental impact: the difference between what happened and what would have happened without the initiative.
• Enterprise value: the durable effect on revenue, margin, retention and the multiple the market applies to them.

Most reporting systems are excellent at the first two, adequate at the third, and silent on the fourth and fifth. That silence is where growth strategies go to die.

Consider a familiar example. A retail bank runs a spending campaign offering cashback to cardholders who transact three times in a month. Response is strong. Transacting customers rise. The team reports a double-digit lift and asks for a larger budget. Then someone builds a holdout group and discovers that a large share of the responders were already habitual monthly spenders. The campaign paid a reward for behaviour it did not create. The reported lift was real. The incremental lift was a fraction of it, and in some segments the programme was value-destructive once the reward cost was netted off.

A campaign can perform well and still create almost nothing. Performance measures the response. Incrementality measures the difference.

This is not a marketing problem. The same confusion appears in product, in pricing, in sales coverage and in partnerships. Any time an intervention is aimed at a population that was already inclined to act, measured response will overstate created value.

Across startups and large enterprises, I have observed the same five failure modes. They are structural rather than personal, which is the good news, because structures can be redesigned.

First, too many priorities. When a leadership team commits to twenty initiatives, it has not made a choice, it has made a list. Each initiative receives a fraction of the funding, talent and attention needed to reach the threshold where an effect would even be detectable. The organisation then concludes that nothing works, when in fact nothing was properly tried.

Second, no quantified value hypothesis. Ask a team what the initiative is worth and you will often hear a strategic rationale rather than a number. A quantified hypothesis is a simple sentence: for this customer group, this intervention should change this behaviour by roughly this much, which is worth roughly this amount over this period. Without it there is no way to size the bet, no way to design a test with adequate power, and no way to be wrong.

Third, functional ownership instead of customer ownership. Growth failures are usually handoff failures. Marketing owns acquisition, product owns activation, service owns retention, finance owns the P&L, and nobody owns the customer's journey end to end. The gaps between the functions are exactly where value leaks.

Fourth, success measures selected after launch. If the metric is chosen once results are visible, the organisation has guaranteed itself a favourable answer. Post hoc metric selection is the single most common way that companies deceive themselves at scale.

Fifth, scaling before validating. National rollout is treated as a reward for a good pilot rather than as a separate decision requiring its own evidence. Pilots are typically run in favourable conditions by motivated teams. Neither condition survives contact with the full estate.

The consulting evidence on large change programmes points in the same direction. McKinsey has argued for years that roughly 70% of complex, large-scale change programmes fail to reach their stated goals, a figure the firm attributes largely to employee resistance and insufficient management support rather than to poor strategy. That number deserves care. It is a practitioner estimate drawn from consulting experience and survey work, not a controlled measurement, and the definition of failure varies across studies. Treated as a law of nature it becomes an excuse. Treated as a warning about execution capacity, it is useful.

McKinsey's global survey work on transformations adds a more actionable finding: fewer than a third of respondents said their organisations had succeeded in improving performance and sustaining those improvements over time, and the transformations that did succeed captured a far higher share of the financial value they had identified, around two thirds, compared with roughly a third for the rest. The gap between those two numbers is not strategy. It is the discipline of execution, ownership and measurement.

Executives worry about initiatives that fail. I worry more about initiatives that appear to succeed. A visible failure is cheap: it is stopped, and the capital returns to the pool. A false positive is expensive in four ways at once.

• It consumes budget that had a better alternative use.
• It consumes management attention, which in most organisations is scarcer than money.
• It becomes a precedent. Once a programme is declared a success, questioning it carries a political cost.
• It corrupts the planning model. Next year's forecast is built on an effect size that was never real.

One of the most important lessons I learned while building Carasti was that the cost of a false positive is not the money spent, it is the option you no longer have. When capital is constrained, a channel that looks like it works absorbs the next round of spend automatically. By the time the truth surfaces, the alternative you should have funded has been overtaken by someone else.

My views here were formed in three settings that could hardly be more different, and the same lesson appeared in all of them.

Advising executive teams early in my career taught me that most strategic disagreements are actually measurement disagreements in disguise. Two leaders arguing about whether to invest in a segment are usually holding different implicit baselines. Make the baseline explicit and the argument frequently resolves itself in twenty minutes.

Scaling a regional enterprise software business taught me the difference between a pipeline and a plan. Pipeline is a claim about the future made by people with an incentive to be optimistic. A plan is a set of resource commitments that only makes sense if certain conversion rates hold. When I started stress-testing the conversion assumptions rather than the pipeline total, forecast accuracy improved and, more importantly, so did the quality of the deals we chose to pursue.

Building Carasti across four markets taught me the value of a baseline you cannot argue with. In an asset-heavy subscription business, utilisation and residual value are unforgiving. You can tell yourself a story about brand awareness for a while, but the fleet tells you the truth every month. What looks straightforward in a strategy presentation often becomes considerably harder once real assets, real customers and real cash conversion are involved.

Leading analytics-driven commercial growth since then has reinforced the final point. The organisations that grow fastest are not the ones with the most sophisticated models. They are the ones where a model output reliably becomes a decision, the decision becomes an action, and the action is measured against a credible comparison.

This is the operating loop I use with teams. It is deliberately simple, because complexity is what allows organisations to skip stages without noticing.

The loop closes because stage five feeds the next stage one. Each cycle should leave the organisation with a better prior about which levers work for which customers, and that accumulated prior is a genuine competitive asset. Most companies throw it away every planning cycle.

Adopting a loop is easy. Changing the conditions that made the loop necessary is the real work, and it sits squarely with the executive team.

Governance. Growth reviews should open with the value hypothesis and the baseline, not with a status update. A useful rule: no initiative reaches an investment committee without a named decision owner, a stated baseline and a pre-registered success measure.

Incentives. If people are rewarded for launching, they will launch. If they are rewarded for verified value, they will measure. Most incentive systems reward the first while the strategy document asks for the second, and people are rational enough to follow the incentive.

Portfolio management. Growth initiatives should be managed like a portfolio with explicit risk tiers: proven scaling bets, validated but unscaled, and genuine exploration. Each tier deserves a different evidence threshold. Applying experimental rigour to exploratory bets kills innovation, and applying exploratory tolerance to large scaling bets destroys capital.

Test design capability. Very few commercial teams have access to someone who can design a properly powered test and interpret it honestly. This capability is inexpensive relative to the spend it protects, and it should sit close to the business rather than buried in a central analytics function.

A learning repository. Every test, its hypothesis, its design and its result, kept in one searchable place. Without it, organisations pay repeatedly to learn the same thing, and institutional memory leaves with each reorganisation.

Not every strategic investment can or should be evaluated through a controlled experiment. You cannot randomise a market entry, an acquisition, a brand repositioning or a multi-year platform rebuild. Insisting on experimental proof in those cases produces paralysis, and paralysis is its own form of value destruction.

Where experimentation is impractical, the discipline shifts rather than disappears. Make the assumptions explicit and write them down. Use scenario analysis to identify which assumption the decision is most sensitive to. Phase the implementation so that early stages generate information rather than only commitment. Define leading indicators that would tell you within months, not years, whether the thesis is holding. And agree in advance what evidence would cause a redesign or an exit.

There is also a genuine risk of over-measurement. Organisations can spend more designing the test than the decision is worth, or optimise so tightly for measurable short-term effects that they underinvest in slower-compounding assets such as brand, trust and talent. Proportionality matters: the rigour applied should scale with the capital at risk and the reversibility of the decision.

Finally, incrementality analysis is only as good as the comparison. A badly constructed holdout, a contaminated control group or a test run during an unusual trading period can produce a confident answer that is simply wrong. Scepticism should be applied to favourable results with at least the same energy it is applied to unfavourable ones.