Business Growth

Don’t invest in marketing, invest in a growth engine

3 min read

Published on: 2024-07-16

Tom Goddard

Head of Growth

What is a growth engine?
How to move towards the growth engine model

We recently spoke at an event for startup founders about forecasting realistic customer growth. At the end of the talk, a founder asked: “what is the top priority to maximise customer growth?”

This is a very valid question and in our experience, it’s not asked enough. 8/10 times, the conversation with a startup founder focusses only on questions like “is my budget better spent on Facebook ads or PR?” Essentially, “how can I get in front of the most people?”.

It’s easy to see why filling the top of the funnel as fast as possible is often the priority. Both investors and the competition encourage a “grow at all costs” mentality. It’s much less common that we get questions about keeping customers happy, and this is a problem. While on the surface it might look great to have acquired a tonne of new customers, when you dig deeper and look at the profits, that falls away fast.

Take the on-demand cleaning start-up Homejoy, for example. After starting out with a period of explosive growth, they abruptly collapsed after 2 years. Amongst a slew of other factors, a fundamental contributor was poor product-market fit. It pushed up the cost of acquisition and heavily affected their profits. Focussing on growth above all else meant that any efforts to address retention fell to the wayside. There’s a great article which goes into more detail about it here.

Rather than thinking about marketing as just piling leads into the top of the funnel, startup founders should think about the entire customer lifecycle. To illustrate this idea, here’s the difference in output when the top of the funnel is prioritised compared to the bottom of the funnel.

A useful framework to support this thinking replaces the idea of “marketing” with that of a “growth engine”.

What is a growth engine?

This video was recorded at a WeWork Labs event for startup founders.

Much like an engine consists of a number of parts that work together efficiently to convert fuel into motion, the growth engine components also work collectively, to convert prospects (the fuel) effectively into sustainable growth (forward motion).

Here’s a brief overview of each part of the engine.

1. The customer success component

Whilst it’s tempting to give more attention to prospective customers, making sure existing customers value your product or service should always be the first priority. If your customers aren’t successful, you won’t survive. Your customers are the ones who will become your advocates and help you grow; a paid search campaign alone isn’t nearly as impactful. Finally, your customers will provide insights to refine your growth strategy and improve product-market fit.

2. The lead generation component

We’ve seen a lot of approaches to lead generation that follow the spray and pray philosophy (although it’s never admitted). By leveraging insights from customers, personas can be validated and improved upon. That means more accurate targetting and lower acquisition costs. In other words, fewer, but highly qualified leads are more likely to convert and turn into successful customers.

3. The customer conversion component

Freshly generated leads are not always ready to buy. For the best chance of converting into them into customers, they need to be kept engaged through messages that resonate with them and their buyer journey stage. It’s amazing how often this function is covered by just a sales team with a limited capacity to manage the needs of each lead. By using insights from the customer success component, the right tactics and messages can be used.

None of these individual concepts are new, but thinking of them together as a growth engine promotes joined-up thinking. If you can fluidly connect each of the 3 components, you have a much better chance of growing your customer base and increasing customer lifetime value. Here’s that diagram again:

How to move towards the growth engine model

Anything worth having takes time. Transitioning to a growth engine model requires a commitment to undertake two fundamental long-term projects. To balance that and help build momentum, here are some quick win suggestions too.

Fundamentals: Long-term projects

Implement a flexible CRM:

In an ideal world, a CRM holds all relevant customer and lead data. It should map basic behavioural information like site visits, email opens, ad clicks and video plays to a contact. For this to happen, it must have tight integrations with all relevant touchpoints (that’s where an open API becomes very important). This information builds a strong picture of where a contact is on the buyer journey and therefore how best to interact with them.

Most importantly, a CRM tracks the transition of leads into customers. This facilitates accurate attribution and the CRM can be mined for insights and trends. A CRM only needs to be as sophisticated as your business; there are plenty of free or low-cost options. We like HubSpot and Salesforce.

Validate personas:

It's common sense to understand the people you’re trying to sell to. It’s also remarkable how often this is lacking, considering it underpins the entire messaging and positioning strategy.

Identifying the traits and drivers of your best customers so you can find more of them is a piece of live research to regularly add to. A well-maintained CRM comes in mighty handy for running this kind of analysis. Once you have a list of objective demographic and behavioural characteristics of successful customers, you can apply them further up the funnel to target, score and qualify leads.

Quick wins: short-term projects

Set up a review programme:

Online reviews have a high impact on sales. Set up a simple automated email asking customers to leave you a review on your chosen platform.

Assign personas to new leads:

If you already have a good sense of your personas, you can ask leads to self-identify through a form or chatbot. For example, add a field asking “What is your biggest challenge?” and provide pre-set answers which correspond to different personas. Then tailor the follow-up using the new persona knowledge.

Collect customer feedback:

Set up a pop-up survey visible only to customers on the website or in-app to take a quick pulse check. Limit to 3 closed-ended questions. More about website survey best practice here.

Find out why customers leave:

Ask customers to provide a reason for leaving your product/ service. An automated email will work, or it can be built into the subscription cancellation process.

You might also be interested in

The Rise of the AI-Generated Training Data Industry and Its Impact on Machine Learning

AI training data has emerged as a crucial and rapidly expanding component of the artificial intelligence ecosystem. While most attention often focuses on the impressive advancements in AI models and computing infrastructure, the underlying data industry that powers these models is experiencing its own transformative boom. Leading startups like Mercor, Surge AI, and Handshake AI are pioneering a new frontier in AI development. These companies specialize in providing meticulously curated, expert-annotated, high-quality training data. Unlike earlier crowdsourcing efforts, the modern training data market emphasizes skilled professionals creating domain-specific annotations, rubrics, and environments to optimize AI learning. This trend is driven by the limitations of conventional large-scale model training approaches that rely on generic datasets. To push AI systems towards real-world competence, especially in complex domains like software engineering, law, finance, and medicine, developers require highly specialized data that captures task-specific nuances and criteria. The AI data industry ecosystem now encompasses a diverse array of activities: from detailed rubric design used to evaluate AI outputs, to the creation of customized reinforcement learning environments, to deploying experts with specialized backgrounds who can verify and enable improvements in AI performance. Companies like Mercor have innovated by integrating AI-assisted recruiting and training of software engineers who contribute data annotations aligned with real challenges faced in software development. This not only addresses the need for task-specific expertise but also exemplifies the growing profitability and valuation potential of AI-centric data providers. Meanwhile, established enterprises such as Scale AI and Surging AI continue to expand offerings that go beyond simple labeling, venturing into areas of human feedback, evaluation metrics, and fine-tuning datasets that enhance model reasoning and accuracy. This blossoming market attracts a wide spectrum of professionals, ranging from Nobel laureates and legal experts to mathematicians and even physicists. It represents a fundamental shift where AI development is increasingly a collaborative effort between machine learning technology and human expertise. Investors are betting heavily on this sector, recognizing that AI progress heavily depends on quality human-generated data. The market’s growth foreshadows an economic ecosystem as significant as AI compute hardware, but focused on the human-in-the-loop dimension. While the dream of artificial general intelligence (AGI) envisions systems that require little additional training data, present realities emphasize the ongoing need for specialized, augmentative data creation, ensuring AI systems remain adaptable and capable. In conclusion, the AI training data industry is a foundational yet often underappreciated pillar of AI innovation. Its evolution through sophisticated human intervention marks a profound change in how AI systems are taught, improved, and validated, driving the next wave of machine learning breakthroughs.

How Stack Overflow Navigates AI Disruption: Trust, Community, and the Future of Coding

Stack Overflow, the premier Q&A forum for developers, has faced existential challenges with the rise of generative AI like ChatGPT. This new AI wave upended the traditional way developers seek help—by connecting with peers on Stack Overflow—to how AI tools can answer coding questions instantly. In late 2022, Stack Overflow's leadership perceived this AI change as a crisis, declaring it a disruptive moment. The company reallocated 10% of its team to focus exclusively on AI strategies to respond to this. This led to a fundamental shift from a public Q&A site to a hybrid model emphasizing enterprise SaaS solutions enhanced by AI. Key to Stack Overflow's approach is maintaining its trusted community knowledge base. The company banned AI-generated answers backed by detecting low-quality AI content, preserving the integrity of its platform’s human-curated answers. Simultaneously, Stack Overflow released AI Assist, an AI-powered conversational tool that sources answers first from the trusted Stack Overflow corpus before consulting external AI models, to keep responses authentic. This strategy helps manage AI's hallucination problem while providing developers with natural language tools they demand. The company recognizes the enduring importance of human interaction for complex problems. Despite AI's prevalence, many developers still prefer connecting with real people for difficult technical challenges, evidenced by Stack Overflow's continued engagement on thorny topics. Stack Overflow monetizes through three primary channels: the enterprise SaaS platform powering AI assistants with company-specific data, licensed community data for model training, and advertising targeted at developers. Remarkably, the data licensing deals with AI labs such as OpenAI and Google have become a vital revenue stream. Despite widespread AI use, trust remains a major challenge. Surveys show over 80% of Stack Overflow users use AI for coding aid, but only 29% trust AI-generated outputs fully. This mistrust comes from AI models’ tendency to hallucinate or provide inaccurate results, complicating integration. Looking ahead, Stack Overflow is focused on building a knowledge intelligence layer for enterprise use, enabling trustworthy AI agents within companies. Alongside this, the public platform continues to evolve with new engagement modes like discussion chat rooms and coding challenges to support developer growth and community. Stack Overflow’s story is an example of a legacy tech platform adapting amidst AI disruption by balancing AI-powered innovation with the irreplaceable value of human expert curation and trust. The future of software development is likely hybrid, with humans and AI collaborating, and Stack Overflow aims to remain central to this new ecosystem.

The Booming AI Data Annotation Industry: Powering the Future of Machine Learning

The AI data annotation industry is rapidly growing, fueled by an insatiable demand for high-quality, specialized training data to power advanced machine learning models. Young entrepreneurs like Brendan Foody have launched companies such as Mercor, which use automated processes to hire software engineers overseas for data labeling tasks, quickly scaling to multimillion-dollar revenues. This sector is vital to AI progress, as models require precisely labeled datasets prepared by experts in fields like programming, finance, and medicine to improve reliability and performance. Traditional crowdsourcing platforms proved insufficient due to quality concerns and lack of domain expertise, leading to the rise of companies like Scale AI and Surge AI, which focus on recruiting experts and maintaining strict annotation standards. The industry is also witnessing unprecedented investment, with valuations in the billions and new startups emerging regularly. These companies are diversifying into related services such as evaluation testing and specialized reinforcement learning environments, which are designed to train AI models more effectively across complex real-world tasks. Despite criticisms of the AI model economy and uncertainties about achieving artificial general intelligence, data annotation remains a lucrative and essential field. AI labs continue to pour billions into data acquisition, even as they refine training techniques on smaller, tailored datasets. Experts believe this focus on quality human data will be a key bottleneck and driver of future AI development. The future of AI may well depend on the continued expansion of this data annotation ecosystem, with predictions that data annotator roles could become among the most common jobs globally. The intricate work of creating highly granular training rubrics, hiring domain specialists, and developing custom environments underscores the complexity and promise of this emerging industry.