Data-driven decision-making isn’t about blindly following algorithms; it’s about strategically leveraging data and models while honing human expertise
In the age of ubiquitous AI and data-driven insights, a critical question for sales leaders emerges: how much should we rely on these automated recommendations? The answer, as with most things in life, is nuanced and demands a deep understanding of the interplay between decision stakes, data reliability, and human judgement.
Consider the diverse scenario: an AI suggests Cloud services to a medium-sized company, a LinkedIn list pops up for a financial role, and a model proposes downsizing a pharmaceutical sales force. These examples highlight the pervasiveness of data-driven insights, but also the inherent complexities. Data accuracy, model quality, and decision consequences can vary dramatically, influencing how we should engage with these “predictions.”
Effective sales decision-making hinges on understanding two key factors:
- Decision Stakes: How impactful is this choice? Will a mistake be costly and hard to fix? Examples of high-stakes decisions include new product channel selection, incentive plan design, and key team hiring profiles. Low-stakes decisions, like a salesperson’s product suggestion, have minimal downsides and are easily revisited.
- Data-Driven Insight Reliability: How trustworthy is the underlying data and model? A salesperson’s incentive pay, calculated by a model, carries high reliability and modest decision stakes. Conversely, an AI scoring LinkedIn profiles for sales jobs offers moderate reliability at best, due to profile biases and incompleteness.
By mapping these two factors, we can navigate the data-driven landscape intelligently:
- Modest Stakes & High Reliability: Embrace the insights readily. Incentive calculations or AI-generated meeting summaries fall in this category.
- Modest Stakes & Moderate Reliability: Use the insights as helpful suggestions, but apply scrutiny and judgement. LinkedIn-based candidate evaluations fall under this umbrella.
- High Stakes & Moderate Reliability: Treat the insights with caution. Analyse the model’s rationale, assess data quality, and involve human oversight before implementing the suggested action. An AI’s customer interaction recommendation is a prime example.
- High Stakes & High (but not perfect) Reliability: Consider the insights valuable, but prioritise supervision and judgement. Salesforce size and structure recommendations, with imprecise inputs and uncertain future scenarios, illustrate this category.
Navigating with Nuance
Evaluating data-driven insights for optimal use involves several strategies:
- The Sniff Test: Does the insight make sense? Use experience and external benchmarks to assess its plausibility. Doubling your sales team to 500 based on a model’s suggestion? Compare it to competitor sizes for a reality check.
- The Explanation Factor: Can the recommendation be justified? A Cloud cross-sell suggestion is more convincing if explained with supporting data, like similar companies successfully adopting the solution. Transparency in model reasoning builds trust.
- Data Quality Matters: Is the data relevant, accurate, complete, and timely? Recommendations based on LinkedIn profiles are inherently vulnerable to data flaws, while salesperson route planning utilising GPS, traffic, and appointment data benefits from accurate and timely inputs.
- Model Quality Insights: While assessing technical models directly can be challenging, understanding the modeller’s experience and industry knowledge is crucial. Model building is an art, and the human element significantly impacts model design and performance.
Data-driven decision-making isn’t about blindly following algorithms; it’s about strategically leveraging data and models while honing your own expertise. We must avoid confirmation bias, seeking only data that supports our beliefs. Instead, embrace non-intuitive insights and continuously improve your data-driven skillset.
Remember, AI-based recommender systems evolve over time. Adoption rates often climb from 40% to 80% within two years, reflecting both model improvement and user proficiency. The future of sales lies in a judicious blend of model insights and human judgement, navigating the dynamic digital landscape with both data and wisdom.