Six months after enabling predictive lead scoring, a consultant notices that scores no longer align well with which leads actually convert, since the client's target market shifted following a new product line launch. What should the consultant recommend?
Select an answer to reveal the explanation.
Short Explanation
When the market shifts underneath a model, like a new product line changing who tends to buy and why, the model's old lessons stop matching reality, and the fix is to give it a refresher course using the newer outcomes rather than assuming it will just keep working forever unchanged. Throwing out the AI approach altogether and going back to manual rules gives up ground unnecessarily when the actual problem is stale training data, not a flawed concept. And wiping out the historical records does not help either, since that erases useful signal instead of adding the fresh signal the model actually needs to catch up with the new buying patterns. The sensible move is straightforward: feed it current outcomes and let it relearn.
Full Explanation
The correct answer is D. Predictive models reflect the patterns present in the data they were trained on, and when the underlying buying patterns shift, such as after a new product line changes who converts and why, the model's accuracy will drift unless it is reviewed and retrained or refreshed using more recent outcome data that captures the new patterns. Option A is incorrect because it rests on a false premise; models are not permanently fixed and do require periodic review as business conditions evolve, otherwise their recommendations grow stale and misleading. Option B is incorrect because abandoning predictive scoring entirely in favor of manual rules discards a working, learnable capability over one accuracy issue, when retraining directly addresses the root cause without losing the benefits of the AI-driven approach. Option C is incorrect because deleting historical records destroys valuable training data, including patterns that may still be relevant, and does nothing on its own to incorporate the newer post-launch outcomes the model actually needs to realign its predictions.