Understanding Data Confidence

AskBiz uses AI and statistical models to generate predictions, anomaly scores, and segmented insights. These models require data — the more historical data available, the more confident the model can be in its output. A business that connected Shopify six months ago has far less data for AskBiz to learn from than one that connected three years ago. Confidence levels communicate how much to trust a specific output, so you can decide how to act on it.

High confidence (green indicator): the model has sufficient historical data and the pattern is consistent. Treat this output as reliable for decision-making. Medium confidence (amber indicator): the model has some data but patterns are variable or the history is short. Use this directionally — it points you in the right direction, but do not make high-stakes decisions based on it alone. Low confidence (grey indicator): limited data, new connection, or highly variable history. Treat as a rough signal only. More data over time will improve confidence automatically.

The main factors: data history (longer history → higher confidence), data consistency (regular, complete sync history → higher confidence than gaps), metric complexity (simple totals have high confidence by default; multi-variable predictions like churn score need more data), and business volatility (highly seasonal or fast-changing businesses are inherently harder to model, so confidence may stay medium even with good data). You cannot manually set confidence levels — they are calculated automatically.

Low confidence does not mean the metric is wrong — it means the model is uncertain. In practice: do not use a low-confidence churn prediction to make individual customer decisions (calling someone to retain them when the model is not sure they are at risk). Do use a low-confidence metric to identify an area worth investigating manually. As you use AskBiz and more data accumulates, confidence levels typically improve within 2–3 months.