How Faherty Hit Sub-10% Forecast Error on a $107M ecommerce Target

Faherty is an omnichannel apparel brand with ~75–78 physical retail stores nationwide. For 2025, leadership set an ecommerce target of ~$107–108M with a fixed marketing budget. Previous planning relied on historical averages, without scenario modeling, actuals benchmarking, or a way to test downside assumptions.

Saras Analytics built a four-input scenario-based forecasting model using customer behavior data and delivered it in Excel for the finance team. The model projected ~$98M, leading to a strategy adjustment. Actual ecommerce revenue reached ~$109M by year end, with ~8% error on new customer revenue and ~10% on repeat revenue.

No segmented revenue model
New and repeat customer revenue were not modelled separately, making it impossible to test each stream against historical performance or identify which growth assumptions were realistic.

No benchmarking against actuals
Forecast accuracy was never measured after the year closed, so there was no feedback loop and no basis for refining assumptions over time.

No downside scenario testing
Leadership could not evaluate the target under lower-spend or lower-retention scenarios, so planning trade-offs lacked a quantified data foundation.

Fixed spend made validation essential
A target set under fixed marketing spend cannot be achieved by increasing investment. It must be validated against what existing customer behaviour can deliver within that envelope.

New vs. repeat customer behaviour
Customer acquisition and purchase history modelled separately for each segment, reflecting the distinct revenue dynamics of each group.

Retention and repurchase patterns
Repeat revenue projected using cohort-level retention rates and repurchase behaviour.

Spend, discount, and revenue correlation
Measured relationships between marketing investment, discounts, and revenue, so fixed spend constraints became a genuine model input.

ARIMA/SARIMA statistical modelling
Time-series modelling for seasonal patterns and trend dynamics, producing statistically grounded projections.

Excel-first delivery
Embedded directly into Faherty's existing FP&A workflow with no new tooling required.

~8% forecast error, new customer revenue
Projected using acquisition patterns and spend correlation, measured against full-year actuals.

~10% forecast error, repeat customer revenue
Modelled using cohort-level retention and repurchase data. Sub-10% on the segment most sensitive to behavioural assumptions.

~$98M projection triggered a strategy adjustment
A data-supported view of what the customer base and spend envelope could deliver, giving leadership the basis to adjust strategy before the year began.

Actuals came in at ~$109M by year end
Within the range the scenario model defined as achievable with targeted adjustments.

Planning accuracy is now measurable and improvable
8-10% error rate is the baseline benchmark. Each cycle, the model can be refined where assumptions diverged from outcomes.