Business Forecasting & Predictive Analytics
Haven & Grain · Outlook to FY2027
Anita Analytics | Data & Business Intelligence | October 2026
HAVEN & GRAIN
Executive summary
Revenue is on track for $17.9M this year, with a $19.9M base case for FY2027
2
$17.9M
FY2026 revenue (forecast)
+13% vs FY2025
$19.9M
FY2027 base forecast
+11% vs FY2026
2.0%
Monthly revenue MAPE
Target below 5%
65,200
Q4 2026 orders forecast
+35% vs Q3
How reliable
Revenue forecasts landed within 3% every month this year, with a small 1% under-forecast.
Where it misses
Furniture is over-forecast by 3.4% and textiles under-forecast by 4.5%, feeding both excess stock and stockouts.
What we recommend
Plan Q4 for 65,200 orders, rebalance purchasing toward fast movers, and re-forecast on clear triggers.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Agenda and forecasting scope
From revenue outlook to demand, accuracy review and the decisions that follow
3
1
Sales & revenue forecasts
FY2026 outlook, FY2027 scenarios and category view
2
Demand forecasting
Weekly Q4 demand, seasonality, drivers and models
3
Forecast versus actual
Accuracy, bias and where forecasts missed in 2026
4
Assumptions & implications
Sensitivities, scenarios and business actions
HORIZON
Oct 2026 to Dec 2027
HISTORY
36 months of orders and revenue
GRANULARITY
Weekly demand, monthly revenue
METHODS
ETS, SARIMAX and boosting ensemble
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Sales & revenue forecasts
What the data says about revenue through FY2027
SECTION 01
HAVEN & GRAIN
Revenue is up 13% and tracking ahead of 2025
Q4 carries about a third of annual revenue; October to December is forecast at $6.0M
5
+13.1% year to date
Jan to Sep revenue of $11.94M against $10.56M in 2025.
One third lands in Q4
Q4 is 33% of annual revenue in both 2025 and the 2026 forecast.
Q4 forecast: $6.0M
October $1.6M, November $2.1M and December $2.3M, up 13% on 2025.
Source: OMS and finance ledger. Oct–Dec 2026 values are forecasts.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
FY2026 revenue is forecast at $17.9M
Every quarter is growing 11% or more year on year
6
$17.9M
FY2026 revenue (forecast)
+13.1% vs FY2025
Quarter | Growth vs 2025 |
Q1 | +13.2% |
Q2 | +15.0% |
Q3 | +11.3% |
Q4 (forecast) | +13.2% |
Source: finance ledger and forecast model. Q4 2026 is forecast.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
FY2027 revenue range: $19.0M to $20.8M
The base case of $19.9M assumes growth slows from 13% to 11%
7
$19.0M
Low case, +6% on FY2026E
$19.9M
Base case, +11% on FY2026E
$20.8M
High case, +16% on FY2026E
Source: forecast model, scenario runs. FY2026E base is $17.9M.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Growth shifts toward lighting and textiles
Faster-turning categories lead FY2027 growth; furniture and wall art grow slowest
8
Category | FY2026E | FY2027F | Growth | Main driver |
Furniture | $5.4M | $5.8M | +7% | Mature range, supply-constrained, modest price rises |
Lighting | $3.6M | $4.1M | +14% | New brass and rattan lines, higher social reach |
Textiles | $3.9M | $4.6M | +18% | Repeat purchases and room-bundle sales |
Ceramics | $2.6M | $2.9M | +12% | Wholesale channel added in H2 |
Wall art | $2.4M | $2.5M | +4% | Saturated range, limited refresh planned |
Total | $17.9M | $19.9M | +11% | Base case |
Lighting and textiles add $1.2M of the $2.0M total increase, and they turn over fastest.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Demand forecasting
Weekly order demand, seasonality and the drivers behind it
SECTION 02
HAVEN & GRAIN
Q4 demand peaks at 8,600 orders in one week
Black Friday week runs at about twice a normal week, within an 80% interval of ±12%
10
Black Friday week
8,600 orders, about double a normal November week.
65,200 orders in Q4
35% above Q3; the plausible range is 60,000 to 70,000.
Capacity risk
Pick and pack already runs at 91%, so peak weeks need extra shifts.
Source: demand forecast model, run 12 October 2026. Interval is 80% prediction range.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Seasonality: Q4 carries a third of demand
November and December run 42% and 58% above an average month
11
Peak: December at 1.58
Gift and hosting demand lifts lighting, ceramics and textiles most.
Trough: February at 0.72
A natural window for stocktakes, range reviews and process work.
Ramp starts in October
Demand rises 16% above average before the November peak arrives.
Source: 36-month order history, ratio-to-moving-average decomposition.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Seasonality and promotions drive the forecast
A third of the model's signal comes from levers we control: promotions and social content
12
33%
of the signal comes from promotions and social reach, which we can plan
Uncontrollable drivers, such as macro and weather, make up 17% and are treated as scenario inputs.
Source: gradient boosting feature importance, weekly category-level model.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
From raw data to published forecast
A weekly pipeline blends three models, with planner review before release
13
1
Data
Orders, traffic, promotions and prices, 36 months
2
Features
Calendar effects, lags, price and event flags
3
Models
ETS, SARIMAX and gradient boosting run in parallel
4
Ensemble
Weighted blend based on recent backtest error
5
Review
Planner review with documented overrides
36 mo
of order and traffic history feed every run.
Weekly
refresh each Monday, with monthly revenue roll-ups.
1,860
active SKUs rolled up to category and site.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
The ensemble beats every single model
A weighted blend cut backtest error to 4.1%, against 9.8% for a seasonal baseline
14
Model | Backtest MAPE | Strength | Limitation | Role |
Seasonal naive | 9.8% | Simple benchmark | Ignores promotions and trend | Benchmark |
Holt-Winters (ETS) | 6.2% | Captures trend and seasonality | Weak on promotions | Component |
SARIMAX | 5.4% | Adds price and promotion inputs | Needs stable history | Component |
Gradient boosting | 4.6% | Handles many drivers | Weak on new SKUs | Component |
Ensemble (selected) | 4.1% | Blends all three models | Harder to explain | Production |
Backtest: rolling 12 months, weekly order demand at category level. MAPE = mean absolute percentage error.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Forecast versus actual performance
How accurate were we in 2026, and where did we miss?
SECTION 03
HAVEN & GRAIN
Revenue forecasts landed within 3% every month
Jan to Sep 2026: $11.82M forecast against $11.94M actual, a 1.0% under-forecast
16
2.0%
Monthly revenue MAPE, target below 5%
+1.0%
Actual above forecast over nine months
9 of 9
months landed within ±3% of forecast
Source: finance ledger and December 2025 forecast archive. Variance = (actual − forecast) ÷ forecast.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Errors are small and lean one way
Seven of nine months beat forecast; May and September are the two misses
17
Small, steady under-forecast
Seven months beat forecast by 1.4% to 2.7%; the model is slightly cautious.
May: promotion fell short
A spring event drew less traffic than modelled, 2.1% below forecast.
September: supply limits
Furniture stockouts held revenue 2.0% below forecast.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Category accuracy shows where bias hides
Textiles and lighting are under-forecast while furniture and wall art are over-forecast
18
Category | Weekly WAPE | Actual vs forecast | What it means |
Textiles | 5.2% | +4.5% | Under-forecast: stockout risk |
Lighting | 5.8% | +3.0% | Under-forecast: stockout risk |
Ceramics | 7.1% | +1.5% | Slightly under-forecast |
Wall art | 8.4% | −2.6% | Over-forecast: trim buys |
Furniture | 11.6% | −3.4% | Lumpy demand, over-forecast: trim buys |
Bias is the fixable problem. Correcting the category baselines targets the same imbalance that leaves furniture with 114 days of stock.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Planner overrides added error, not accuracy
Overrides raised error from 4.1% to 4.8%, so each one needs a documented reason
19
−5.7 pts
Error cut by the ensemble vs naive
+0.7 pts
Error added by manual overrides
Keep overrides for known events, such as a planned launch, and log the reason so we can test it later.
Source: forecast accuracy log, 12-month backtest. Forecast value added = error vs seasonal naive.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Assumptions, uncertainty & implications
What could change the forecast, and what to do about it
SECTION 04
HAVEN & GRAIN
Forecast assumptions and tested ranges
Seven drivers underpin the base case; each is stress-tested in the scenarios
21
Assumption | Base case | Tested range | Evidence |
Site traffic growth | +14% year on year | +8% to +18% | 36-month trend, social reach |
Conversion rate | 2.9% | 2.6% to 3.2% | Trailing 12-month average |
Average order value | $90 (Q4: $92) | ±$4 | Q3 actual of $90 |
Promotion depth | 15% off, 4 events | 10% to 20% | 2025 promotion calendar |
Return rate | 6.9%, easing to 6.0% | 6.0% to 8.0% | Q3 actual of 6.9% |
Supplier lead time | 38 days | ±8 days | Supplier on-time-in-full 86% |
Macro demand | Stable | −5% to +3% | Consumer confidence index |
Source: forecast model configuration, October 2026.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Traffic and conversion swing the outcome most
Impact on FY2027 revenue in $M when each driver moves across its tested range
22
Top two drivers: 53%
Traffic and conversion make up more than half of the total swing.
Supply risk is lopsided
Delays cost $0.5M, while faster supply adds only $0.2M.
Content is the lever
Social and search content feed traffic, the largest single driver.
Source: one-at-a-time sensitivity runs on the FY2027 base case ($19.9M).
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Three scenarios for FY2027 revenue
The base case carries 60% weight; the 80% interval runs from $19.0M to $20.8M
23
DOWNSIDE | 20%
$19.0M
+6% vs FY2026E
BASE | 60%
$19.9M
+11% vs FY2026E
UPSIDE | 20%
$20.8M
+16% vs FY2026E
Probability-weighted expected value: $19.9M, matching the base case.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
What the forecast means for the business
Four operating decisions follow directly from the Q4 and FY2027 forecasts
24
Inventory and purchasing
Build Q4 cover for 65,200 orders, trim furniture and wall art buys by about 15%, and add textile and lighting stock.
Staffing and capacity
Plan about double normal pick capacity for Black Friday week and add temporary shifts at the pack stations.
Marketing and content
Weight spend and social content toward lighting and textiles, and publish campaigns four to six weeks ahead.
Cash and finance
Working capital need peaks in October and November; align supplier payments and hold a buffer for the downside case.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Monitoring and re-forecast triggers
Clear thresholds tell us when the forecast needs a refresh, not just a review
25
Metric | Healthy | Trigger | Action |
Weekly demand MAPE | Below 6% | Above 8% for two weeks | Retrain models, check inputs |
Rolling four-week bias | Within ±3% | Beyond ±5% | Adjust category baseline |
Monthly revenue variance | Within ±3% | Beyond ±5% | Re-forecast the quarter |
Planner overrides | Reason logged | Any without a reason | Review in forecast meeting |
Traffic and conversion | Inside tested range | Outside for three weeks | Re-run the scenarios |
Weekly
Demand refresh every Monday, with the accuracy log updated.
Monthly
Revenue re-forecast and variance review with finance.
Quarterly
Assumptions, scenarios and model weights refreshed.
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026
Decisions requested
Three decisions turn the forecast into action
1
Approve the Q4 capacity and stock plan
Plan for 65,200 orders, a peak week of 8,600, and headroom toward the upper interval.
2
Rebalance purchasing toward fast movers
Trim furniture and wall art buys by about 15% and add textile and lighting stock.
3
Adopt the monitoring rules
Weekly refresh, monthly re-forecast, trigger-based reviews and documented overrides only.
Appendix: metric definitions
Accuracy metrics used throughout this report, calculated from the forecast archive
27
Metric | Definition |
MAPE | Average of |actual − forecast| ÷ actual across periods |
WAPE | Sum of absolute errors ÷ sum of actuals, weights large periods more |
Variance vs forecast | (Actual − forecast) ÷ forecast; positive means actual beat forecast |
80% prediction interval | Range expected to contain the actual outcome 80% of the time |
Forecast value added | Error of a method minus error of the seasonal naive benchmark |
Seasonality index | Month average ÷ average month, so 1.00 is typical |
Haven & Grain | Business Forecasting & Predictive Analytics | Oct 2026