Pet Treats Example
Pet Food & Treats · $25.3M
Forecast accuracy
76% → 82%
Cash released
$235K
Annual benefit
$175K
Cash cycle
−5.5 days
Fill rate
96.1%
Act 1 — Where you are today
Every buying decision trades service against cash.
Buy more and you protect revenue — the product is on the shelf when the order comes. Buy less and you free cash — less of it sitting in the warehouse as working capital. Forecast error is what forces the trade, because the less confident the number, the more stock you have to hold to get the same service. Here is where that leaves you today.
Forecast accuracyest
76.3%
SKU × month, lag-1
Fill rate
95.5%
$531K revenue at risk
Inventory
$3.09M
72 days on hand
Cash cycle
76 d
DIO 72 + DSO 42 − DPO 38
E&O write-off
$87K/yr
2.8% of inventory
Inventory turns
5.1×
COGS $15.7M
The number above is an average, and averages hide the problem
| Portfolio segment | Share of volume | Accuracy today | vs. practical ceiling |
|---|---|---|---|
| Mature SKUs — baseline demand | 78.8% | 80.5% | |
| Mature SKUs — promoted demand | 12.8% | 66.5% | |
| New items — first 12 months | 8.4% | 52.7% |
A single blended accuracy figure lets the hardest part of the portfolio hide behind the easiest. Here the weak point is new items — first 12 months at 52.7% — 8.4% of volume being planned close to guesswork. That is where the trade goes wrong in both directions at once: too much cash in the items you over-call, and lost sales in the ones you under-call.
Where these numbers come from
Read straight off the demo dataset you are about to walk through: 119 active SKUs, 702,624 units shipped in the trailing 12 months (anchor 2025-12), forecast 769,075 units for the next 12 (+9.5%), portfolio split Dog 58.6% / Cat 41.4%. Dollars come from an assumed average selling price of $36.00 per unit — change it below and every figure on this page moves with it.
Act 2 — What the model actually does
Six things, each of which you are about to see working.
This is not a black box. Each capability below is a page in this tool, and each one closes a measurable share of the gap between where your forecast is now and the best it can realistically be.
Dashboard
Statistical baseline, trend & seasonality
+2.2 pts
Decomposes every SKU into level, trend and seasonal shape instead of carrying last year forward with a growth factor. Removes the single largest source of ordinary error.
Dashboard — review table
Bias detection & correction loop
+1.2 pts
Scores every SKU on rolling accuracy AND bias, then ranks the items that actually ran you short or long so a planner fixes the twenty that matter, not all two thousand.
Top-Down + Bottom-Up
Top-down / bottom-up reconciliation
+0.9 pts
Forces the finance number, the sales number and the sum of the SKU forecasts to be the same number. Kills the version-of-truth argument that quietly inflates buys.
Smart Sales Stats
Promotion & event lift modeling
+0.8 pts
Separates baseline demand from promoted demand and learns the lift each mechanic delivers, so a planned promo raises the buy by the right amount instead of a guess.
New SKUs
Attribute-based new item forecasting
+0.5 pts
Builds a launch curve from analogous SKUs that share attributes, rather than from a number someone typed into a launch deck. This is the single biggest accuracy win in most portfolios.
Cannibalization
Cannibalization & portfolio overlap
+0.5 pts
Models how a new item steals from existing ones instead of treating its volume as purely incremental — the classic reason a launch year ends with both a stockout and a write-off.
BOM Demand
BOM-level component demand
Explodes the finished-good forecast through the bill of materials so raw material and component buys move with real demand, instead of each buyer holding their own private buffer.
$17K of component stock
Act 3 — What good looks like
76.3% → 82.4% +6.0 pts
Improvement is modeled as closing the gap to a realistic ceiling, not as a flat promise. A company already forecasting well gets a smaller number here — that is the model working correctly.
How the points are earned
76.3% starting point
+2.2 pts Dashboard
+1.2 pts Dashboard — review table
+0.9 pts Top-Down + Bottom-Up
+0.8 pts Smart Sales Stats
+0.5 pts New SKUs
+0.5 pts Cannibalization
82.4% achievable
The amber line is the practical ceiling for your industry (92%). Demand noise makes the remaining gap genuinely unforecastable — anyone promising to close it is selling you something.
Where the gain actually lands
| Segment | Before | After | Lift | |
|---|---|---|---|---|
| Mature SKUs — baseline demand | 80.5% | 85.1% | +4.7 pts | |
| Mature SKUs — promoted demand | 66.5% | 77.0% | +10.6 pts | |
| New items — first 12 months | 52.7% | 64.7% | +12.0 pts |
The blended figure understates the story. The biggest movement is always in new items and promoted volume — the two segments that drive your worst buying decisions.
Error is what you actually buy inventory against
Forecast error today
23.7%
the number safety stock is sized on
Forecast error after
17.6%
same service level, less buffer
Error removed
25.5%
the lever on every number below
Safety stock scales with forecast error, not with forecast accuracy. Moving accuracy from 76.3% to 82.4% looks like +6.0 pts — but it removes 25.5% of the error you are currently buying protection against. That asymmetry is the whole business case.
This is not a projection you have to take on faith.
The model above says this portfolio should reach 82.4%. On the dashboard you are about to open, the tool is actually delivering 83.6% at lag-1 on exactly this data, with bias held to +0.5 pts. Rolling accuracy runs 89.3% at three months. The estimate and the live result agree within 1.2 points.
Act 4 — What it is worth
$409K in year one.
One result, measured three ways. Different functions are held to different numbers and paid on different outcomes, so the same improvement has to be shown in whichever terms the person in the room is accountable for.
Working capital released
$235K
one-time cash out of inventory
Annual P&L benefit
$175K
0.69% of revenue
Cash conversion cycle
−5.5 d
76 → 71 days
Year-one cash impact
$409K
release plus annual benefit
Annual benefit, built up line by line
| Carrying cost on inventory no longer held | $45,046 | $235K × 19.2% (obsolescence excluded — claimed below) |
| E&O / obsolescence avoided | $42,682 | 49% of $87K written off today |
| Margin recovered from fewer stockouts | $28,205 | $74K revenue × 38% margin |
| Expedite freight avoided | $58,766 | 12% of freight is demand-miss driven |
| Total annual benefit | $174,700 |
Program cost
$65K/yr
platform plus services
Return on program
2.7×
annual benefit ÷ annual cost
Payback
4.5 mo
before the one-time cash release
Net annual benefit
$110K
after program cost
Read this before you quote these numbers
Starting forecast accuracy was estimated from industry benchmarks and your portfolio shape. Replacing it with your measured SKU-level number is the single highest-value input on this page.
Your data and process maturity inputs reduce the achievable lift. The gap between this number and the best case is mostly a sales and marketing feedback problem, not a modeling problem.
Act 5 — Now put your numbers in
Fill in what you know. We will benchmark the rest.
Nothing here is required. Anything you leave alone uses a mid-market benchmark for your industry, and everything above recalculates as you type.
Who you are
Optional. We read your site for categories, product lines and sizes. Add anything it would miss — or name them outright, one per line: categories: Pantry, Kitchenware, Merch
Steady repeat demand, heavy retailer promo calendars, protein-driven raw material buys with real shelf-life exposure.
$
%
Revenue less COGS. Type it as a percentage, e.g. 38.
Portfolio shape — this is what makes forecasting hard
More SKUs means a longer low-volume tail.
Items launched in the last 12 months. The single biggest driver of poor accuracy.
Distinct deals or events across the portfolio. Promoted volume is structurally harder.
How far ahead you must commit cash.
1.0 is flat; 2.0 is a sharp season.
Higher means noisier month-to-month demand.
Working capital
$
Leave at 0 to derive it from COGS and days on hand.
%
Share of inventory sitting in components. Set 0 if you are a pure distributor.
%
Capital, storage, insurance, shrink, obsolescence.
Current performance — leave blank if you do not measure it
%
SKU × month, lag-1.
%
Case or line fill achieved.
$
Obsolete and expired stock.
Most companies quote an aggregate annual accuracy number that flatters them by 10–15 points. If yours is not measured at SKU × month, leave it blank — the benchmark will be closer to the truth than the number in the deck.
Data & process maturity — how much of the lift is actually reachable
Achievable lift multiplier: 0.79×
You are leaving accuracy on the table before the model even runs. The switches above are worth more than any change to the algorithm — a forecast cannot predict a promotion nobody told it about. Turning all of them on takes this scenario to 85.3% accuracy and $350K of released cash.
How do you want to spend the improvement?
At this setting you release $235K of cash and move fill rate to 96.1%. Slide left to buy service instead of cash.
$
Platform plus services. Used only for the ROI and payback figures.
Act 6 — Make it your data
See it running on your business.
The numbers above are the case. The rest of the tool is where you would actually work — and it is far easier to judge when the SKUs, categories, seasonality and lead times on screen are yours. Enter your business above and every page rebuilds around it, so you can see how your portfolio would be planned rather than someone else's.
Pet Treats Example
Pet Food & Treats · $25.3M revenue · 119 SKUs · 17 launches/yr
Accuracy 76.3% → 82.4% · $409K year-one impact
Build my business takes what you entered above — your industry, SKU count, launch cadence, promotional intensity, seasonality, demand variability and lead times — and builds a portfolio that behaves like yours behind every page: products named in your industry's own vocabulary, four years of demand history, a scored backtest so the accuracy you see is measured rather than claimed, a 12-month forecast, a multi-level bill of materials, and an inventory position with open POs, customer orders and a buy plan. Takes about 15–25 seconds, and rebuilds identically every time, so you can come back to the same numbers.