Sales forecast, built.
Also called: Demand forecast · Unit projection · Revenue model · Bottom-up sales plan
An estimate of units and revenue over time, built up from real channels and conversion rates, expressed as a range with its assumptions on show.
Start from the bottom: how many people each channel reaches, what fraction realistically buys, what that earns. Show it as a low, mid and high range with the assumptions written down. A defensible forecast you can argue with beats one confident number nobody believes.
What a sales forecast is
A sales forecast estimates how many units you will sell and what they earn, across a defined period. The useful kind is built from the bottom up. You take each channel you can actually reach, estimate the audience, apply a realistic conversion rate, and add up the units. The number that falls out is a consequence of stated assumptions, not a wish.
The trap is the top-down version: “the UK home-baking market is worth millions, we only need one percent.” That sentence has never been built from anything. It starts with a big total and works backwards to a figure that feels modest, which is exactly why it persuades people and exactly why it is useless. One percent of a market you have no route into is zero.
Build it the other way and the forecast becomes a model you can argue with. Reach times conversion times price, per channel, summed, over time. If someone disputes the number, they have to dispute a specific assumption, which is a far more honest fight than trading hunches.
Why it has to be a range
- Conversion is uncertain. A 1.5% store-conversion rate could turn out to be 0.8% or 2.5%. A single point estimate hides that; a low-mid-high range puts it in front of you.
- It forces honest assumptions. To draw three lines instead of one, you have to say out loud what would make the low case happen. That is the most valuable part of the exercise.
- It survives contact with reality. In my experience the mid case is usually optimistic and the low case is what actually happens in month one. Planning against the low case is how you avoid ordering stock you cannot sell.
A worked forecast canvas
The clearest way to build a forecast is to fill five rows honestly, channel by channel. Here is the proofing box’s Year-1 forecast, the one we built in the pilot, so you can see the shape of a good answer rather than a generic template.
Notice the forecast names its own weak points. Every number sits on an assumption you can attack, and the first production run is sized against the low case, so a soft launch costs the team unsold stock rather than the company.
Top-down hope versus bottom-up range
The same Year-1 forecast, framed two ways. The first feels ambitious and tells you nothing. The second is duller, smaller, and actually usable.
- “The market is huge; we only need a tiny slice.”
- One confident number, no range.
- No channel named, no conversion shown.
- Curve bends sharply up in month three for no stated reason.
- Nobody can dispute it because nothing in it is specific.
- Built channel by channel from reach and conversion.
- Low, mid and high case, each a separate line.
- Every assumption written down beside the number.
- Seasonality and warm-versus-cold conversion modelled.
- Disputing it means disputing one named assumption.
The top-down version is the one that gets a room nodding. The bottom-up version is the one that tells you how many units to order. Only one of those keeps the team out of trouble.
How it fits the bigger picture
Sales forecast is activity 10.10.01 in the framework, opening Stage 10 Deliver. It draws on the pricing, audience and channel work from earlier stages, and it feeds straight into the CRM system (10.10.02), which captures the real leads and orders that test whether the forecast holds.
What it can do
It turns a vague ambition into a model. It sizes the first production run against evidence rather than optimism, names the assumptions most likely to be wrong, and gives the team a low case to plan stock and cash against. When the real orders arrive, you can check them against the forecast and learn fast which assumption was off.
What it can’t do
It can’t predict the future, and a forecast treated as a promise rather than a model is dangerous. It also can’t replace a real demand signal. Until pre-orders or actual sales land through the CRM, every conversion rate in it is a hypothesis, and the most likely outcome is that the first month underperforms the mid case.
See the full 10-stage process →
Try it yourself
List every channel you can genuinely reach in the next twelve months. For each, estimate the audience, then apply three conversion rates: pessimistic, realistic, optimistic. Multiply by price, add the channels up, and you have a low-mid-high range. Write the assumption behind each conversion rate next to it. Then size your first order against the low case, not the mid.
Want a structured first pass? Start the Free Sprint → and the GPT will help you frame your channels and demand before you commit to a run.
Your forecast checklist
Project notes: pre-orders before the first run
▸ From the notebook · optional reading
Building the proofing box forecast with Dan and Anna Hartley in Stockport, and the pre-order signal that stopped them ordering 3,000 units on faith.
3 min read · click to open
Dan arrived with a forecast already written: 3,000 units in Year 1, £447k of revenue, a clean upward curve. It was the mid case from a spreadsheet, and he had quietly started treating it as the plan. The first thing I asked was the most awkward: “Where does the curve come from?”
Pulling the number apart
It came from nowhere we could defend. So we rebuilt it from the channels they actually had. DTC traffic was modest and unbought. The real asset was the Sourdough School relationship, a warm list of around 12,000 with maybe 3,000 serious bakers in it. We pushed the conversion rates apart into a range: warm-list buyers at 2% low to 7% high, cold DTC visits an order of magnitude lower.
Three lines came out. Low around 1,400 units, mid around 3,000, high near 4,800. Dan’s original number was sitting on the mid line, which meant his entire plan assumed the optimistic-realistic case landed perfectly. That is a fragile place to commit a first production run from.
The decision that changed
The forecast on its own would not have shifted him. The range did. Once the low case was a visible line at 1,400 rather than a vague worry, the question changed from “how many do we make?” to “how do we find out before we commit?” We agreed a pre-order campaign to the Sourdough School list first, sized the first run at 500–1,000 against the low case, and treated the pre-order conversion as the real test of the warm-list assumption.
The pre-orders came in above the low case and below the mid, which is roughly where I expected. It meant the first run of 750 was right, the BOM at £38–55 held its margin, and nobody had £200k of unsold ceramic in a Stockport unit. The mid-case 3,000 stayed on the page as the Year-1 target, earned channel by channel rather than assumed.
— Deliver stage, project notes, 2026
— Next in Deliver → CRM system