When I went back into my 2024 Filgap spreadsheet, I noticed something I had almost forgotten.
I had started with an annual revenue number.
The spreadsheet then divided that total into product categories, spread the sales across the months, applied prices and calculated how many units those sales would represent.
It looked quite detailed once all the formulas were working.
But the annual number at the top was still a rough number.
At the time, I was trying to get the overall business plan into shape and I had not finished every product price in enough detail. Starting with a sales total gave me a quick way to see the size of the business and continue building the rest of the model.
For that purpose, it helped.
If I were doing the detailed forecast now, I would still look at the big number early. I just would not want to leave it there without asking what has to happen underneath it.
How many people actually buy? What do they buy? At what price? How often do they come back? Does the answer change depending on the season?
Those questions do not make the forecast certain. They at least make the uncertainty easier to see.
I still look at the big picture first
I normally want some sense of scale before I spend too much time on detail.
With a café, that could mean looking at the location, possible customer traffic, the number of seats and how often those seats might turn over. I can compare that rough picture with what I know about the surrounding market and ask whether the sales level I am imagining is even plausible for that kind of operation.
With a Farmers’ Market, I would look at different things.
How many people usually come through on a Saturday? What seems to happen at the food booths? How busy are they at different times? I can watch those things before Filgap starts selling and get some feeling for the scale.
That is still only the top view.
Once I start going into the details, I prefer to work from the bottom too. Products, prices, quantities, customer count, average purchase, selling days. Depending on the business, I can approach the same sales number in several different ways.
They may not give me exactly the same answer.
I do not think that is necessarily a problem. If one approach gives me a number that is far away from the others, I want to know why.
That difference may be more useful than forcing all of the models to agree.
For Filgap, I can start with one Saturday
The first Farmers’ Market forecast does not need to begin with a yearly sales target. I can start with one market day.
- How many people come to the market?
- How many actually come near the booth?
- How many try a sample?
- How many buy something?
- When they buy, is it usually one bag or more than one?
- What does the average purchase look like?
I do not know those numbers yet.
Before selling, I can still gather some clues. I can go to the market and watch how people respond to other food vendors. I can see where people stop and where they do not. That gives me more than sitting at home and inventing a conversion rate.
But I also have to be careful with what I compare.
Someone selling food that can be eaten immediately is not doing exactly the same thing as someone selling whole-bean coffee for people to take home. From what I have observed, whole bean by itself does not automatically sell at the same pace as ready-to-eat food or drinks.
That matters to me because whole bean is the centre of the current Filgap plan.
So market traffic alone is not enough. I need to learn how many of the people passing through are actually interested in coffee for home, how many are willing to taste it, and how many decide to buy it.
At the beginning, those are all assumptions.
What I really need to learn is whether people come back
A first purchase is useful information, but repeat customers are much more important to the business I am trying to build.
A visitor can buy one bag while passing through Squamish. I am happy if that happens. But the part I care about more is whether local people who make coffee at home decide that they want another bag later.
I have not decided on one formula for repeat purchase.
I could divide customers into groups with different buying frequencies and see what kind of sales range that produces. I could also look at a household’s coffee consumption and work back from there. There are probably several reasonable ways to test the assumption.
I would rather look at more than one of them than choose one repeat rate because the spreadsheet needs a percentage.
This is also why I do not want to read too much into the first few market days.
I am building the website before launch. I plan to use social media, and I want people to sample the coffee before regular selling begins. If that work goes well, there may be some curiosity when Filgap first appears at the market.
That would be good, but it could also make the first weeks unusual.
For myself, I would rather watch something closer to three months before I start feeling that the early pattern means very much. I want to see whether some of the people who bought once actually return.
Three months is not a rule. It is simply a period that feels more useful to me than judging the business from the first four Saturdays.
Some assumptions will be weaker than others
Product mix is one of them.
In previous coffee businesses, I used market research and operating experience to make an initial sales-mix assumption. Then the actual numbers changed it.
I expect the same thing here.
Whole bean is the clearest product in the current Filgap plan. Other products may become important, but they have less real sales evidence behind them right now. I do not want to give them large, confident percentages just to make the spreadsheet complete.
Seasonality has the same problem in a different way.
The Farmers’ Market does not run in exactly the same way all year, and Squamish itself changes through the seasons. Visitor traffic is different in summer, and there are periods when the market is not operating in the same way.
I can put a first seasonal assumption into the model because I need something to plan with. I just have to remember what it is.
After the first full selling cycle, I should have a much better basis for changing it.
At this stage, roasting capacity is not the first limit I see
A sales forecast still has to make operational sense.
For the Farmers’ Market stage, though, I do not currently think roasting capacity is the main problem. With one selling day a week, the roasting setup I am planning around gives me enough room for the volume I can realistically imagine at the beginning.
If sales are stronger than expected, my own time may become the more immediate issue. More roasting means more packaging, more market preparation and more hours of work.
That is different from a future permanent roastery and tasting room. In that situation, service speed, equipment, staffing and the space itself may create more obvious limits.
For the first forecast, I mainly want to ask a simple question after I estimate demand: if this number happened, could I actually prepare and sell that much?
One forecast is not enough for me
I still need a base forecast. I need something to use when I think about inventory and the rest of the plan.
But I also want to see another version where sales are better, and one where they are worse.
The lower case matters a lot because that is where I start thinking about what I would have to do if the business takes longer to build than I hoped. I do not want to go too far into that here because the risk side deserves its own discussion later.
For now, it is enough to change some of the assumptions and see how much the sales number moves.
Maybe fewer people buy. Maybe the product mix is different. Maybe repeat purchasing takes longer. Maybe one part of the season is weaker than I expected.
The useful thing is that I can see which assumptions are doing most of the work.
Then selling starts, and the model changes again.
I would not rewrite everything after the first week. I would rather collect actual market days and see what happens over a little more time: visitors, samples, purchases, bags per order, product mix, repeat customers.
Before launch, I am already trying to make the blank spaces a little less blank. That is part of the reason for the website, social activity, market observation and sampling. None of those things tells me what future sales will be. They give me something to work with before I have actual sales data.
After selling begins, I want the spreadsheet to contain less guessing than it did before.
That is probably the main difference between a forecast I can use and one that only looks impressive to me.
The first one does not have to be right. I need to be able to see what I assumed, watch what actually happens, and change the number when the business gives me a better answer.
