The fashion world has always loved people who stick to their gut. “I just knew this would be a big seller.” “I woke up one day and it popped into my head, and then we had the new It Bag.” The problem is for every time someone says this there are thousands of times where their gut was wrong.

Fashion suffers from a problem of blindly following the gut. Data is often seen as an anathema that interrupts the creative process. The gut manifests itself from design to sourcing to buying to selling. No wonder the retail world is struggling and inventories are up. Nothing ends creativity like running out of money to make clothes. But data can help.

Below is an outline for running a new type of fashion brand that’s entirely data-driven. While the gut is still important—data doesn’t make decisions, it only informs them—decisions are no longer made blindly. Some of this might seem crazy, but the point is simple: the data won’t lie. Data-based learning comes from a company’s culture. If the culture doesn’t justify and reinforce the importance of data, no one will value it. But if it’s an integral part of the process, it will pay dividends.

Picking the brand

Most brands start because someone either wished they had a specific product or because they saw a hole in the market. This is healthy. The problem is that people often decide right away to focus on one specific idea, often with no validation that people actually want it.

Instead of picking one brand and diving in, come up with a handful of ideas and test them all. Make a Squarespace for each, throw up some photos, a rough brand identity and a clearly defined value proposition. Hook it up to Optimizely to run your A/B tests through. Run $300 of Facebook ads for each and see which converts the best. Maybe the call to action is signing up for a wait list. Maybe it’s filling out a survey. Maybe it’s seeing if someone will get to checkout. Test all that you can. You will quickly find out which brands resonate and which do not.

Finding the customer

So you found a brand and value proposition that works. Now you need to figure out the best way to acquire that customer. Target different groups of people with Facebook ads. Find similar brands and target people who like those brands. Run some Instagram ads too. Try some Google Adwords ads. The goal is to find the most efficient channel and customer.

Let’s say you want to start a womenswear brand that uses technical materials. Maybe you test it out among 18-29 year olds, only to find out that the brand resonates more among 25-35 year olds. Information like this is incredibly valuable, and will affect every process. Test price points, colors, messaging, you name it. Test if the product is something a guy buys this for his girlfriend or a girl buys for herself. Everything important is testable.

Finding the right products

Now you have a brand and a general price point figured out, maybe more. Next you need to test out different products and more specific price points. Find some product shots that look similar to what yours would be and throw them up on the site. Use Optimizely to partition visitors to land on different pages with different price points for the same product.

Test demand for different sizes, what colors people want, what they would pay for shipping, what return policies they are comfortable with, what shipping times are suitable and so on.

What’s amazing is that you have not developed or produced a single physical item at this point, which means you haven’t sunk thousands of dollars into something that might not work. These are all just tests. The potential customer doesn’t know they are tests, but your risk is so low right now because you’re just messing with Squarespace and Optimizely.

Minimizing inventory risk

Now you have a ton of data about price points and colors and products. It’s time to make something real. Many people (especially buyers) will say that early fashion brands need full collections and tons of SKUs. This is mostly false. If you make a better tee shirt than everyone else, you don’t need to make pants or coats. Just produce what you have tested and can execute on. This is likely one or two SKUs and maybe a few colors. It will look like you made too little, but this is good.

Holding inventory is one of the riskiest parts of running a brand that sells physical things. Inventory is dead money. It exists in its current form and that will never change. You can’t unmake a shirt or an iPhone. It also will never be worth more to you. It’s value can only decrease. Inventory is pure risk. Inventory rots. Minimizing your inventory burden is always important, but it is crucial early on. Sinking too much money into the wrong item could be disastrous.

Now that you know what product or two you will be making, you need to plan your inventory. Luckily, you have a ton of data to help. You might know that you had a lot of interest for Smalls and Mediums so you can make more of them. You know that people really liked the color red. This is crucial info for inventory planning.

Early on, under forecast your demand. Selling out is good. If you think you can sell 50 pieces, make 25. That will be enough work for you. But since you have meticulously tested each aspect of your offering, there’s now a much better chance these items will sell since you’ve demonstrated demand for it.

And the best part is that your only costs here are only $12 a month for Squarespace and whatever you spent on Facebook ads. Optimizely, which is an amazing tool, is free for simple uses such as this.

You’ve spent a minimal amount of money (under a thousand dollars) and gathered an immense amount of data. This is infinitely more valuable that your gut telling you to make 100 pieces of one SKU that might not sell because you’re just relying on your gut.

Everything described above doesn’t need to end after you launch. This is a mindset for every process at every stage of a company.

Data is the best.