How a Beauty Startup Cut Initial Inventory Risk by About 65% by Building Demand Before Stock

A friend of mine was preparing to launch an online cosmetics store with roughly $30,000 CAD in initial capital.

The obvious approach was straightforward: choose a broad assortment of products, purchase inventory, build the website, launch advertising, and see what sold.

On paper, it looked reasonable.

In practice, the first inventory order could have consumed close to $20,000 CAD before we had learned anything meaningful about customer demand.

If the store launched with around 70 to 80 SKUs and carried only 8 to 12 units of each product, a significant portion of the available cash would already be sitting on shelves.

That was the problem.

Beauty is an unusually difficult category for a new retailer to predict. There are thousands of possible SKUs, trends can change quickly, established retailers already compete heavily on price and attention, and a product that looks attractive to the founder may have very little actual demand.

We therefore reversed the normal sequence.

Instead of asking:

What products can we source and sell?

We started with a different question:

What are customers already showing us they want to buy?

That change eventually transformed the launch model from:

Buy → Stock → Advertise → Hope it sells

into:

Discover demand → Validate → Build audience → Build distribution → Stock winners → Scale

We treated inventory as the last major commitment, not the first

Our first recommendation was simple: do not buy inventory yet.

We began by studying best-selling products across several large beauty retailers and marketplaces.

We reviewed roughly 120 frequently appearing products across six major categories, including complexion, lips, eyes, skincare-adjacent beauty products, tools, and trend-driven accessories.

We were not looking for products to copy mechanically.

We were looking for repetition.

If similar products, formats, shades, price points, or use cases appeared repeatedly across multiple bestseller lists, there was at least some observable evidence that customers were spending money in that area.

After the first review, we reduced the universe from more than 100 potential products to a shortlist of approximately 25 products worth testing.

That was already useful.

Instead of allocating capital across 70 or 80 assumptions, we now had 25 hypotheses.

The next step was to determine which of those hypotheses belonged to the customer we actually wanted.

Choosing a customer narrowed the product strategy

Originally, the customer definition was essentially "people who buy cosmetics."

That was too broad to be operationally useful.

A 19-year-old university student buying a viral lip product behaves very differently from a 38-year-old professional buying premium skincare. The products may differ, but so do the acceptable price points, content formats, influencers, purchase triggers, and channels through which trust is built.

We chose to focus initially on women roughly 16 to 23 years old.

That decision immediately narrowed several other decisions.

Instead of attempting to compete across the entire beauty market, we concentrated primarily on products that could sit within an accessible $15 CAD to $45 CAD retail price range.

We also designed the early assortment around products that were visually demonstrable and content-friendly. A product that could be shown in a tutorial, comparison, before-and-after video, event demonstration, or user-generated post had more strategic value than a product that relied entirely on paid advertising to explain itself.

The goal was not to remain permanently within one age group.

The goal was to create product-audience fit somewhere before trying to create it everywhere.

We tested 25 products without putting 25 products into inventory

Normally, a retailer validates products after purchasing them.

We wanted to validate as much as possible before purchasing them.

We built landing pages and product pages for the shortlisted products and used several forms of commitment:

• Notify Me
• Waitlist registration
• Pre-order
• Add to Cart
• Checkout initiation

This distinction was important.

Asking someone whether they like a product produces weak evidence.

Getting them to click on it produces better evidence.

Getting them to leave their contact information is stronger.

Getting them to begin checkout is stronger again.

And getting them to pay is the strongest signal available.

Over the initial testing period, approximately 7,500 people visited the test pages through a combination of paid traffic, social content, creator exposure, and early organic distribution.

Around 2,100 visitors moved from discovery pages into specific product pages, giving us a product-level intent signal.

About 800 users joined a waitlist or requested product notifications.

Most importantly, the tests generated approximately 145 paid pre-orders.

At an average pre-order value of roughly $42 CAD, that represented just over $6,000 CAD in validated customer spending before we had committed to a broad inventory position.

The distribution of that demand was much more important than the total revenue.

Seven products generated most of the demand

The 25 products did not perform evenly.

That was exactly what we hoped to discover.

The top seven products generated roughly two-thirds of paid pre-orders, while the weakest group of products attracted clicks but produced very little meaningful purchase intent.

Several products looked strong when measured by social engagement but weak when measured by checkout behavior.

That distinction changed our view of what a "popular product" actually meant.

A product with 2,000 video views and almost no purchase intent was less interesting commercially than a product with 600 views and a disproportionately high number of waitlist registrations and pre-orders.

We therefore created a simple hierarchy of demand signals.

Views were interesting.

Clicks were useful.

Waitlist registrations were meaningful.

Payments mattered most.

Products that performed poorly on the stronger signals were removed before inventory was purchased.

The initial inventory order was therefore concentrated primarily around seven validated SKUs, rather than spread across 70 or 80 speculative products.

Our first meaningful inventory commitment was approximately $6,500 CAD.

Under the original broad-assortment plan, we estimated that the store could easily have committed around $18,000 CAD to $20,000 CAD to its opening inventory.

Demand validation therefore reduced the amount of capital exposed in the first inventory cycle by roughly 65%.

The objective was not simply to spend less.

It was to preserve capital until we had earned the right to deploy it.

The first inventory cycle gave us another signal

The validation process did not end once products were purchased.

Pre-orders and waitlists told us what customers intended to buy. Actual inventory movement told us whether those intentions translated into sustained demand.

Within roughly the first eight weeks after the validated products became available, approximately 75% of the initial units had sold.

Five products justified relatively quick replenishment.

Two products sold adequately but more slowly than expected.

The difference mattered because we now had a repeatable rule.

A product did not remain in the assortment because we liked it or because a supplier offered favorable terms.

It remained because customers continued to justify the working capital allocated to it.

This started to turn inventory purchasing from a merchandising opinion into an evidence-based capital allocation decision.

While testing products, we were also building the audience

There was another problem with the traditional sequence.

A store that buys inventory first and starts marketing afterward has two risks at the same time.

It does not know exactly what customers want, and it does not yet have reliable access to those customers.

We wanted to reduce both risks before scaling.

So while product testing continued, we began building a small beauty-focused audience around the chosen customer segment.

The content was not designed simply to advertise products.

It included tutorials, product comparisons, makeup trends, short educational videos, creator collaborations, customer demonstrations, and user-generated content.

Within the first several months, the brand built an audience of roughly 6,000 followers across its primary social channels, alongside an owned email and notification list of approximately 1,400 people.

That audience was small compared with established beauty brands.

But it was strategically useful because it was concentrated around a specific customer profile.

More importantly, it gave every future product test a lower-cost starting point.

We no longer needed to purchase every first click.

When a new product was being considered, part of the validation traffic could come from people who had already interacted with the brand.

We were gradually building distribution before expanding inventory.

The next question was harder: how could a small beauty retailer compete for attention?

Product validation reduced inventory risk, but it did not solve the larger competitive problem.

Cosmetics is crowded.

A small retailer can stock a good product and still disappear among hundreds of stores selling similar products.

We therefore stopped thinking of marketing purely as online advertising.

The more important strategic question became:

Could distribution itself become part of the business model?

One experiment emerged from that question: small, targeted beauty events.

We designed events to do four jobs at once

A normal promotional event might be judged primarily by immediate sales.

We wanted ours to produce several forms of value simultaneously.

Attendees could test products.

Creators and customers could produce content.

We could observe reactions to products in person.

And the brand could build trust in a category where trial and recommendation strongly influence purchase behavior.

The early events averaged roughly 40 to 45 attendees each.

Across the first four events, approximately 170 people attended, many of them closely matching the 16 to 23-year-old target segment.

The direct economics were not spectacular on their own.

The events cost roughly $500 CAD to $700 CAD each after venue collaboration and basic activation expenses.

Across four events, approximately 60 first purchases could be connected to attendees within the event or shortly afterward.

That put direct first-purchase acquisition cost around $40 CAD.

Paid online acquisition at the time was somewhat lower, generally around $25 CAD to $30 CAD per first customer.

If we had evaluated the events only as an advertising channel, that comparison would have made them look inefficient.

But the events were producing things digital advertising was not.

Across the same four events, customers and creators produced close to 100 usable pieces of short-form content, hundreds of individual product trials took place, and we collected direct feedback on shades, packaging, price sensitivity, and purchase objections.

The event was simultaneously functioning as:

Customer acquisition + product research + content generation + trust building

Once we accounted for those functions, the economics looked very different.

The salons were more valuable than the venues

Beauty salons were natural partners for the events because they already had something a new ecommerce brand did not: existing customer relationships.

Initially, the partnership seemed like a customer acquisition opportunity.

The salon brought access to an audience. We brought products, content, and an event concept.

But another opportunity became visible after the first events.

If customers responded strongly to a product inside a salon, why should the salon only host the event?

Why could it not also sell the product?

That changed the role of the salon.

A venue could become a customer.

A customer could become a distribution partner.

We began developing a simple monthly partnership model in which the brand would collaborate with approximately one salon per month on an event or activation.

Products would first be validated through direct consumer demand.

Only the strongest products would then be offered to salon partners.

This prevented us from pushing speculative inventory into a wholesale channel merely to create revenue.

The same demand validation system controlled both direct-to-consumer and B2B expansion.

Five active salon partners created a second revenue channel

Over the following months, the network grew to five active salon partners that were purchasing selected products for resale or client recommendation.

The numbers were still small, but strategically meaningful.

By this stage, salon partners were generating approximately $6,000 CAD to $7,000 CAD in combined monthly wholesale revenue.

Direct ecommerce sales had reached roughly $15,000 CAD per month, generated from around 320 monthly orders with an average order value close to $47 CAD.

Together, the business was approaching approximately $21,000 CAD to $22,000 CAD in monthly product revenue without needing to build a large speculative catalog.

Direct-to-consumer gross margins were approximately 45% to 50%, while wholesale margins were lower, closer to 30% to 35%.

That tradeoff was acceptable.

The salons contributed lower margin per unit, but they provided local distribution, product discovery, credibility, customer access, and another source of repeat demand.

This was not simply a second sales channel.

It was the beginning of a distribution network.

The business became increasingly selective about what it stocked

After several cycles of testing and replenishment, the store still carried a relatively narrow assortment.

That was intentional.

Instead of expanding rapidly toward 100 products, the business concentrated most working capital in fewer than 15 actively supported SKUs, with the strongest products receiving deeper inventory positions.

Products could earn more inventory through demand.

They could also lose it.

When new products were considered, the process remained largely unchanged:

Market signal first.

Customer interest second.

Behavioral validation third.

Inventory fourth.

Distribution expansion afterward.

This created a completely different relationship with working capital.

In the original model, inventory would have been the cost of discovering demand.

In the new model, demand was discovered before most of the inventory cost was incurred.

What changed in the first several months

The most important result was not simply revenue growth.

It was the quality of the decisions behind the revenue.

The original launch plan could have placed approximately $20,000 CAD into a wide opening assortment.

Instead, the first concentrated inventory commitment was closer to $6,500 CAD.

Roughly three quarters of that first validated inventory moved within about eight weeks.

The brand developed an audience of around 6,000 social followers and approximately 1,400 owned contacts.

Paid customer acquisition eventually moved toward roughly $25 CAD per new customer as the brand accumulated stronger creative assets, retargeting audiences, and user-generated content.

Monthly ecommerce sales reached approximately $15,000 CAD, while the initial salon network contributed another $6,000 CAD to $7,000 CAD per month.

None of these numbers represented enormous scale.

That was not the achievement.

The achievement was that the business had begun growing without making one large blind inventory bet.

Why the model worked

The strategy changed the order in which risk was accepted.

A traditional retail launch often commits capital first and learns afterward.

We tried to learn first and commit capital afterward.

That does not eliminate risk.

A successful pre-order does not guarantee a successful long-term product. A viral product can lose momentum. A salon partnership can stop producing orders. A social audience can become less responsive.

But each stage gives the business another opportunity to stop before the next, more expensive commitment.

A weak product can fail at the landing-page stage instead of failing with $3,000 CAD of inventory sitting in storage.

A product with strong online engagement but weak payment intent can be rejected before launch.

A good direct-to-consumer seller can prove itself before being introduced to salon partners.

And inventory can be replenished after evidence appears rather than purchased because someone hopes demand will appear.

The larger lesson was about distribution, not cosmetics

At the beginning, the business appeared to have a product-selection problem.

Which cosmetics should we buy?

That was only part of the issue.

The larger challenge was building a system capable of repeatedly answering two questions:

What does this customer want right now?

How can we reach that customer without depending entirely on increasingly expensive advertising?

The demand-validation process helped answer the first question.

The audience, events, creators, and salon partnerships began answering the second.

That distinction matters in crowded markets.

A competitor can often buy the same product.

They can copy a price.

They can imitate a landing page.

What is harder to copy quickly is a working system that continuously identifies demand, tests it cheaply, produces customer relationships, creates content, and pushes successful products through several distribution channels.

The competitive advantage was therefore not necessarily having a completely different product.

It was learning what to stock faster and building distribution around the winners before committing large amounts of capital.

That is a much more defensible capability than simply having a larger catalog.

Questions for other business owners

If you are launching or expanding a product business, several questions from this case are worth considering.

How much of your current inventory exists because customers demonstrated demand, and how much exists because someone internally predicted demand?

What customer behavior could you measure before placing your next large purchase order?

Could you test 20 product ideas and fund only the five or six that generate the strongest commitment?

How much of your customer acquisition depends on channels you have to pay for every time?

Are there businesses that already serve your target customer and could become both acquisition partners and distribution partners?

And if your competitors can sell essentially the same products, what part of your demand discovery or distribution system would actually be difficult for them to reproduce?

For businesses facing a similar inventory decision, the most useful place to start may not be negotiating with suppliers. It may be measuring what customers are already trying to tell you before you place the order.