Customer-Demand Assortment Flywheel
Use unmet customer searches to choose the next products to launch
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 96%
Treat customer activity on the existing storefront as a continuously updated market-research system. Record what shoppers search for, identify frequent queries the current assortment cannot satisfy, and rank those gaps by demonstrated demand. Investigate the features customers actually seek, then apply the company's proven sourcing and value proposition to a controlled launch in the leading category. If the new range succeeds, it attracts more shoppers and creates more search behaviour, which reveals the next unmet need. That feedback creates the flywheel: broader relevant assortment produces more traffic and data, and better data improves the next capital-allocation decision. The framework also protects focus by favouring observed behaviour over supplier enthusiasm or heavily promoted trends.
Origin
As Kogan expanded beyond televisions, site searches showed what customers wanted but could not yet buy there, including luggage. The company used those signals to choose new categories and used the lack of searches for 3D televisions to resist industry hype.
Core principles
- 01Observed customer behaviour is stronger than industry hype
- 02Unmet searches reveal demand the current range does not serve
- 03Every successful category can improve the signal for the next one
- 04Allocate capital toward repeated demand evidence
How to run it
- 1
Collect behavioural demand
Track customer searches, purchases, and service feedback, paying particular attention to needs the current range does not meet.
Pro tip Keep raw query language because it often reveals the feature customers value.
Watch out Do not substitute media coverage for customer behaviour.
- 2
Rank the unmet needs
Compare repeated demand signals and prioritize gaps with clear relevance to the customer base and business model.
Pro tip Look for agreement across searches, calls, and purchases when possible.
Watch out High query volume can still represent research rather than buying intent.
- 3
Define the valued features
Determine which product attributes customers are seeking instead of copying the industry's promoted specification list.
Pro tip Translate search terms into a short feature hierarchy.
Watch out A fashionable feature may be bundled into products without causing the purchase.
- 4
Launch with proven capabilities
Use the existing sourcing, pricing, and fulfilment model to test the highest-ranked opportunity with controlled capital.
Pro tip Start near categories where current operational strengths transfer.
Watch out Demand evidence does not guarantee that the business can source or serve the category profitably.
- 5
Feed results back
Use the new category's traffic, searches, and sales to improve the next opportunity ranking.
Pro tip Review both fulfilled and still-unmet searches after every expansion.
Watch out Do not keep expanding when the newest category weakens service or economics.
In the wild
As Kogan's range grew, the company monitored searches for products it did not stock. Repeated searches for suitcases showed that existing customers wanted luggage, so the company applied the same direct-to-consumer methodology it had used for televisions to that category and others.
→ Customer behaviour guided assortment expansion rather than category selection by guesswork.
Manufacturers promoted 3D televisions heavily, but Kogan said customers on his site were not searching for them. They were searching for built-in recording, pausing, and programme-guide features instead, so the company focused on those needs.
→ The company avoided spending substantial attention on a promoted feature that its customer data did not support.
Common mistakes
Following manufactured hype
A feature can appear popular because every supplier includes it, not because customers choose the product for it.
Expanding from a single weak signal
One search or anecdote is not enough. Prioritize repeated behavioural evidence and test with controlled capital.
Is it for you?
Best for
It is best for retailers with meaningful first-party search, purchase, or customer-service data.
Not ideal for
It is not ideal for businesses without enough behavioural data or where searches cannot distinguish curiosity from purchase intent.
From the transcript
“what are people searching for that we don't yet have”
“having that really valuable insights from your customers is the best market research you can possibly do”
From the episode
536: He Made $450M Selling TV's for $0.01
Ruslan Kogan