In retail, getting the selection or assortment right is a continuous pursuit in response to the changes in the market and is driven by both a quantitative and qualitative understanding of the category and, therefore, there is no clear science to it. A category should maintain a constant selection pipeline to drive customer value – driven by better, faster and cheaper product. So, selection has to be monitored in three sub-pipelines of a better product at cheaper prices and faster delivery (pricing, latest technology and services, and better spec-to-price ratio). In the online world especially, selection should be specifically monitored under exclusive and non-exclusive selection and the category owners should separately drive both these selections.
How do you know when to add selection?
Adding selection is a continuous pursuit (like adding sales) and one should always monitor the metrics of selection at a category, sub-category and spec level (and exclusive and non-exclusive). However, there are some signs when you will know actively that you have to add selection.
- Signs of inadequate selection (session times, bounce rates, conversions, clues from long tail search queries). Some of the signs of inadequate selection are:
- Increase in bounce rates
- Drop in session times
- Reduced conversions from PLP to PDP and on detail page to checkout
- Drop in absolute conversions (reduced traffic)
- Increase in number of long tail queries on search bar
- Increase in visits of second, third and subsequent pages
- When the latest selection is getting launched in the category! Adding the latest selection (latest tech and services) available in the market. Basis the latest available change in the market, the assortment has to change and a constant pipeline of latest selection has to be maintained in sync with the market. Example of a source: tradeshows and industry conferences
- Growth from existing portfolio is not as high as expected, no major top line addition from existing portfolio. While this is a reactive approach, during annual planning, the category owners do take a target on new selection and sales from the new selection.
- Inputs from continuous internal tracks such as spec-price gap-finding exercises, competitive intelligence (from business relations, vendors, etc.), and changing customer needs (search queries, offline retailer merchandising changes, etc.)
- When gaps are identified while monitoring selection at a sub-category and spec level – each sub-category to have adequate selection in brands, latest spec and BAU spec.
- When gaps are identified in pricing and spec w.r.t. competition and customer needs
How much selection is right selection in a category? And when to add selection?
Building selection or assortment is a tricky problem for any retailer. This is especially tricky for horizontal ecommerce retailer such as Amazon as the platform targets a broad set of consumers with any buying need. So, how do retailers measure their success in selection: What is the right selection? How do we know our selection is improving? Do we just keep adding more SKUs?
The first step is to understand the key relevant attributes that matter to the consumers looking to buy a product. This can be deduced from buying behaviour, filter usages and search keywords. Additionally, the common knowledge from offline retailers and manufacturers should be taken into consideration. This whole knowledge base will help to build a skeleton for selection called Market SKU. Every product category will have a market SKU and a market SKU is the smallest unit of demand.
For example, for a shirt, these attributes could be price, brand, fabric, sleeve length and pattern. A Market SKU could have many individual products, but these products would be similar enough that the consumers would consider them to be “choices” for the same product. For example, Van Heusen cotton full-sleeve plain formal shirts would be a “Market SKU”.
We also needed to deduce, through data, how many such “choices” per market SKU we need to carry. Again, based on data from browsing behavior, conversion behavior and data from offline and online retailers, the minimum threshold for the number of choices for each Market SKU was determined. This is referred to as the “depth” of selection.
One should always go by the fact that the current selection is always inadequate (especially in the e-commerce world) and therefore always be in pursuit to add the selection from customer requirements, competitive intelligence and on-ground feedback from delivery and customer experience teams. Also, the category owners should always accommodate for ‘test selection’, selection that you are still evaluating if it will sell or not (a certain budget should always be allocated for it).
We use machine learning to match our product assortment/selection to the evolving consumer demand. This helps us identify the selection gaps and provides the insights to our business leaders and sellers to fill these gaps.
Next, we developed a statistical model to evaluate the quality of each of the product listings based on, among other things, attributes of the seller and the product, catalog quality, historical performance of the product and the consumer rating of the product. With these, we would be able to evaluate at scale, the width, depth, and quality of selection we carry.
Finally, to tie all of this together using an outside-in view, we survey the consumers and ask them to rate our selection for the specific products they are looking for at Flipkart. This measure is referred to as the Selection Index which is the consumers’ point of view of our selection.
In essence, if we are indeed moving the needle on selection width, depth, quality, and relevance, the true validation is an improvement in the selection index.
This approach has been piloted for a set of categories for which we have been able to move up the consumer Selection Index by over 10% in the last few months. This is just a tip of the iceberg. The goal is to completely transform our selection over the next few quarters.
– Ravi Vijayraghavan, VP, Flipkart.”
Gathering customer requirements, competitive intelligence and on-ground feedback:
- Do a dip-stick or customer immersion programs once in 3 months and understand what are all the various factors and use-cases considered and the weight given to each of those factors and use-cases to understand the whole battery of attributes (factors) important for customers in the category. This can be developed using conjoint or fairly simple analysis to develop a category feature index – a weighted index that signifies how important a feature is to the customer in purchase decisions.
- Understand the existing product issues from analysis of ratings and reviews, customer experience teams and servicing teams. Develop future selection and existing products that fill these gaps.
- Benchmark with customer trends from other benchmark countries (can change for category to category) such as Malaysia, Philippines and China.
- Offline Retailer Visits – conduct offline retailer visits to understand
- Insights from search queries and long tail queries on Google
- Understand the upcoming launches and where the category is moving towards from trade shows, expos, etc. Understand the new launches from companies at product development stages itself.
- Spec and Price Segment-wise benchmarking with market
- New trends on emerging sub-categories
Drive selection without carrying inventory and minimum risk
- Negotiate for sale or return agreements with brands and vendors
- Negotiate for only listing and then order stocks when customer order comes in (ensure that the customer lead time to delivery is slightly on the higher side during this testing phase)
- Negotiate with the vendor to put a warehouse next to your warehouse – this way you don’t need to hold the inventory and at the same time you secured the inventory
- Get liquidation and margin protection support in case of price-drops to sell the product.
Adding selection in startup and initial stages of category
Selection, originally when the company is in initial stages, is always limited by capital. One knows many products that will sell, however, there is only limited capital to invest and make profits. So, an entrepreneur is initially very cautious about selection in the initial stages. The entrepreneur only starts with selection that he is confident that it will sell and make a quick buck. When managing categories, one should still think in the same way. The ideal SKU that an entrepreneur wants to add in selection is the SKU that sells the fastest, the most and gives max profit. The one that sells the fastest and the most are the same. So, essentially we should have a rank within the Market SKUs in terms of growth contribution, profit contribution, traffic contribution, etc. However, it is easier said than done. Usually, the one that sells the most doesn’t give profits and might need higher inventory (lower turns) and the one gives the max profit doesn’t usually have high volumes (and that is why the margins are higher). Slowly as the company becomes big and capital is no more a hard constraint, then selection is limited primarily by discoverability (will users find it) and saleability (will the selection sell atleast one unit) – we are looking at one unit because that is the minimum number of units one will stock and as per the law of cost of capital you will atleast look to sell it in the next 20 days once you stocked it. (20 days is just a cost of capital benchmark in India basis the interest rates).
So theoretically speaking, if one unit of an SKU can get sold in say 20 days of time (18 rotations, cost of capital 1.5% per month), then one should include that SKU in one’s selection. We can stock only one unit of that SKU, however it will add to the sales by one unit. So theoretically, the selection that a category can have is: (sales in units for 20 days)/(avg. sales in units per selection) This implies that if you are confident that you can sell a unit of an SKU in the next 20 days then we should have that in selection. So, the maximum selection one should maintain in a category is the maximum selection that can be maintained with every SKU selling at least one unit in 20 days.
But, does that mean that we will keep adding selection of every SKU that we can atleast sell one unit of in the next 20 days. The answer is ‘No’ and is because traffic is ‘finite’. So, we first decide on how much traffic can come in realistically and the conversion that is optimistic and realistic and basis that we will get a certain unit sales number. Now, use that unit sales number in the above formula with a small twist.
Not every SKU that you have in your selection has the same avg. sale in 20 days (or a fixed period) and therefore weight as per the same. So, the formula for ideal selection is: (sales in units for 20 days basis maximum traffic and conversion)/(weighted average sales in units per selection). Say selection which is a GMS driver like 32” LED TV (bread and butter segment) will have a higher sale than a premium segment. So, for example the sales units for 20 days could come out to be 5000 units and the weighted average sale per SKU comes out to be 20 units, then the ideal selection is 5000/20 = 250 SKUs.
Obviously, the accuracy of this calculation depends on the accuracy of the numerator and the denominator (weighted average and the number of sub-categories and sub-segments considered carefully).
While we say that fast sales and profit is important to decide on selection, sometimes companies and category owners take a call on building exclusive selection and sacrifice the profit in the short-term and gain desired turnover (GMS) and margins in the long-term.
Last, but not the least: In selection, it is very difficult to know if a selection is valuable or not, however, it is slightly more easy to know the answer to the question – ‘is this selection a duplicate of what we already have?’ So, one should always monitor duplicate selection for a kind of false positive check. The question that category managers should answer is: ‘why is this not a duplicate selection?’