Personalised Training Plan

Active data collection: e.g. surveys

Passive data collection: e.g. transactional data – across all touch-points/all channels (single customer view)

Data quality/usefulness: complete, accurate, timely, allows for managerial decision-making

Data for decision-making (increasing sales):
(Point of sales data: what are they buying, where/how are they buying, what triggered the purchase, what do they have in their basket, etc.)

1. Impact of promotions (both short-term and long-term): who buys our products on promotions; how many of our customers are cherry-pickers; will cherry-pickers become loyal (& if so, how); are today’s promotions cannibalising future promotions/sales

2. Impact of display (retail, catalogue, online): which types of display work best to optimise sales

3. Sales within and across categories: which products (or categories) are substitutes; which products (or categories) are complements; which products are working well together and which are cannibalising each other

4. Impact of pricing: standard list price, discounted price, membership discounts/launch discounts, currency pricing overseas

5. Effectiveness of new customer recruitment: how did first time buyers find us; how much did they purchase; which products or offers will optimise their purchases

Data for decision-making (improving profitability):

[TBC]

Data for decision-making (reducing risk):

[TBC]

Terms Related to Matching and Ranking

Ranking (a function which selects and returns results ordered based on some kind of relevancy algorithm, e.g. synonym matching to product database)

Two-step ranking (as above but with two stages involved in the selection and ranking process, e.g. synonym matching followed by sales best-seller ranking)

Text string (e.g. the search query typed into the search box in an online library catalogue or a book-seller’s website)

Matching (e.g. matching the search text string to a pre-defined list of categories in the database of books held/sold)

Controlled vocabulary index (where there are only a certain number of words which can actually match like for like, e.g. a subject index for all the books in a library, or all the books for sale on a website)

Synonym matching (where matching will be done not only on the exact words in the text string but also relevant synonyms for the exact words searched, e.g. where searches in a bookseller’s website returns relevant matches ranging across a number of possible/similar categories)

Parent category (e.g. high-level subject index)

Child category (e.g. sub-topics within the high-level subject index)

Tree of categories (the ordered list of parent and child categories)

Relevancy ranking (ranked according to some pre-defined relevancy algorithm, likely based upon both synonym matching and best-seller ranking)

Terms Related to Selling

Best-seller ranking (best-seller = as defined for all those products falling within a specific sub-category only; ranked according to highest levels of sales within specific sub-category)

Co-occurring sales (sales orders for multiple products placed by one customer in one order)

Co-occurrence sales data (all data related to multiple sales orders being placed by any one customer at any one time)

Co-occurrence matrix (all the co-occurrence data mapped out structurally to allow better analysis and sales-optimising decision-support)

A/B (split) testing (testing two different but similar options to different customer cohorts simultaneously to gather empirical evidence as to which option drives more sales)

Terms Related to Browsing (but not Selling)

Co-occurrence clickstream data (all data related to multiple views/clicks/browses by any one customer during their current visit that results in the sale of at least one book to that customer, i.e. what other products have you viewed together with the product you went on to purchase, even though you didn’t actually buy the viewed product) (e.g. Amazon “What other items do customers buy after viewing this item” recommendations)

Also mapped to a Co-occurrence clickstream matrix

Mapping the Above to a Product Database

Each item in a product database should have the following defined fields:

The full tree of categories the product fits into, including:

–   All relevant high-level subject categories

–   All relevant sub-topics within those high-level categories

Sales best-seller ranking

Cross-referencing categories (e.g. genre, author, year of publication, etc.)

Up-selling / cross-selling recommendations (e.g. Amazon’s “frequently bought together” and “customers who bought this items also bought” recommendations)

Advanced databases may include possible cross-sell / up-sell discount %’s to offer when making the recommendations to encourage sales (Note: % discount levels to be offered on cross-/up-sell will be dependent on empirical review of selling metrics for those products i.e. whether or not offering a % discount increases sales volumes or not. It may be subject to frequent change dependent on season, customer type, etc.. As such, the business may not wish to hard-code recommended % discounts into the product database, or if they do so, to have functionality to allow regular review and rapid updating in order to optimise sales.)

Such up-/cross-selling metrics should be tracked at all times (e.g. what products do customers buy together – from co-occurrence sales data; what has driven that buying behaviour; what impact do different marketing offers and % discounts have on sales volumes and values; what marketing offers and % discounts should be provided to optimise sales values; what other products do customers view when they buy this product – from co-occurrence clickstream data;). The most up-to-date metrics should be maintained – likely an independent database table which regularly feeds into the product database – to allow automation of the sales optimisation process.

Hence product database for each product will be dynamically updated for latest data on:

1. Best recommendation for up-sell/cross-sell, with or without sales optimising % discount offer

2. What other products are purchased when this product is purchased

3. What other products are purchased when this product is viewed

Summarised as:

Frequently bought together metric, possibly with addition of sales-optimising % discount

Recommendation engines (one based on multiple purchases made by others based on this product purchase, another based on single purchases made by others based on this product viewed)