Search, recommendation and social: how consumers discover products

One vacuum flask can reach four buyers by four routes. The first types 500ml flask into a search engine. The second sees a short-video review and taps the product link. The third receives a group-buy link in a friends' chat. The fourth reopens a shop bought from last year. The product is identical, but the intent, the tool and the cost per order are different. Discovery is not one channel; it is a portfolio, and each channel does a different job.

Start with the website, because every other channel eventually points somewhere. Section 3.3.1 gives four functions: establishing brand identity and consumer expectations, informing and educating the consumer, shaping the customer experience, and anchoring the brand in an ocean of messages from social, email and apps while driving revenue. Tesla's site is the textbook example of expectation-setting: it stresses battery technology and range, so the buyer imagines a particular experience. The site is the anchor, not because it is the only page, but because all other pages should land on the same promise.

Search is the channel that captures intent. Section 3.3.2 defines search engine marketing as using search engines to build and sustain brands and support direct sales. It has two legs. Paid search, often called pay-per-click, buys keywords through a bidding process; the ad appears when the query matches, and the advertiser pays when the ad is clicked. Organic search, or SEO, earns ranking by improving content and site structure. The textbook reports click-through of about 1% to 4%, averaging near 2%, and argues that search ads are four to ten times as effective as traditional ads because a query reveals a specific intention to shop and compare.

The cost logic differs between the two legs. Paid search is a variable cost that starts and stops with the budget. It is fast, measurable and competitive: if more sellers bid on the same keyword, the price rises. Organic search is a fixed and slow investment in content, structure and authority; it does not stop when a campaign ends, but it depends on algorithms the seller does not control and it may take months to compound. Neither is free. A page that ranks organically still cost someone to research, write, design and maintain. The useful question is blended acquisition cost per order, not the sticker price of a click.

Recommendation is the channel that creates interest without a query. Section 3.5.3 describes big-data marketing as collecting large amounts of behavioural data to find the target audience and predict the content, time and form of advertising. A practical way to explain the system is a pipeline: behaviour signals, candidate generation, ranking, placement and feedback. Signals include views, clicks, dwell time, searches, add-to-cart, purchases and returns. Candidate generation narrows millions of items to a manageable pool. Ranking scores the pool for this user, often using predicted click or purchase probability plus business rules such as stock, margin and sponsored placement. Placement produces new behaviour, which updates the signals.

The pipeline has limits, and saying so is part of understanding it. A new product has no behaviour, so it suffers from cold start. Popular items get more exposure, which gives them more data and more exposure, a popularity bias. A feedback loop can narrow what a user sees, creating a filter bubble. Sponsored slots are paid placements, not evidence of quality. The system is a ranked guess, not a mind read. For a seller, the practical implication is to make the item easy to interpret: accurate titles and attributes, a clear main image, a competitive early click and conversion rate, and few returns. Those are the signals the pipeline can use.

Social is the channel that builds trust through people. Section 3.3.6 reports that companies spent about US$90 billion worldwide on social network marketing and advertising in 2020, roughly 26.5% of all online marketing spend. It also notes US blog scale: about 30 million writers and 85 million readers, and distinguishes key opinion leaders from key opinion consumers. Content seeding - repeated creator posts that build familiarity before a purchase - is the practical form, and viral marketing is its online word-of-mouth version: customers pass the message to friends and family, which lowers acquisition cost and often brings referred customers to profit sooner. Trust comes from the person, not from the banner.

Social commerce is not the same as social marketing. Section 2.6 defines social commerce as ecommerce enabled by social networks and online relationships. Its tools include social sign-on, network notifications, collaborative shopping, social search and buy buttons inside platforms such as WeChat, TikTok, Facebook, Instagram, Pinterest and YouTube. The textbook estimates worldwide social commerce revenue in 2020 at about US$90 billion, citing Business Wire, and notes that the form is still in its infancy. The same figure as the marketing spend is a coincidence of magnitude, not a shared definition. Marketing spend is a cost; commerce revenue is the value of transactions. They cannot be added, compared as if they were the same, or used to claim that marketing pays for itself.

Influencer marketing sits inside social marketing, and it carries a disclosure duty. The textbook describes influencer marketing as using endorsements and product mentions from people with dedicated followers who are trusted as experts or celebrities. When money or goods change hands, the audience is entitled to know. Many platforms require a paid-partnership label, and consumer-protection rules in many markets treat hidden advertising as misleading. This is a professional and ethical standard, not a claim the textbook makes. Disclosure does not weaken the recommendation; it protects the trust that makes the recommendation valuable. A creator who hides the payment is borrowing trust without repaying it.

Channel fit is the practical conclusion. Search is best at capturing existing demand, so it suits products with clear categories and known keywords. Recommendation is best at creating demand and filling attention, so it suits visually demonstrable products with strong early signals. Social is best at building trust and explaining use, so it suits products that need demonstration or social proof. Direct visits and email are best at retention, because the buyer already knows the seller. A small budget should not be spread evenly across all four; it should buy the channel that matches the product's intent and the seller's ability to measure it. Measurement closes the loop. Last-click attribution gives all credit to the final click, which usually flatters search and direct traffic and undervalues content that created the interest. A better test is incremental: run the channel in one group and not another, or pause it and see what happens to orders. The textbook's figures describe market scale, not a guarantee for a seller. The ad-spend, click-through and US$90 billion figures are context, not a forecast. Lesson 13 turns these paths into a seller-side funnel and asks where the leaks are.