Build Personalization Around the Next Best Conversation

1. Campaign Personalization Is a Category Error: Dave distinguishes personalization from using a transaction to populate a mailing list. After buying a Nespresso machine with 100 pods, he received daily pod discounts beginning three days later and another machine offer slightly more than a week afterward. Nespresso already knew his residential address, selected flavors, and delivery date, yet ignored that context. A next best conversation could have asked whether he enjoyed the machine, explained iced and hot coffee options, introduced coffees from different regions, or offered auto-replenishment at an appropriate time. The central strategic shift is from selecting audiences for prebuilt pitches to continuously recognizing customer states and designing interactions that move each person forward.

2. Value-Creating Triggers Improve the Economics: The Verizon example shows how a trigger can serve both the customer and the company. Ryan appears likely to leave because he has kept his phone for a long time and is on an unsuitable plan. Verizon and Meta can use those signals to place a relevant ad, offer a new television with a phone upgrade, collect zero-party data about trade-in preferences, and recognize repeated data-limit overages. The recommendation may raise Ryan’s monthly plan price while reducing his total costs. The journey then connects marketing with billing, fulfillment, and delivery tracking. This is more than message targeting: it is end-to-end experience design built around a specific customer problem and measurable commercial risk.

3. Personalized Education Can Reduce Service Demand: As Chief Marketing Officer at Aetna, Dave confronted widespread confusion about health insurance, including deductibles, coinsurance, copayments, provider networks, and appropriate care settings. Aetna used SundaySky to assemble personalized, largely animated 3.5-minute explanations. A video could identify covered family members, note that a spouse lacked a primary care physician, list five nearby in-network doctors accepting new patients, and show three local urgent care centers as alternatives to an emergency room. Seventy percent of recipients watched the content, call-center volume declined by more than 20%, Net Promoter Score increased, and customers opened later emails more often. The case demonstrates that personalization can educate customers while lowering operating costs.

4. AI Makes Disconnected Data More Usable: Customer context often remains divided among marketing, billing, service, sales, product usage, and regional systems. Orange and other telecommunications companies use Dataorbit to transform call-center and chat conversations into structured data. A complaint about preferential offers for new customers, a competitor mention, or an unresolved service issue can become an attrition trigger rather than disappearing into a recording. Narrative’s Rosetta addresses a different problem by reading database schemas, writing integration code, and standardizing data in a common destination. Pepsi applies it across 40 geographies; The Trade Desk integrates client and platform information; Snowflake uses it as an ingestion front end. AI reduces integration labor, although it cannot correct information never captured.

5. Zero-Party Data Creates Permissioned Relevance: Organizations that lack preference data can ask customers directly. Marriott Bonvoy, under Chief Customer Officer Peggy Roe, asks frequent business travelers where they want to vacation with family and what welcome gifts they prefer, then records answers in visible profiles. Netflix’s thumbs-up and thumbs-down controls are a simpler version. At Aetna, five trade-off questions helped people over 65 navigate complex Medicare choices, including travel coverage and preferred benefits. SonderMind goes further: gamified exercises assess 12 dimensions of how a person’s brain works, support therapist matching and treatment planning, and are repeated three months later to measure progress. Each example gives the customer an immediate reason to share information.

6. Experimentation Personalizes the Decision Rule: A/B testing identifies an average winner, but personalization requires learning which option works for which customer. Shopify’s conversion journey from free trial to paid service included about 10 variable elements and roughly 500 possible message combinations. OfferFit used trial behavior, line of business, and journey information to optimize choices, producing multiple-times daily returns over three months. Noom runs more than 100 tests each month, while Booking.com explicitly treats tests as experiments. Dave recommends directing 10% to 20% of the marketing budget toward new tests. Generative AI can create the required variations, but teams still need hypotheses, controlled deployment, human review, and segment-level measurement.

7. Content Volume Needs Structured Governance: One international airline used generative AI to increase personalized content output by 40 times. Its prompts and controls incorporated brand voice, message tone, and blacklisted words, illustrating why scale cannot mean unrestricted generation. The airline also tagged creative across 44 dimensions and, over 18 months, analyzed 1,000 campaigns and billions of impressions. It learned details such as effective colors, whether to feature a model, where to position the call to action, and when “new” outperformed “free.” Platforms including Adobe, Salesforce, and Braze can support this metadata. The management challenge is to make tagging mandatory so every experiment expands the organization’s reusable knowledge rather than producing an isolated result.

8. Operating Design Sets the Speed Limit: Personalization cannot move rapidly when 13 specialists and multiple agencies must hand work from one function to another. Dave recommends cross-functional pods of five to six people who combine complementary skills, maintain a prioritized backlog, work in two-week sprints, and hold daily standups. Even a highly regulated pharmaceutical company accelerated its process by incorporating compliance into this operating model. At Aetna, five pods focused on specific outcomes, including Medicare sales and small-business lead generation. AI can automate dashboards, generate creative options, and summarize results, but it does not remove organizational friction. Tight teams need shared goals, direct access to performance data, and authority to release controlled experiments.

9. Trust Depends on Purpose and Restraint: BCG consumer research cited by Dave found that 70% of U.S. customers had given more business to companies that used their information to help them. Conversely, two-thirds had stopped doing business with companies they believed used information inappropriately. These findings frame privacy as more than a consent requirement: the perceived purpose and usefulness of each interaction matter. Companies should secure permission, protect data, test potentially sensitive uses with employees and small customer samples, and avoid personalization that merely reveals how much the company knows. The threshold between useful and intrusive varies by person, so disciplined testing and clear customer benefit are safer than relying on a single universal rule.

10. Orchestration Prevents Cross-Product Bombardment: When multiple product teams target the same account independently, even individually relevant offers can create a poor customer experience. Dave recommends taking a customer-level view and ranking perhaps the top three product priorities at a given moment. An orchestration engine can consider current product usage, content consumption, service history, industry vertical, and other signals to recommend whom the sales force should contact and about what. AI can make an initial ranking, but leaders must define starting business rules and resolve trade-offs among products. This approach replaces internal competition for customer attention with a coordinated next best conversation governed by the overall brand relationship.