Section 1
A merge field is not personalisation
Personalisation is worth doing when it changes the argument, not the salutation. A genuinely personalised story references the situation the reader is in, the constraint they are working against, and the objection they are most likely to raise, and it does so because you know something specific about their circumstances. Most systems cannot do that, so they do the visible part instead. The reader notices, because everyone has received a thousand of these. The result is worse than a generic message honestly presented, since the attempt signals effort spent on the wrong thing. The useful question is what you know about this reader that would change what you say, not what you know that you could display. Usually the answer is a situation rather than an attribute, and situations are harder to buy from a data vendor. The wider trend context is covered in [The Future of Storytelling in Business: Trends to Watch](/blog/the-future-of-storytelling-in-business-trends-to-watch).
Section 2
Segment by situation, not by attribute
Industry, headcount and job title are cheap to obtain and weakly predictive. Two companies of the same size in the same sector can be in completely different situations: one is integrating an acquisition, the other is defending a contract renewal. The story that lands is different in each case, and neither is predicted by the firmographic data. Situation-based segments are built from behaviour and stated context. What they searched, what they downloaded, what they told a salesperson, what stage of a cycle they are in. Fewer segments, better fit, and each one deserves a story rather than a variable. Where narrative is the wrong tool entirely is examined in [When Not to Use Storytelling in Business](/blog/when-not-to-use-storytelling-in-business).
Section 3
A variant budget
Every variant is a maintenance liability, so the number you can support is a budget, not an ambition. The table below sets out variant count against the evidence needed to justify it, the update cost per cycle, and the point where accuracy starts to fail.
Section 4
Three versions instead of thirty
Pick the three situations that account for most of your revenue and write a complete story for each. Complete means the problem in their language, the specific obstacle, the evidence from a comparable case, and an objection handled. That is three real assets rather than thirty near-identical ones. Then test whether the distinction is real. Show version A to someone in situation B and ask if it describes them. If they say yes, your segments are not segments and you have a single story with cosmetic differences. Keep a review date on each version. Personalised material rots faster than generic material, because it makes specific claims about a specific context, and contexts move. A variant nobody has read in a year is a liability sitting in a template library. The infrastructure trade-offs behind this are discussed in [The Intersection of AI, Blockchain, and Automation](/blog/the-intersection-of-ai-blockchain-and-automation).
Section 5
Where personalisation turns against you
The first risk is accuracy. Personalised messages fail loudly. A generic email that is wrong is ignored; a personalised email that is wrong about the reader's own business is remembered. The second is the line where knowledge becomes intrusion. Referencing something the reader chose to make public is fine. Referencing behaviour they did not know was tracked reads as surveillance, and the reaction is not proportional to the offence. The third is consent and retention. Every variant depends on data you must be able to justify holding, delete on request, and explain the source of. That obligation is permanent and it grows with each field. The fourth is quiet incoherence. Thirty variants written by different people over two years no longer say the same thing about what the company does, and nobody notices until a buyer receives two of them.