The four things worth knowing before you write
A research pass that takes ten minutes and actually changes the message answers four questions:
- What has changed at this company recently, according to the company itself?
- What problem does that change create, in the language of the person who owns it?
- Who owns it, and how do we know - is that person named in a document, or guessed from a title?
- Why would they act this quarter rather than next year?
Named in a document, or guessed
The difference matters more than it appears. A person named in a filing, a press release, a conference listing or their own company leadership page is a fact about the company. A person inferred from a job title in a contact database is a guess that may be two role changes out of date.
A research system that cannot tell you which of the two it is doing will eventually have a seller address someone by a title they left a year ago, which ends the conversation before it starts.
Where personalisation goes wrong
The failure mode is mechanical personalisation: a template with a variable slot, filled with something scraped. "Loved your post about Q3" reads as automation because it is. The recipient has seen hundreds.
What works is narrower and harder to fake: a specific observation about their situation that could not be true of another company, followed by a claim about what usually goes wrong next. If the first two sentences would survive being read aloud to the recipient, the research was real.
Judging a research tool
Ask it for an account you know well and read what it produces. The questions that separate the useful from the plausible: does every claim link to a document, does it distinguish what the company said from what someone said about them, and does it tell you when it found nothing rather than filling the space?