What happened
Rajalakshmi Srinivasan has been at Zoho for twenty-three years, which is her whole career, and she started as a developer. She opened with a list of words on a slide: human hair, peacock tail, bamboo, ostrich feather, amber, tungsten. They are filament materials, and only the last one worked. Her point was that Zoho is known for its fifty-odd products and that anybody who has been there long enough can list the failures instead. She has worked six months or a year on things that were never released, because the market was not there or had moved while they built.
Her six ways of balancing what customers ask against what you planned were the substance of the session. Weigh one customer against many, and prefer the small feature hundreds want over the large one a single account wants. Make that demand visible somewhere other than support tickets, which get missed: their answer is a public feature request forum per product where customers upvote each other. Define the roadmap for a year and review it quarterly rather than planning five years out. Do not pull people off a half-built feature when something urgent arrives, because people need to see the thing they built get released.
Two of the six were about giving the work away. Extensibility, meaning proper APIs, so a customer who needs a small change can make it, integrate with their own ticketing tool, or build their own interface. And customization, because the same product gets used by a bank, a college and an IT team who each want the fields named differently, and answering that with professional services does not scale. Let them rename things and rebrand it entirely if they want. Both exist to reduce the number of requests that reach you at all.
The fifth was about delays, which she said they have still not mastered after all these years, and where the last mile into production routinely takes longer than the development did. Her rule was that the lead should exceed the lag: if you know in November that a December commitment will slip by a month, say so in November. She has been badly caught by not doing it, and pointed out that customers are people who will accept a reason and will not accept being asked in January what happened to December.
The sixth was permission to say no. She was pointed about "it's on our roadmap", the standard reply, and asked the room whether they would accept it themselves from somebody who knew the thing was not happening for two years. If a request is too complex, or you have not got the people, or it is a niche you do not want, say that. It is fine not to satisfy every customer, and the relationship survives the honest version better than the vague one.
She finished on using your own product, which at Zoho means running it internally in production for around two years before it goes to market. Site24x7 was built to monitor Zoho's own infrastructure and was only sold after a few years of that. The observation underneath it was the sharpest thing she said: they are more afraid of the internal release than the public one, because colleagues complain the moment something breaks, while customers are polite enough to wait. The impatience is the point.