The lesson for startups from Snap's recent share fall when it reported huge losses was not the lack of profitability -- it was that it missed analyst estimates despite a big jump in marketing spend.
The seemingly unstoppable demise of the media industry has forced many companies to be media companies as they try to reach potential customers with their marketing through traditional and social media channels, and across multiple devices.
In this fractured media world the old marketing programs don't work the same way while the new marketing programs might not work the same way the next time you try them.
When I met Microsoft CMO Grad Conn earlier this year, he told me that everything he knew about marketing he had to re-learn and this is coming from a former Procter & Gamble marketing executive -- P&G marketing is almost a science.
Traditional marketing such as advertising, TV, web seminars, trade shows, etc is less effective because potential customers are spending their time with many other types of media: social, games, Instagram, Twitter, forums, etc.
It's a many-media world and company marketing campaigns are getting used to the fact that each type of media has to be approached in its own way which adds complexity.
The fickle nature of online trends means that marketing needs to be vigilant and responsive on a near real-time basis. Microsoft employs more than 150 social media product managers for such tasks.
""You have to be constantly learning," says Conn. Constantly learning because things are constantly changing. Most CMOs would rather manage a process rather than be engaged in the process through leadership by example, Conn is an exception.
This constant change in online trends means that the results of marketing programs cannot be used to a reliably forecast sales.
With traditional marketing programs for example: P&G knows what each $1m spent in TV ads will generate in product sales. But what will $1m in marketing across a many-media landscape generate? Is it better to spend it on podcast sponsorships or maybe public relations?
It's a fog as Sean Maloney, Intel's former CMO once told me when he spoke about the new types of challenges thrown up by this many-media landscape which because of Intel's global reach he had to figure out country by country. Intel currently has a marketing budget of over $2 billion showing the scale of this problem.
When the marketing gets tough the tough get going: CMO job tenure continues to fall (3.5 years compared with CEOs 7.2 years). It's the toughest job in the C-suite.
For IPO: Predictability trumps profitability...
Predictability in revenues is the biggest challenge to tech IPOs -- not profitability as Snap found to its cost.
All pre-IPO companies face the same scaling challenge: they have to demonstrate how invested capital equals predictable growth.
Young tech companies know that revenues are lumpy and have tried to do away with estimates. When Google went public it said it would not issue any financial guidance -- but that didn't work.
Pre-IPO statements such as Snap's warning about the difficulty in forecasting revenues didn't work either -- investors have their own risk assessment based on many factors. If companies don't offer guidance the consensus estimates could be way out of line and so would its effect on the share price.
Large companies face the same problems. A few recent headlines:
It is not fair to blame marketing unpredictability for making every company's financial estimates unreliable -- but it certainly does nothing to help improve accuracy.
Online ads continue to fail in effectiveness and performance metrics are proliferating and confusing. In such an environment it is not surprising that marketing costs are rising says Gartner.
As digital marketing spend overtakes traditional channels we should see: a rise in the number of missed financial estimates across all companies; stock volatility during earnings season will be more common; and the tech IPO pipeline will continue to struggle with successful launches.
And there is no easy solution (tracking is not selling).