A while back, we found ourselves in the pub. A digital planner was talking about her latest campaign, and we asked how many people it reached. "Reach?" she sneered, "it's not about reach any more."

Grrrr… Her Kronenbourg-fuelled outburst betrayed an all too common point-of-view now. In the old days, goes the argument, you needed big audiences. These days, our communications are more finely targeted, more interactive, more engaging. Reach no longer matters.

It's a seductive idea. Agencies like it, because it allows them to experiment. Clients like it, because it appears to save money. Unfortunately, it's just a pipe dream.

Logically, there are only two ways to get a big payback from a small audience. Either you get them to spend a lot. Or they must encourage others to do so. Neither is a sure-fire route to success.

The first strategy is to go for 'brand loyalty'. Target some segment of the market, and extract as much value from them as possible. But data shows that loyalty strategies rarely work. As we discussed before (Admap, April 2010), the main driver of brand growth is penetration.

The second strategy is to go for 'influence'. Spend your money cleverly on targeting a few people, often 'opinion leaders', get them to influence other people for free, who in turn influence others, until you get a chain reaction. Eventually you reach 'the tipping point'. This can happen. Fashions often propagate in this way. So do playground crazes. When SMS took off, its success took the mobile companies by surprise.

But this sort of chain reaction is rare and hard to predict. It can only happen if people are connected in the right way, and if the behaviour in question is highly 'infectious'. In this respect, toys and fashion are unusual - consumption is visible and interesting. But this is not true in most markets. As we've said before, people don't care about brands much.

At first sight, social media data might seem to contradict this. Coca-Cola has over 35 million fans globally on Facebook, for instance. That's a hell of a lot of brand engagement, isn't it? But Coke has around four billion users worldwide, so those Facebook fans represent less than 1% of users. And the figures look similar for other big brands. We estimate that, for the most popular brands on Facebook, only one in 20,000 sales actually leads to someone 'liking' the brand.

And it seldom goes further than that. Facebook's own data reveals that less than 1% of those 'fans' actually talks about the brand. These numbers are far too small to create the big effect that marketers need.

Duncan Watts, principle research scientist at Yahoo!, concludes that it is impossible to reliably generate large effects by targeting a few key influencers. Rather, the optimum strategy is to use mass-marketing, then let viral propagation take the message further. Exponential effects are rare, "But you can double your impact, which is still pretty good," Watts says.

The same is true offline. Famous campaigns get people talking and amplify the message. IPA data suggests that this amplification can double effectiveness, but you still need good reach in the first place.

That's why share of voice (a measure of paid-for reach) tends to correlate more closely with market share than online 'buzz' (the free amplification). In fact, research suggests that the effect of share of voice on sales hasn't changed much in 30 years. Despite the digital revolution, reach still matters.

Smart marketers like Coca-Cola understand this. For them, social media are a clever way to make mass-marketing work harder. But companies that rely too heavily on viral propagation are taking a risk, as Pepsi recently found out to its cost. So the dream of 'target and get the rest for free' unfortunately remains just that - we still need to put our money where the mouths are.

This article originally appeared in the February 2012 issue of Admap. Click here for subscription information.