Sentiment, share of voice and mentions aren't crude, they're just the wrong layer

A question came in through the course feedback on the North Star module last week, and it's exactly the right one to be asking at that point:

"How can we move beyond crude data like sentiment, share of voice and mentions to really measure whether comms is helping to enhance reputation?"

It's a meaty topic, and it's the entire focus area of the next module. But here's some practical advice you can use while we put the finishing touches on that course.

First, a reframe. Sentiment, share of voice and mentions aren't necessarily crude. They're just the wrong layer for what you're trying to prove. They measure the conversation and the coverage. Reputation is a belief sitting in someone's head. Those are two different things, and no volume of coverage data will tell you what people actually think of you.

So here are three moves to get from measuring the conversation to measuring belief.

1. Name whose perception you should be measuring

Reputation isn't one score. You have a reputation with customers, prospects, employees, investors, regulators and analysts, and they're rarely the same. Pick the audiences whose behaviour actually moves your North Star, usually one or two rather than all of them, and don't blend them. Keep each audience distinct in your analysis.

2. Define the belief shift

Write it out as: [audience] currently believes X, we need them to believe Y, so that they do Z.

That's the strategy bridge from the module, pointed at belief rather than position. Now you have something to measure movement against, and Z is the business outcome you'll eventually take credit for contributing to.

3. Instrument it three ways

Solicited: quarterly perception tracking. Awareness, unaided description, consideration, likelihood to recommend. Same wording every quarter so you can actually see the line move. (You can use LLMs here as a way of creating a synthetic audience.)

Unsolicited: organic mentions nobody prompted, third party recommendations, and whether you show up when someone asks an LLM about your category.

Behavioural: the proxies sitting in other teams' data. Branded search volume, deals lost to "we've never heard of them", offer acceptance rates, renewal rates, inbound analyst interest.

That’s it!

Do that and your old metrics get a job again rather than getting binned. Coverage and share of voice become your inputs and outputs, perception shift becomes the outcome, and the business result becomes the impact. That chain is what execs respond to.

On budget, because this is where most people stall: you probably don't need a new research line item. The fastest baseline is three questions that already exist in a survey another team runs. Customer success has an NPS survey, HR has an engagement survey, and marketing may already have a brand tracker. Ask for the three questions, own the analysis, and you've got your first reputation datapoint inside a quarter.

An example IRL

The best example I've seen of this in practice is Genevieve Bramall, Head of Publicity and PR at News Corp Australia and co-host of our podcast, Earning Your Seat. She had six months of media data showing which voices were on air and when, which was useful but not the answer. She went looking for what actually drove purchase, found the growth audience she was missing, and rebuilt her inputs around it. Two years later, sales were growing in a declining newspaper market.

One warning. Perception moves slowly. Run this quarterly, not weekly. Report it monthly and you'll spend your time explaining noise.

If you only do one thing from this post, do the second move, and just for one audience. Write the sentence out and post it in this thread:

[audience] currently believes X, we need them to believe Y, so that they do Z.

Reply with yours and I'll give you specific feedback on it. And you'll get a notification here when the next module goes live.