The first time I saw this, I thought it was absolutely insane.

It’s the dashboard Nicolas Boucher built to monitor the health of his community.

No other creator I know goes to this level of effort to make sure their community actually gets a result.

There are 3,551 members in there. And the data is an effort to ensure people get an immense amount of value from the platform. To the point where they actually achieve the transformation Nicolas promises.

So the dashboard shows him who's commenting, who's engaging, and who hasn't logged in for a while. Which means somebody can get a nudge to help them get value.

It's probably the most mature mechanism behind community performance I've ever seen.

And it's grounded in good data, with logic that has an associated action attached to it. If the data says X, the action is Y.

Last week I wrote about getting too close to the data and stressing myself out. Email opens, email clicks, refreshing every five minutes to see what changed.

This week, working out what data is bulls* to be ignored, and what data is genuinely useful.**

I've got four pillars for this. Here they are, plus what I found when I went digging into how some very big names think about their own numbers.

Specific = data you know is 100% accurate

Directional = data that it’s impossible to ensure accuracy, but is still useful.

It's very difficult to be specifically accurate with data like an email open rate, because the odds are stacked against you. There are too many variables. A lot of opens happen automatically by people's mail clients - Apple Mail being the big one, and that's a lot of people on iPhones.

So if I look at an open rate and try to work out whether a 2% swing either way is good or bad, I'm wasting my time. It is pointless as a gauge of subject line success.

Directional is different.

If I'm running between 40 and 45% edition on edition, that's my average, and I know I'm performing consistently. But if there's a big swing either way, that tells me something.

Drop to 30-35% and either the subject line is truly terrible, or there's been a deliverability issue - spam triggers, promotional tab, that sort of thing. Those big swings let me ask "okay, what went wrong here?"

Go over 50% and that's an outlier I might want to emulate. Although outliers like that are usually off the back of a hot topic, a seasonal trend, or a specific tool everyone's going crazy about — so sometimes you can't emulate a point in time. Which again is why looking at one variable in isolation is pointless.

Nathan Barry (Kit/ConvertKit) wrote the definitive piece on Apple's Mail Privacy Protection, and his point is better than the usual "opens are dead" take.

And their own deliverability lead, Alyssa Dulin, said the specific vs directional thing - probs better than I just did:

The majority of data is useless in isolation, and useful when you can identify a trend.

So you've got to decide what granularity and what frequency you need to make a decent decision.

Take a social post. You can get properly bent out of shape checking the feed every five minutes to see how many impressions it's picked up. On something like TikTok or X, where posts don't last any time at all, maybe impressions-per-minute is genuinely useful. But mostly it's pointless, because it's already been posted. Your only options are delete it or edit it, and both of those can cause you algorithmic problems anyway.

LinkedIn and YouTube are different. The content lasts longer. LinkedIn now acts as a mini website for your brand — the stuff you post gets indexed by search engines and AI engines. So it isn't just about the performance of the initial post, it's about credibility, alignment of the content to your brand, and your experience.

Which is why there's no point me tracking post by post over a week. That post could grow.

I had one that started off semi-viral. About 50k views in the first week. Then somebody else picked it up, started engaging with it, LinkedIn decided to recirculate it — and it went on to do 1.2 million impressions.

Point-in-time data for a single post is useless. You need to look at data strategically, and by that I mean: define the check-in point. Is this a daily data point? Weekly? Monthly?

Some things you can only look at annually. Sales in August are pointless to panic about unless you've got a prior year benchmark, because you're tracking against last year's seasonal dip. There's always a dip.

The most data-obsessed man on the internet does exactly this. MrBeast's leaked internal staff handbook tells his team there are only three metrics - CTR, average view duration and average view percentage - and then adds the rule:

He refuses to compare a day-one number to a lifetime number.

He also does the controlled comparison thing properly. Two videos, identical length. One did 120M views, one did 45M. The difference? People watched the winner 1 minute 38 seconds longer. His conclusion to staff: "OF COURSE IT GOT TRIPLE THE VIEWS."

Most data points are bulls***.

Staying on LinkedIn: impressions is a vanity metric. I've had more people approach me off the back of posts that did a few hundred views than posts that did tens of thousands.

Why? The posts that get more views are broader, and they potentially speak less to my audience. The posts that get fewer impressions are more specific, and they showcase my expertise to exactly the people I want.

Followers is a vanity metric too. It looks good on a profile, but if you actually look at your followers, the proportion inside your target market is tiny. And a follow isn't a signal of buying intent. Someone can follow you because they liked one post about your life and still think this dude is not somebody I'm going to buy from.

Profile views is a better metric, because you can work out what proportion of your posts, DMs and connection requests are actually getting people through to your profile. That's where you can make a tangible change - is my tagline rubbish? Are people reading it and thinking "nope"? Change it, watch the data. Over months, not days.

Then the brilliant data that sits underneath: how many messages, how many connections, how many newsletter signups came off the back of LinkedIn.

It doesn't even need a UTM. If I post with a link once that week, and subscribers from LinkedIn go up that week, the timeline adds up - and that's a format to repeat. That's how it becomes a virtuous cycle.

Same with the newsletter. Opens are vanity. Clicks are getting unreliable because of bots. The only thing that's genuinely useful is registrations and signups.

So when I look at a newsletter, I'm not looking at open and click performance (even though it's directionally useful for spotting outliers). I'm looking at how many people we drew to the webinar off that specific edition. Same with brand partners: the number of registrations or gated form submissions we delivered is the ultimate metric.

You've got to decide what your ultimate KPIs are and ignore the rest, unless it's useful to analyse once a month for outliers and direction.

The best example of someone doing this is Alex Hormozi. He ran a test making broader content (college, relationships, fitness, philosophy) instead of pure business content. Views went up. Subscribers went up. Then a friend running a $1M/month business told him he'd stopped listening: "I'm just not your avatar anymore."

He went back to niche business content. Views roughly halved. Opt-ins and book sales went up (roughly 25% more opt-ins, book sales doubling).

He watched the vanity metric fall by half and called it a win, because the metric underneath became more useful. Reference: results breakdown

The cautionary version is Halsey, who in 2022 said their label wouldn't release a finished song "unless they can fake a viral moment on TikTok" - after eight years and 165 million records sold:

A threshold, with no action attached to it, used as a gate. That's the purest form of BS data I've seen. (The song got released, obviously. Because the complaint about needing a viral moment became the viral moment XD)

This one wasn't in my original three. Nicolas' dashboard is what made me add it.

Because what do you actually do off the back of the data?

There are two things.

So it's either a performance metric to be emulated, a performance metric to be avoided, or a data point that triggers a specific action.

The person who proved this at the top level is Ian Graham, the Cambridge physics PhD who built Liverpool's analytics department. His summary of how most clubs fail is brutal: they "could be doing good work but find that it is not impacting the decision-making." Reference: Sky Sports

Good data with without an action is a missed opportunity.

And my favourite version of the action being deletion: Steven Bartlett bins around 20 fully recorded Diary of a CEO episodes a year. Because the data said publishing them would damage what he calls "invisible trust." Reference: Bartlett

The difference between Bullsh** data and brilliant data isn't volume of metrics.

It’s virtuous metrics the lead to constant improvement over time.

Everything else is just refreshing a shiny, but ultimately sh** dashboard.

Until next time.

Adam

—

P.S - Want me to help you find you best AI use cases? Reply with what you’re struggling with, and I’ll help where I can.

P.P.S - 3 extra goodies for you:

1. Claude for e-mail subject lines:

How I get 52.98%+ opens on a newsletter to 190k subs here

2. Turn followers into brand partnerships (without relying on agencies)

5 day e-mail course on how I've delivered $500k+ in 6-months for influencers here

3. Resource vault

Everything else I've produced over the past 3 years here.