LLenny's Podcast
← All frameworks
MarketingTimothy Davis (Shopify)

Signal vs Noise Metric Selection

Let the campaign's single goal decide which metric is signal — everything else is noise

Difficulty
Moderate
Time to result
~days to results
Steps
3
Confidence
90%

Ad platforms hand you an overwhelming firehose of metrics (impressions, reach, frequency, CPM, CPC, CTR, cost-per-lead, cost-per-MQL). The discipline is to ignore all of them until you name the campaign's objective, then focus on the one metric that objective is defined by. This is both a hiring filter (hire for data judgment, not tool knowledge) and an operating habit.

Origin

Davis adapts the concept from Nate Silver's book 'The Signal and the Noise', deliberately reframing it as 'signal, not noise'.

Core principles

  • 01Anyone can be taught Google or Meta ad mechanics — the scarce skill is judgment about which data matters
  • 02The campaign's purpose dictates the primary metric; other metrics are distractions
  • 03A funnel-stage lens keeps you honest: awareness = people who saw it, consideration = people entering the funnel, conversion = actual conversions

How to run it

  1. 1

    State the campaign's single purpose

    Before touching any dashboard, force the answer to 'what is this campaign for?' — awareness, consideration, or conversion.

    Pro tip Use this as an interview question: throw a pile of data points at a candidate and see if they first ask what the campaign's goal is instead of drowning in metrics.

  2. 2

    Map the goal to its defining metric

    Awareness campaigns are judged on how many people saw the ad; consideration on how many entered the funnel (white-paper downloads, demos); conversion on conversions themselves.

    Watch out If the goal is conversions, impressions/reach/frequency are noise — chasing them optimizes the wrong outcome.

  3. 3

    Feed the platform only the business-result signal

    Because platforms optimize toward whatever action you emphasize, tell it to chase the action that equals a real business result — otherwise it gets very good at delivering worthless volume.

    Pro tip When hiring, screen for this data judgment first; the platform mechanics can be trained later.

    Watch out An action that doesn't equal a business result trains the algorithm on noise and hurts you at scale.

In the wild

The interview data-dump test

Davis throws a bunch of data points at candidates and asks what they'd optimize. Most get overwhelmed; the ones he likes first ask 'what's the purpose of the campaign?' and, on hearing 'drive conversions', focus on clicks and conversions rather than impressions or reach.

The question reliably separates data-judgment hires from metric-reciters.

Common mistakes

Optimizing to vanity reach when the goal is conversion

Reporting impressions, reach and frequency for a conversion campaign optimizes the wrong metric and sends the platform a misleading signal.

Is it for you?

Best for

Performance marketers and hiring managers building analytical discipline in a metric-saturated ad account

Not ideal for

Situations that genuinely require broad diagnostic dashboards rather than a single-objective read

From the transcript

what is the signal and what is the noise and let's make sure we're focusing on the right signal versus the right noise

42:30

if the goal of the campaign is to get conversions that's what we should be doing if it was awareness it's how many people saw…

47:00

there was a a book written by uh Nate silver called the signal and the noise

40:30

From the episode

The ultimate guide to performance marketing

Timothy Davis (Shopify)