How to Read a Recommendation You Don't Fully Understand
Someone tells you to increase your budget. Or pause a campaign. Or "focus on retargeting." It's said with total confidence, and it's usually said fast — a line in a report, a comment in a call, a Slack message from your agency.
Most people react to the confidence, not the reasoning underneath it. That's understandable — reasoning takes longer to evaluate than a confident tone takes to absorb. But confidence and correctness are two different things, and a recommendation can carry a lot of the first without much of the second.
You don't need to become a marketing analyst to tell the difference. You need five questions.
Question one: what did they actually find?
Every real recommendation starts with an observation — something that was actually seen in the data. "Conversion rate dropped 12% this week." "Cost per lead is up across all campaigns." "This audience segment stopped clicking."
If you can't identify the specific finding underneath the recommendation, that's worth noticing. Not because the recommendation is necessarily wrong — but because you can't evaluate advice you can't trace back to something observed.
Ask directly: "What specifically did you see that led to this?" A recommendation grounded in a real finding survives that question easily. One that isn't tends to get vaguer, not more specific, when you ask.
A recommendation without a stated finding underneath it is just an opinion wearing the costume of an analysis.
Question two: what does that finding actually mean?
A finding and its meaning are not the same thing, even though they often get presented as if they are.
"Cost per lead is up" is a finding. "Cost per lead is up because our top-performing ad fatigued" is one possible meaning. So is "cost per lead is up because a competitor started bidding on the same keywords." So is "cost per lead is up because of a seasonal dip in search volume." All three are consistent with the same finding — they're not the same explanation, and they don't point to the same next step.
When someone moves straight from finding to recommendation without stating what they think the finding means, they've skipped the step where you'd actually be able to disagree, or ask a better question. That skip is often not deliberate — it's just faster to talk in recommendations. But it's worth asking for the missing piece: "What do you think is causing that?"
Question three: what are they actually telling you to do?
This sounds obvious, but recommendations are often vaguer than they feel in the moment. "Focus on retargeting" is not the same as "shift 20% of the prospecting budget into a retargeting campaign targeting site visitors from the last 30 days." The second version is something you can actually act on, evaluate later, and hold accountable. The first version isn't — it's a direction, not a decision.
Vague recommendations aren't always wrong. But they put the work of turning direction into action back on you, often without you noticing that's what happened. A recommendation worth acting on should be specific enough that you could describe, in one sentence, exactly what will be different afterward.
Question four: how confident should you actually be?
Not every recommendation deserves the same level of conviction, even when it's delivered with the same tone.
A recommendation based on three months of consistent data, across multiple campaigns, with a clear mechanism behind it, deserves more weight than one based on a single week that could easily be noise. But tone rarely signals the difference — confident delivery is available regardless of how strong the underlying evidence actually is.
Ask what the confidence is actually based on: how much data, over what time period, how consistent the pattern has been. A source that can answer this precisely is doing real analysis. A source that answers with more confidence, rather than more specifics, usually isn't.
Question five: what's the evidence, specifically?
This is the question that ties the other four together. Not "trust me" or "this is standard practice" — the actual number, chart, or comparison that supports the claim.
You don't need to be shown a dashboard for every decision. But you should be able to ask "can I see that?" and get an actual answer — a specific number, a before-and-after comparison, a named pattern in a named time period — not a restatement of the recommendation in more confident language.
If the evidence, when you actually look at it, is thinner than the confidence suggested, that's useful information. It doesn't mean the recommendation is wrong. It means you now know how much weight it can bear.
What to do with all five
You don't need to interrogate every piece of marketing advice you receive. Most day-to-day recommendations are small, low-stakes, and fine to act on without a formal review.
These five questions earn their keep on the decisions that matter — a meaningful budget shift, a change in direction, anything you'd have to explain later if it didn't work. For those, running through finding, meaning, action, confidence, and evidence takes a few minutes and gives you something concrete to evaluate, instead of just a tone to trust.
The goal isn't to become suspicious of every recommendation you receive. It's to be able to tell, quickly, which ones are built on something solid — and which ones just sound like they are.
Curious how Lonko Digital thinks about evidence and recommendations? Learn more about our approach →
This article reflects Lonko Digital's general approach to evaluating marketing recommendations and does not reference specific client data or outcomes.