Data beats opinions. — LinkedIn Wallpaper

The Tuesday pipeline meeting that this banner is really about

Picture a standard weekly revenue review. Twelve people on the call. The VP of Sales says the new outbound sequence "clearly isn't working — reps hate it, I've heard it from three of them." Nobody has the sequence data open. The room nods, the sequence gets killed, and the team reverts to the old cadence. Three weeks later someone pulls the numbers and finds the sequence had a 4.1% reply rate against the old cadence's 2.6% — it was working, it was just harder to run.

That meeting is the exact failure the Wallpaper is arguing against, and it is worth being precise about *why* it failed. It did not fail because the VP was stupid or acting in bad faith. It failed because three reps complaining is genuinely available information, it arrived first, and it came from someone with authority. Daniel Kahneman's name for this in *Thinking, Fast and Slow* is WYSIATI — "what you see is all there is": the mind assembles a confident narrative out of whatever information happens to be in front of it and almost never pauses to ask what's missing. In a meeting, the information in front of you is whoever spoke first or ranks highest. Analytics writer Avinash Kaushik gave the resulting pattern its enduring nickname — the HiPPO, the *highest-paid person's opinion*, which wins arguments by org-chart position rather than by evidence.

The reason a slogan like "Data beats opinions" travels — on a LinkedIn cover, on a wall, on a slide — is that it gives a room a *non-confrontational script* for interrupting that pattern. Saying "I disagree with you" to a VP is a political act. Saying "what would we need to see in the reply-rate data to settle this?" is a procedural one. The banner does not make anyone smarter. It pre-negotiates a norm, so that when the moment comes, the person asking for evidence is invoking a shared rule rather than picking a fight.

Note carefully what the phrase does *not* claim. It does not say opinions are worthless. In the pipeline example, the VP's instinct was the most valuable input in the room — it identified exactly where to look. The failure was treating a seasoned instinct as a *finished conclusion* rather than a starting hypothesis. That distinction is the entire operating principle, and everything below is a way of making it routine.

How the rule actually works as a decision procedure

The phrase only becomes useful when you convert it from a sentiment into a sequence you can run in a meeting. The sequence has four moves, and the order matters more than the sophistication of any single step.

Move one: state the claim as something falsifiable. "The sequence isn't working" is not checkable. "Reply rate on Sequence B is below 3%" is. The conversion step alone resolves a surprising share of disagreements, because two people often discover they were arguing about different claims — one meant reply rate, the other meant rep morale, and both were right about their own metric.

Move two: name the threshold before you look. Write down, in advance, the number that would change your mind. "If B's reply rate is under 2.6%, we kill it; at or above, we keep it and fix the rep experience." Setting the bar *before* seeing the data is what separates a test from a rationalization. Skip this and you will find yourself explaining, after the fact, why 2.4% is "actually fine given the segment."

Move three: check whether the question is measurable at all. Some claims can't be settled with data on any reasonable timeline — "will this brand move damage trust in five years?" is a real question with no cheap test. Routing those into the experiment pipeline wastes weeks. Route them explicitly into judgment, and say out loud that that's what you're doing.

Move four: act on the result, then measure the outcome of acting. The loop closes only when the decision itself gets a follow-up measurement. Otherwise you build a culture that runs tests and still can't tell whether its decisions are getting better.

The loop back from "measure the outcome" to "restate the claim" is the part most teams drop. A decision that was correct in March may be wrong in September because the underlying market moved. Treating every settled question as permanently settled is how an evidence-based culture calcifies into a different kind of dogma.

Real numbers: the banner spec, the test sizes, and the thresholds

Two sets of numbers matter on this page — the ones that govern the Wallpaper itself, and the ones that govern whether a test can actually beat an opinion.

The banner spec. LinkedIn's profile cover slot is 1584 × 396 pixels, a 4:1 ratio. That is the number to build to and the number to check before uploading; anything else gets cropped or letterboxed, usually straight through your text. This Wallpaper ships as an SVG, a scalable vector, which means it stays sharp at any display size and on any pixel density — no re-export for retina screens. It is free to use with no attribution required. Practical consequences of the 4:1 ratio: keep the readable text inside roughly the middle 60% horizontally, because LinkedIn's profile photo overlaps the lower-left region and mobile crops the edges more aggressively than desktop. A square post or a story frame will not accept this aspect ratio without a re-crop.

Test sizes that actually settle an argument. The honest answer is that the sample you need depends on the size of the effect you're trying to detect and the baseline rate you're starting from — and the relationship is punishing. Detecting a large effect (a doubling of conversion) takes far fewer observations than detecting a small one (a 5% relative lift), and detecting any effect on a rare event (2% conversion) takes far more traffic than the same effect on a common one (40% conversion). Two practical implications:

• If you are trying to detect a *small* lift on a *low-rate* metric, the required sample often exceeds anything you can collect in a quarter. Recognize that early and stop pretending the test will resolve it. Either go find a bigger lever, or accept that this decision belongs to judgment.
• If your opinion-holder is claiming a *large* effect ("this will double our reply rate"), that claim is cheap to test. Large claims are the easiest to falsify, which is why insisting on the threshold up front is so effective — it converts vague confidence into a specific, checkable number.

The thresholds worth writing into policy. A "no data, no decision" rule needs a trigger level or it becomes bureaucracy. Pick a dollar figure and a customer-count figure that are meaningful at your scale — for a small team that might be anything above a few hundred dollars or affecting more than a hundred customers; for an enterprise it may be two orders of magnitude higher. The specific numbers matter less than the fact that they are written down and applied consistently. Below the line, decide fast on judgment and move on. Above the line, either produce evidence or explicitly label the choice a bet.

Time thresholds. Give every test a pre-committed end date. A one-week pilot with a single segment, a two-week A/B on a landing page, a single sequence run against a defined list — the point is that the stop condition exists before the test starts. Tests without end dates get read early, get read repeatedly, and get stopped the moment they show the answer someone wanted. That is not evidence; that is an opinion with a dashboard.

Trade-offs: when the opinion should actually win

An honest version of this principle has to name its own exceptions, and there are three that hold up.

The data is too sparse or too noisy to carry the weight. Twelve visitors from a niche segment is not a signal that should override someone's five years of direct experience with that audience. The failure mode here is *false precision* — a number computed to two decimal places from a sample of twelve looks more authoritative than an expert's qualitative read, and it is not. When the sample is thin, say so explicitly and treat the number as one input among several rather than as the verdict.

The metric being optimized is the wrong metric. This is the more dangerous case, because the data is perfectly sound and still leads you astray. If your dashboard says page-load time doesn't correlate with conversion, but you know from watching session recordings that mobile users are abandoning mid-form, the quantitative signal may simply be measuring the wrong thing — or measuring it on a population that excludes the people who already left. A metric that is easy to collect is not the same as a metric that matters to revenue. Ask what the number *would* look like if the thing you fear were true; if the answer is "exactly the same," the metric can't settle the question.

The decision is about ethics or brand, not optimization. Aggressive dark patterns reliably lift signup numbers in the short run. Data will tell you that. Data will not tell you what it costs you in trust, churn, or regulatory exposure over three years, because those effects arrive on a timescale your test doesn't cover. Here the correct move is to state plainly that the metric is not the objective and let judgment govern.

The rule you are actually adopting when you post this Wallpaper is therefore narrower and more defensible than the slogan: data wins when the question is measurable and the measurement is sound. That version survives contact with a skeptical colleague. The unqualified version doesn't, and someone will use its weakness to dismiss the whole idea.

Common pitfalls: opinions wearing a data costume

The most expensive mistakes are not opinions that lose to data. They are opinions that got dressed up well enough to pass as data. Three disguises account for nearly all of it.

The expert-experience disguise. "Based on my twenty years, this is the right move." Experience is genuinely valuable, but it is an opinion *shaped by* experience, not evidence — and human memory is selective in a specific direction: we remember the calls we got right far better than the ones we got wrong. The counter is a single concrete question: "which comparable situation, with a measurable outcome, supports that?" If no specific example can be named, what's on the table is a hypothesis. Ask it neutrally and it works; ask it sarcastically and you've just started the political fight the Wallpaper exists to prevent.

The anecdote-as-signal disguise. "I talked to three customers and they all want feature X." Three voices out of a customer base of thousands is a sample of three. The two questions that defuse it: how many did you reach, and how were they selected? Self-selected power users who answered your email are systematically different from your median account — they're more engaged, more tolerant of complexity, and more likely to want features nobody else will use. Treat the three conversations as a *lead to investigate*, which is real value, not as a roadmap item.

The cherry-picked-distribution disguise. Someone presents the deals that closed in eleven days and omits the ones that stalled at ninety. The fix is to always ask for the whole distribution, not the highlight reel: show me every deal in the period, the median close time, and the range. Medians and ranges are much harder to cherry-pick than a handful of examples, and asking for them routinely — for every claim, not just the ones you dislike — keeps the request from reading as an attack.

A fourth pitfall worth naming: the decorative dashboard. A team builds a beautiful single-source-of-truth dashboard, links it in every channel, and then continues deciding exactly as it did before. The tell is simple. Ask, before any decision meeting: "what data would change my mind?" If the honest answer is *nothing*, you are opinion-driven with a data decoration on top. This is why the habits matter more than the tooling — a shared spreadsheet that people actually consult beats an elaborate BI stack nobody opens.

And the pitfall specific to the banner itself: posting a Wallpaper that says "Data beats opinions" while your experience section contains no evidence-driven examples is a mismatch a recruiter will notice. The banner is a claim about how you work; the profile below it is the evidence. If you're going to make the claim visually, back it with at least one specific instance — a test that changed your mind, a metric you moved, a decision you reversed when the numbers came in. That pairing is what makes the graphic function as a credibility signal rather than a slogan.

Making the evidence legible enough to actually win

Analysis buried in a spreadsheet never beats anything, because nobody acts on it. A finding only wins the argument if a non-analyst can follow it in under a minute, and a four-beat structure does that reliably:

Context — "we were choosing between cadence A and cadence B." Data — "we ran both against 200 matched prospects each over two weeks; A returned 2.6% replies, B returned 4.1%." Insight — "B outperformed at a sample large enough to take seriously, and the gap is wide, not marginal." Action — "we keep B, and we separately fix the rep-experience complaints that started this, because those were a real signal about *effort*, not about *results*."

Notice what that last beat does: it gives the original opinion-holder a win. The reps' complaint was legitimate information about workload, and honoring it costs nothing while making the next evidence-based reversal far easier to land politically. Teams that use data to humiliate people stop getting honest opinions, and then they lose the hypothesis-generation that made the whole loop work.

Two further practices make findings travel. First, publish the threshold alongside the result — "we said we'd switch at 3%, we got 4.1%" is dramatically more persuasive than "4.1% is good." It proves the bar wasn't moved after the fact. Second, when a test changes your mind, say so publicly. A short LinkedIn post describing a decision you reversed because of evidence does more for the credibility your Wallpaper is claiming than the Wallpaper does — and it attracts the people who work the same way.

Related questions

What size should a LinkedIn banner be?

LinkedIn's profile cover slot is 1584 × 396 pixels, a 4:1 ratio. Build to that exact size. Keep readable text within the middle horizontal band — your profile photo overlaps the lower-left corner, and mobile crops the outer edges more aggressively than desktop does.

Is SVG or PNG better for a LinkedIn cover?

Use SVG as your master file — it's a vector, so it scales and edits without blurring. Export a PNG at 1584 × 396 for the actual upload, since that's the format the profile slot handles most predictably across devices and browsers.

Does a banner actually affect how my profile performs?

Not on its own. A clean, on-message cover strengthens how a profile reads and signals a point of view, but reach and engagement depend on the substance and relevance of what you post. Treat it as a credibility signal rather than a growth tactic.

How do I apply "data beats opinions" with almost no data?

Write down the threshold that would change your mind, then run the cheapest test that could clear it — a two-week A/B, a one-week pilot on one segment, a properly sampled short survey. You need a decisive result, not a large dataset.

What's the difference between data-driven and data-informed?

Data-driven implies the number decides. Data-informed acknowledges that evidence is one weighted input alongside domain expertise, ethics, and context. The second is more honest for questions where the metric is a rough proxy or the sample is too thin to carry the decision.

FAQ

What does "Data beats opinions" actually mean for a decision?

It's a tiebreaker rule. When a claim can be checked against measurable evidence and that evidence disagrees with someone's gut, you go with the evidence — regardless of seniority. It doesn't ban intuition; intuition is how you decide what's worth testing in the first place. It only stops untested belief from overriding results you can verify.

What size is this Wallpaper, and where does it fit?

It's built to LinkedIn's official profile cover spec of 1584 × 396 px and ships as a scalable SVG, so it stays crisp at any display size. That makes it a direct fit for the cover slot. For a square post or a story frame you'd need to re-crop it, since the wide 4:1 ratio won't fill those shapes.

Is it free, and can I use it commercially?

Yes — free to use with no attribution required, including on a company or client profile. The thing to avoid is reselling the graphic itself as a standalone product. Using it as your banner, dropping it into a deck, or including it in a post is all fair game.

Can I change the colors and text?

Yes. Use the recolor picker on this page to match your brand colors or set a transparent background, then download the SVG or PNG. To change the wording, open the SVG in Canva, Figma, or PowerPoint — because it's a vector, you can edit text and layout without it going blurry. Keep contrast high: dark backgrounds need light text, light backgrounds need dark text, and avoid busy patterns behind the words.

Who is this banner actually right for?

It fits roles where you're expected to argue from evidence — RevOps, analytics, product, growth, and revenue leadership. It works less well if your profile makes no supporting claim; a cover asserting "data beats opinions" over an experience section with no measurable outcomes reads as decoration rather than as a position.

Doesn't this slogan risk shutting down debate?

It can, if it's used as a weapon. The healthy version invites disagreement and just insists it be settled with evidence on measurable questions. On questions that aren't measurable — ethics, brand, long-horizon bets — the phrase doesn't apply, and pretending it does is a misuse of it.

Sources

• LinkedIn Help — Add or edit a background photo on your profile
• Harvard Business Review — "Big Data: The Management Revolution" (McAfee & Brynjolfsson)
• Occam's Razor by Avinash Kaushik
• American Statistical Association — Ethical Guidelines for Statistical Practice
• Nobel Prize — Daniel Kahneman biographical (behavioral judgment research)
• W3C — Scalable Vector Graphics (SVG) specification
• W3C WAI — Contrast (Minimum) understanding document
• Wikipedia — Sample size determination

Related on PULSE

• ["Data beats opinions." — Quote Card](/knowledge/gb0135)
• ["Data beats opinions." — LinkedIn Banner](/knowledge/gb0053)
• ["Data beats opinions." — LinkedIn Banner](/knowledge/gb0274)
• [Turn data into decisions. — LinkedIn Wallpaper](/knowledge/gb0401)
• [NRR Beats New Logos — Revenue Law Banner](/knowledge/gb0460)
• [Outcome pricing beats seat pricing — RevOps Banner](/knowledge/gb0425)