Top of the Funnel: Why Most Businesses Get It Wrong (and the Fix)

What you'll build

Your cost per lead doubled last quarter while demo requests stayed flat. The retargeting budget is quietly carrying the whole pipeline, and most businesses call that a bottom of funnel problem. It isn't.

Top of funnel is not a traffic problem. It is a memory problem. The job is to be known as a member of your category, linked to the situations that trigger buying, and trusted before anyone raises a hand. Cheap clicks fail because CPC is a metric that correlates with none of those three things.

In this tutorial you'll build a working top of funnel system: Category Entry Points pulled from real customer conversations, eight trust-first assets mapped to them, a distribution cadence you can sustain, and a five-line dashboard that proves the whole thing moved. No fluff. No retainer required.

Prerequisites

• Read access to Google Analytics (ideally a BigQuery export) and Google Search Console.
• Eight customers and five lost deals you're allowed to interview for 20 minutes each.
• A content home you own: newsletter, blog, or podcast feed. Social alone doesn't count.
• Roughly four hours a week for 90 days. Boring consistency beats bursts.

Step 1: Accept that 95% of your market isn't buying yet

Start with the number that should reframe every budget conversation: the 95/5 rule, popularized by LinkedIn's B2B Institute and built on Ehrenberg-Bass in-market research. At any moment, roughly 95% of buyers in a category are not in-market. Only about 5% are actively shopping.

If your entire top of funnel is high-intent keyword bidding, you are renting the 5% at the moment of maximum competition and building zero demand for the rest. That is not awareness. That's a bidding war with a content problem attached.

The supporting evidence is just as blunt. Gartner's buying journey research found buyers spend only about 17% of their purchase journey time meeting with suppliers.

The other 83% is independent research, peer conversations, and internal consensus building. Translation: most of the trust that closes a deal is built in content, not on calls.

And it's not a straight line. Google's 'Messy Middle' research describes a loop between exploration and evaluation, governed by mental shortcuts and urgency.

Buyers fall back into exploration long after they've shortlisted you. Which means your awareness content has to be re-encountered, not consumed once and filed away.

Step 2: Kill the three habits that fake awareness

Before you build anything, remove the patterns that generate activity without memory.

1. Buying the cheapest inventory. Programmatic remnant placements cost a fraction of LinkedIn's typical CPMs (commonly $25 to $40, with CPCs of $5 to $12+), and they deliver near-zero brand effect. A neutral or negative association is worse than invisibility.
2. Gating everything. Gating top of funnel trades reach for a form fill. Gate the comparison guides and benchmarks further down, and leave awareness content open.
3. Optimizing for MQLs. MQL targets reward capture over demand. Sales then works low-intent leads, close rates drop, and marketing gets cut. The failure is self-fulfilling.

Here's where it shows up in the numbers. A cheap click that never converts is the easiest way to feel productive for a quarter, which is exactly the pattern behind a Meta ads teardown. Fixing top of funnel usually means paying more per impression to reach fewer, better-matched people.

Step 3: Mine trigger events, not keywords

Keywords describe your category. Category Entry Points describe your buyer's life. Those are different lists, and only one of them generates demand.

Run this with a real example. Call it Helio, a 40-person field-service SaaS at about $6M ARR spending roughly $10k a month, with most of it pointed at high-intent search and retargeting.

MQLs had plateaued and CAC was climbing. In two weeks they interviewed eight customers and five lost deals and logged the exact wording around the moment they started looking for a solution: 'we hired our 20th tech,' 'we failed an audit,' 'our dispatcher quit.'

Those quotes are the asset. They become your content calendar, and a top of funnel plan instead of keywords nobody says out loud.

Build an AI answer baseline in 30 minutes

Then record what the machines say. Ask Google's AI Mode, ChatGPT, and Perplexity the 20 questions your buyer asks before they know vendors exist.

Screenshot which brands and sources get cited. Revisit in 90 days. That screenshot deck is your rough equivalent of share of search in AI answers, and it's free.

Step 4: Make content AI can't flatten

This is the part most teams are still getting wrong. According to Pew Research Center's 2025 study of AI summaries in search, users clicked a link in roughly 8% of visits when an AI summary was present, versus about 15% without. Sessions ended about 26% of the time with a summary, against 16% without.

An Ahrefs study the same year estimated AI Overviews cut click-through rate for the number one organic position by roughly 34%, with informational queries hit hardest. And SparkToro's zero-click research puts roughly 58% to 60% of US Google searches ending without any outbound click. Treat those figures as directional and re-verify, but the direction is settled.

So the content that wins now has to contain something an LLM can't average together from other pages:

• Proprietary data. Survey 100 operations managers and publish the raw numbers, not the interpretation.
• A named opinion. 'Most dispatch software over-indexes on scheduling and under-indexes on payroll reconciliation.' Disagreeable and specific.
• Lived teardowns. Real dashboards, real before-and-after numbers, permission in writing.

There's a second reason this works. The Edelman and LinkedIn thought leadership research series has consistently found a majority of decision-makers (around 55%) use thought leadership to vet organisations they hadn't previously considered.

Your awareness content has a vetting function, not just a reach function. And it increasingly happens on Reddit, where buyers ask questions they'd never post under their own name on LinkedIn.

Step 5: Wire the measurement before you publish

Nothing kills a top of funnel program faster than last-touch attribution making it look worthless while retargeting looks heroic. So set the leading indicators first.

The cleanest cheap KPI is branded session share. Pull it from the GA4 BigQuery export:

-- GA4 BigQuery export: branded vs non-branded session share, monthly
WITH s AS (
  SELECT
    PARSE_DATE('%Y%m%d', event_date) AS d,
    CONCAT(user_pseudo_id, '-', CAST(
      (SELECT value.int_value FROM UNNEST(event_params) WHERE key = 'ga_session_id')
      AS STRING)) AS session_key,
    (SELECT value.string_value FROM UNNEST(event_params)
      WHERE key = 'page_location') AS url
  FROM `proj.analytics_123456.events_*`
  WHERE _TABLE_SUFFIX BETWEEN '20260101' AND '20260630'
    AND event_name = 'session_start'
)
SELECT
  DATE_TRUNC(d, MONTH) AS month,
  COUNT(DISTINCT IF(REGEXP_CONTAINS(LOWER(url), r'helio|heliofield'),
    session_key, NULL)) AS branded_sessions,
  COUNT(DISTINCT session_key) AS all_sessions,
  ROUND(SAFE_DIVIDE(
    COUNT(DISTINCT IF(REGEXP_CONTAINS(LOWER(url), r'helio|heliofield'),
      session_key, NULL)),
    COUNT(DISTINCT session_key)), 4) AS branded_share
FROM s
GROUP BY 1
ORDER BY 1;

Expected output: one row per month with three columns. For Helio, month one read 0.09 branded share, month three read 0.13.

That is a real signal, and it costs nothing to track. If you don't have BigQuery wired up yet, the honest cost of doing it yourself versus paying someone is covered in this breakdown of server-side tracking costs.

Pair it with a five-line dashboard, reviewed monthly, nothing else:

1. Branded session share (above).
2. Branded impressions in Search Console versus non-branded.
3. Newsletter reply rate, not open rate. Replies are the trust signal.
4. Self-reported attribution: add a required free-text 'How did you hear about us?' to your demo form and count how many name an awareness touch.
5. AI answer citations across your 20 baseline prompts.

Step 6: Two channels, 90 days, no negotiating

Six half-channels beat nothing. Helio picked exactly two: founder-led LinkedIn and a monthly newsletter, plus 15 minutes a week commenting in three communities.

Weekly cadence: Monday a founder post tied to one Category Entry Point, Tuesday and Thursday two short clips, Wednesday the newsletter, Friday the comments. Amplify the single best-performing organic post each month as a Thought Leader Ad, around $1,500. Paid pointed at content that already worked beats paid pointed at polished brand creative in most B2B feeds.

The budget split on $10k a month looked like this: 40% production (writer, editor, survey costs amortized), 30% distribution and amplification, 20% tools and data, 10% held for experiments. High-intent search dropped to 60% of its former share, not to zero.

That last part matters. Binet and Field's 60/40 finding from the IPA Databank is contested for B2B, but the direction holds, and cutting brand spend in a bad quarter produces share losses that show up four to eight quarters later.

Common pitfalls

• Treating this as a traffic project. If your success metric is sessions, you'll drift straight back to cheap clicks.
• Publishing 'what is X' filler. That is precisely the content AI summaries now absorb.
• Pitching in the first touch. Once the audience learns your content is an ad, unlearning it is expensive.
• Optimizing before you have traffic. If your site is small, fixing obvious conversion leaks beats running split tests, and this piece on low-traffic A/B testing explains when to test and when to just fix.
• Letting good attention rot. Awareness that never gets followed up is wasted, and lead reactivation done badly is how a working top of funnel quietly becomes an expensive list.

Next steps

1. Book the eight customer interviews this week and pull trigger-event quotes verbatim.
2. Log your 20-prompt AI answer baseline and screenshot it.
3. Ship one asset from the Step 4 list before you redesign anything.
4. Run the BigQuery query and write down your branded share today, so you have a starting line.

Then leave it alone for 90 days. Top of funnel rewards patience and punishes improvisation. The teams that win here aren't cleverer, they just stopped rebuilding the strategy every three weeks.

Rather have this built for you?

If the system above makes sense but the 90 days of consistency is the part that keeps sliding, that's the gap we close. Start with the free AI audit to see where your funnel leaks in minutes, and you'll know which of these steps is actually your bottleneck before you spend anything.

Cover photo by Vitaly Gariev on Pexels.