learn-metric-decomposition-with-phoebe / Builder session 6 of 10
Learn Metric Decomposition with Phoebe · Builder track · Session 6 of 10

Branding deep dive

Brand is the metric everyone calls unmeasurable, then quietly measures badly. It feels soft - awareness, preference, love - but it still decomposes. A reached population is filtered through awareness, consideration, preference, and intent until a fraction of it buys because of the brand. Today you build that brand funnel as a live tree, find the leading signal that predicts market share months early, and confront why brand metrics get gamed more than any other.

🟡 Builder track Analysts · PMs · founders · ops Live tree simulator + SQL 45 min
0-3 · Recap 3-20 · The brand funnel 20-42 · Leading signal & the drop 42-45 · Q&A
Part 0

Where we are

In b5 you decomposed marketing revenue - the funnel that turns spend into won deals. Branding is the funnel that feeds the top of that one: it decides how much of the market even considers you before any campaign runs. It is the hardest deep-dive because the metric is soft, the lag to sales is long, and the temptation to game it is huge. If you can decompose brand, you can decompose anything.

Live - presented in session Self-study - read after class ▶ Live tree - editable & runnable Framework sources covered
★ What you walk out with today A brand funnel written as Reached x awareness x consideration x preference x intent x purchase x value, a second live tree for Share of Search - the best-validated leading proxy for future market share - and a clear-eyed view of why brand metrics get gamed and how the long lag to sales invites Goodhart trouble.
Part 1 · covers the multiplicative pattern applied to a soft metric

Even brand decomposes 7 min live

The hard case first. Brand feels like a single fuzzy quantity, but brand-driven revenue is a funnel just like marketing revenue - only the stages are mental states instead of ad events. A reached population is filtered stage by stage: how many are aware, of those how many consider you, then prefer you, then form intent, then actually purchase because of the brand. Multiply the survivors by deal value and you have brand-driven revenue.

Reached 5,000,000 Aware 2,000,000 Consider 500,000 Prefer 150,000 Intent 30,000 Buy 3,000 x 40% x 25% x 30% x 20% x 10% Each band is the survivors of the rate above. 3,000 brand-driven buyers x $150 value = $450K brand revenue.
🔍 Click to zoom - the brand funnel filters a reached population down to brand-driven buyers
LiveWhat each stage really means3 min

The stages are mental states, and each maps to a real-world signal you can actually measure - which is what stops brand from being hand-waving.

  • Reached - the population your media and presence physically touched. Impressions, distribution, footfall.
  • Aware - they know you exist. Aided and unaided awareness surveys, search interest.
  • Consider - you make their shortlist. Consideration surveys, add-to-comparison, category search including your name.
  • Prefer - you are their first choice. Preference surveys, branded-search share, direct traffic.
  • Intent - they mean to buy. Cart adds, demo requests, quote requests, wishlist saves.
  • Buy - they purchase because of the brand, not a discount. Branded-organic conversions, direct sales.
Real world

The leak is almost never at the top. Big brands pour money into Reached and Aware, then lose most of it at Consider and Prefer - a huge audience that knows you and still would not pick you. Decomposing brand exposes exactly that: awareness without preference is a bucket with a hole in the middle.

Self-studyWhy brand is still multiplicative2 min read

Every stage is a conditional rate - the fraction who consider you given they are aware, the fraction who prefer you given they consider you. Multiply the conditional rates and you get the share of the reached population that survives to purchase. Soft inputs, hard arithmetic. The pass-through rule from b2 holds exactly: preference falls 20%, brand-driven revenue falls 20%, no matter how strong awareness is.

  • Conditional rates multiply - each is a "given the last stage" probability.
  • The weakest stage caps the funnel - a great awareness number cannot rescue a broken consideration rate.
  • Soft does not mean unmeasurable - it means each stage needs a named proxy signal, not a survey alone.
Part 3 · covers Goodhart risk, vanity metrics

Why brand metrics get gamed 4 min live

Brand is the most gamed funnel because the top stages are cheap to inflate and the payoff (sales) is far away, so nobody is caught for months. Knowing the traps is half of reading brand honestly.

LiveThe three traps3 min
  • Vanity reach - buying enormous cheap impressions to make Reached and Aware soar, while consideration and preference flatline. A big top with no middle.
  • Awareness without consideration - being known but never shortlisted. The stage-to-stage rate from Aware to Consider quietly collapses and the top-line looks fine because awareness carries it.
  • The long-lag Goodhart trap - because sales lag brand by months, a team can hit its brand target this quarter with tactics that never convert, and the bill only arrives later. When a proxy becomes the target, it stops being a good proxy.
The defence Watch the stage-to-stage rates, not just the top of the funnel. A healthy brand grows Consider and Prefer, not only Reached and Aware. And triangulate the leading signal (Share of Search) against the lagging one (sales) - if search share is flat while awareness spend soars, you are buying vanity, not brand.
Demo 1 of 2

Diagnose a brand-funnel drop ★ 12 min · everyone builds

Here is the brand funnel as a live tree. Read the baseline, then press Simulate a drop - one stage falls and its path up to brand revenue lights coral. Name the stage that fell, and say which real-world signal you would check to confirm it.

{
  "unit": "$",
  "root": {
    "label": "Brand-driven revenue", "op": "x",
    "children": [
      { "label": "Reached",       "value": 5000000, "unit": "people" },
      { "label": "Awareness",     "value": 0.40, "pct": true },
      { "label": "Consideration", "value": 0.25, "pct": true },
      { "label": "Preference",    "value": 0.30, "pct": true },
      { "label": "Intent",        "value": 0.20, "pct": true },
      { "label": "Purchase",      "value": 0.10, "pct": true },
      { "label": "Value / buyer", "value": 150, "unit": "$" }
    ]
  }
}

Read the baseline: 5,000,000 reached, filtered stage by stage to 3,000 brand-driven buyers, x $150 = about $450K.

Press Simulate a drop. One stage falls. Follow the coral trail - is it awareness, consideration, preference, intent?

Map the stage to a signal: awareness to search interest and surveys, consideration to shortlist and branded search, preference to direct traffic and preference surveys.

Say it: "Brand revenue fell X% because [stage] fell Y% - I would confirm it with [signal]." Press Reset and repeat.

Real world

A famous pattern: awareness holds, consideration cracks. A well-known brand can stay top-of-mind for years while quietly falling off shortlists as a fresher rival appears. The top-line brand number looks stable because awareness is sticky - but the Aware-to-Consider rate is bleeding, and sales follow a year later. The tree catches it the moment you look at the stage rates instead of the top.

Demo 2 of 2

Your turn ★ 10 min · everyone builds

Three tasks: edit the brand funnel, build a Share-of-Search tree, then a short thinking exercise.

LiveQ1 · Awareness up, consideration down3 min

In the tree above, raise Awareness from 40% to 55% and drop Consideration from 25% to 15%. Watch brand revenue. Even though more people now know you, fewer shortlist you, and because the stages multiply, the funnel shrinks. This is the "reached but not converting" brand: loud, familiar, and losing - the exact shape vanity spend produces.

The point A rising top stage cannot rescue a falling middle stage in a multiplicative funnel. Awareness is the cheapest stage to buy and the easiest to fool yourself with. Preference is where brands are actually won or lost.
LiveQ2 · Build a Share-of-Search tree4 min

Share of Search is your search volume as a fraction of the whole category. You cannot divide inside the tree, so model the category as an additive tree - total category search is the sum of every brand's searches, and yours is one leaf. Read your share by eye: your leaf over the total. Edit your leaf up and watch your slice of the category grow.

{
  "root": {
    "label": "Category searches / mo", "op": "+",
    "children": [
      { "label": "Your brand",     "value": 90000,  "unit": "searches" },
      { "label": "Competitor A",   "value": 140000, "unit": "searches" },
      { "label": "Competitor B",   "value": 110000, "unit": "searches" },
      { "label": "Long-tail rest", "value": 60000,  "unit": "searches" }
    ]
  }
}

At baseline your share is 90,000 / 400,000 = about 23%. Push your leaf to 130,000 and your share crosses 30% - the signal that tends to precede a market-share gain months later.

Self-studyQ3 · Why does Share of Search lead market share?take home

Write a paragraph explaining why a rise in Share of Search tends to precede a rise in market share, rather than just reflecting it. Ground your answer in the funnel: search is an upstream mental act - it captures demand at the moment it forms, before it converts into a purchase weeks or months later. Because it is measured as a share of the whole category, it strips out seasonality and market growth, isolating the change in your relative brand strength. That relative strength is what eventually shows up as share of wallet - so watching search share is watching the funnel fill before the sales register rings.

Homework

Try it yourself - this week ◐ 20-30 min total

Framework sources

Frameworks this session draws on

Brand measurement borrows from marketing science and the metrics-design literature. This page draws on:

Brand funnel & Share of Search (marketing-science canon) - reach to preference to purchase, and the leading search proxyParts 1-2 · the brand funnel and its best-validated leading signal
Goodhart's law - when a measure becomes a target it stops being a good measurePart 3 · why the long lag invites gaming
Amplitude North Star Playbook - leading inputs versus lagging outputsPart 2 · search leads, sales lag
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · The brand funnel (Reached x awareness x consideration x preference x intent x purchase x value) is which pattern?

Each stage is a conditional rate (consider given aware, prefer given consider). They multiply, which is why a rise in one stage cannot rescue a fall in another and the weakest stage caps the whole funnel.

2 · Which is the best-validated leading proxy for future market share?

Share of Search captures demand as it forms and, being a share, strips out seasonality and market growth. It tends to turn up months before market share. Raw volume moves with the whole category; surveys lag and cost more.

3 · Why do brand metrics lag sales and invite gaming?

The top stages are cheap to inflate and the payoff (sales) arrives months later, so gaming is not caught for a long time. That long lag is exactly the Goodhart trap - the defence is watching stage-to-stage rates and triangulating leading against lagging signals.

Builder session 6 cheat sheet · pin this

Brand funnelReached x awareness x consideration x preference x intent x purchase x value = brand-driven revenue.
Soft, still multiplicativeStages are conditional rates. They multiply, so the weakest stage caps the funnel.
Each stage has a signalAwareness = search + surveys · Consider = shortlist · Prefer = direct traffic · Intent = cart or demo.
Share of SearchYour share of category search. The best-validated leading proxy for future market share.
Share, not volumeRaw search rises every season for everyone. Only your share isolates relative brand strength.
The vanity trapBuying Reached and Aware while Consider and Prefer flatline. A big top, no middle.
Long-lag GoodhartSales lag brand by months, so a gamed proxy is not caught until later. Watch stage-to-stage rates.
Running skillBuild the brand funnel, find the leading signal, spot the gaming. Next: b7, internet and traffic.