The rules you never chose on purpose
Last session we watched one $92 Lumen journey produce eight different "winners" depending only on the rule we picked. Today we open up the first six of those rules - the heuristics - plus two B2B cousins, and look at what each one flatters and what each one hides. The word to hold onto: a heuristic is a hardcoded credit-splitting rule on an ordered path. It does not look at your data. It does not learn. It was decided once, by a person, and then applied to every journey forever.
Single-touch: first versus last 6 min live
The two simplest rules give 100% of the credit to a single touch. They are mirror images, and between them they run most of the marketing world. Here is the same Lumen journey - and what each single-touch rule does to it.
LiveFirst-touch, last-touch, and the one your platform runs3 min▶
Four of the eight rules are variations on "pick a single touch". The differences are small but the budget consequences are not.
- First-touch gives 100% to the opener. It flatters awareness channels - display, social - and is beloved by brand teams. It throws away everything the customer did on the way to buying.
- Last-touch gives 100% to the final touch. This is the default of virtually every ad platform - which is exactly why every platform's own dashboard claims it drove the sale. It over-credits demand-capture channels standing near the finish line.
- Last non-direct is last-touch with one tweak: if the final touch is "direct" (someone typing your URL), skip it and credit the touch before. This was the legacy default of Google's Universal Analytics. The logic: "direct" usually means the customer was already sold, so don't reward it.
The reason your Meta dashboard, your Google Ads dashboard, and your email tool each "prove" they drove the most revenue is that each runs last-touch on its own touches - so the same $92 gets claimed three times. Nobody is lying. They are all running the same single-touch rule on a different slice of the journey.
Self-studyWhen single-touch actively misleads you2 min read▶
Single-touch is not always wrong. On a one-touch impulse buy - see ad, click, purchase - last-touch is basically correct because there is only one touch. The danger scales with the journey.
| Sales cycle | Is single-touch safe? |
|---|---|
| Impulse / one session | Fine - there is barely a path to split |
| Lumen-style (7-21 days, 5 touches) | Risky - it discards 80% of the journey |
| Long cycle (over ~60 days) | Actively misleading - the opener and the nurture vanish entirely |
Rule of thumb: the longer the consideration window, the more a single-touch model lies to you. Above roughly a 60-day cycle, treat any single-touch report as one narrow opinion, never the verdict.
Multi-touch heuristics: linear, time-decay, position 6 min live
The next three rules stop crowning one touch and spread the $92 across all five. They feel fairer - and they are - but every one of them is still a rule you asserted, not a pattern anyone learned. Watch the same $92 land three different ways.
LiveThe three multi-touch rules, and their hidden knobs3 min▶
- Linear splits credit equally - $18.40 to each of five touches. Its story is "every touch mattered the same". Its blind spot: an impression is treated as worth exactly as much as the final click.
- Time-decay gives more credit to recent touches, using
weight = 2^(-days / half-life). On Lumen's 7-day half-life, display gets ~$6 and paid search ~$29. The catch: the half-life is a hand-set knob. Change 7 days to 30 and display suddenly earns three times more - same journey, same data, different number, because you turned a dial. - Position-based (U-shaped) gives 40% to the first touch, 40% to the last, and splits the remaining 20% across the middle. Its story: "creating and closing matter most, nurturing matters least". Defensible - and still a story you decided in advance.
Self-studyAll eight heuristics in one table3 min read▶
| Model | Fixed rule | Best for |
|---|---|---|
| First-touch | 100% to first touch | Awareness-led, short paths |
| Last-touch | 100% to last touch (the ad-platform default) | Impulse / one-session buys |
| Last non-direct | Last touch, but skip "direct" (GA Universal Analytics legacy default) | When direct = already-sold |
| Linear | Equal split across all touches | Simple, even-nurture journeys |
| Time-decay | Weight = 2^(-days/half-life); recent touches win | Short cycles, recency matters |
| Position / U-shaped | 40% first + 40% last + 20% split middle | Create-and-close funnels |
| W-shaped | 30% first + 30% lead-creation + 30% opportunity + 10% rest | B2B with CRM milestones |
| Full-path / Z-shaped | 22.5% x4 to first / lead / opp / closed-won + 10% middle | Long B2B sales cycles |
Notice the shape: every rule is a fixed set of percentages on an ordered path. None reads your data. That is the definition of a heuristic - and the reason none of them deserves the word "data-driven".
The B2B path models: W and Z-shaped 4 min live
Lumen is DTC, so its journeys are short. But if you ever run B2B - or a high-consideration DTC set purchase - you will meet two more heuristics built for long, milestone-driven journeys. They are still fixed rules; they just have more anchor points.
LiveW-shaped and Z-shaped, and why they need CRM milestones3 min▶
Single-touch and U-shaped only know about the first and last touch. B2B journeys have named stages in between - a lead gets created, an opportunity opens, a deal closes. The W and Z models anchor credit on those stages.
- W-shaped: 30% to the first touch, 30% to the touch that created the lead, 30% to the touch that created the opportunity, and 10% split across everything else. Three peaks instead of two - hence the W.
- Full-path / Z-shaped: adds a fourth anchor - closed-won. It gives 22.5% each to first touch, lead-creation, opportunity-creation, and closed-won, with 10% for the middle. Built for long enterprise cycles where the deal-close touch genuinely matters.
- The catch: both need CRM milestone tags - your system has to know which touch created the lead versus the opportunity. No milestones, no W or Z. That is why these live in B2B stacks (Salesforce, HubSpot) and rarely in DTC dashboards.
A B2B team switched from last-touch to W-shaped and watched their "hero" channel flip overnight - from the demo-request paid-search campaign (last touch) to the webinar that created the lead months earlier. Nothing about the deals changed. They just added two more anchor points to the same fixed rule, and suddenly the top-of-funnel content team got the recognition their pipeline actually depended on.
Self-studyWhy more anchors is still not "learning"2 min read▶
It is tempting to think W and Z are "smarter" than first-touch because they consider more of the journey. They do - but they are not smarter in the way that matters. The 30/30/30/10 and 22.5-x4 splits are still numbers a human wrote down. If the webinar in your funnel is genuinely worth 45% and not 30%, W-shaped will never find that out. It cannot. It has no mechanism to read your data and adjust. Every heuristic - from last-touch to Z-shaped - shares this exact ceiling: fixed rules, applied identically to every journey, forever. The models that break that ceiling are next session.
Match the model to your business ★ 10 min · pen and paper
No spreadsheet. Decide, out loud, which heuristic fits a business - and which one would actively mislead it. The point is to feel that model choice is a business judgment, not a technical one.
Take a short-cycle business first - Lumen's impulse-serum buyer, one session, sees an ad and buys. Which heuristic is basically fine here? (Last-touch - there is barely a path to split, so crowning the one touch is roughly correct.)
Now take a long-cycle business - a B2B software deal, 90 days, a webinar, three emails, two demos, a final search. Which heuristics are still safe? (Not single-touch - it discards the webinar that created the lead. You need W or Z-shaped, and only if your CRM tags the milestones.)
Write the boundary you just found: above roughly a 60-day cycle, single-touch models are misleading because they throw away the touches that did the persuading.
Now place your own business on that line. What is your real consideration window - hours, weeks, months? Say the model that fits it, and the model your tools are probably running instead.
Spot the default: find what your tool runs - and what it flatters ★ 10 min · on your reports
Every reporting tool has a default attribution model, and almost none of them announce it. Today we find Lumen's - and the flattery baked into it. Bring your own tool next; today we use Lumen's stack.
Open Lumen's channel report. Paid search shows 41% of "attributed" revenue, paid social 12%, display 3%. Ask the first question: which model produced these numbers? (It is last-touch - the platform default. Nobody labeled it, but the shape gives it away: capture channels on top, awareness channels at the bottom.)
Now ask what the default flatters. Last-touch flatters whoever stands nearest the sale - paid search, retargeting. It starves whoever opened the journey - display, social. On Lumen's data, 68% of those paid-search conversions had an earlier display or social touch that got zero credit.
Find the toggle. In GA4 the default is now data-driven; in your ad platforms it is last-touch on their own touches; in Universal Analytics it was last non-direct. Write down, for each tool you use, the model it silently runs.
Say the verdict a leader should say: "Our main report is last-touch, it flatters demand-capture, and it is a fixed rule - not data-driven. Before I move budget on it, I want to see the same channels under a data-driven model."
A brand assumed its "AI-powered attribution" was data-driven because the vendor said so. It was time-decay with a 7-day half-life - a hardcoded rule with a dashboard. When they changed the half-life to 30 days on a whim, every awareness channel tripled its credit. Nothing about the customers changed. They had been steering a $4M budget with a hand-turned knob and calling it intelligence.
This week ◐ 25 min total
- Name every default. List each reporting tool your team uses and write down the attribution model it runs by default. If you cannot find it, that itself is the finding - ask your analytics owner.
- Catch a "data-driven" that isn't. Find one place in your stack (a vendor deck, a dashboard label, an internal report) where a heuristic is being called "data-driven" or "AI". Bring it to Session 3.
- Do the half-life test on paper. Take your longest customer journey and split it under time-decay with a 7-day half-life, then a 30-day half-life. Feel how much the "winner" moves when you turn a knob no customer ever touched.
- Optional: skim the Google Analytics attribution-model reference (Skillshop) - it names all these rules and shows exactly where GA4 changed the default.
Three questions before you go 🎯 ◐ 90 seconds
1 · Which attribution model is the default of virtually every ad platform?
Ad platforms default to last-touch on their own touches - which is why each platform's dashboard claims it drove the sale. The same conversion gets credited by Meta, Google, and your email tool all at once.
2 · In time-decay attribution, the half-life (e.g. 7 days) is best described as...
The half-life is hand-set. Move it from 7 days to 30 and awareness channels earn far more credit on the exact same journey. A dial you turn is the signature of a heuristic.
3 · A vendor calls their first/last/linear/time-decay attribution "data-driven". What is the accurate correction?
Heuristics are hardcoded credit-splitting rules. They apply the same fixed percentages to every journey and never read your data. Calling them "data-driven" is the most common leader-level error in the field.
What this session covers
This session distills the heuristic-models chapter of the leading attribution courses into a leader-first framing - naming each rule, its default status, and its bias. Certificates and full video courses stay with their official sources.