(7) HubSpot Inbound Marketing Certification 7: Maximizing ROI With Marketing Attribution & Experiments

Full Set from Section 7 derived from official transcripts. Author name: The transcript says “James C. Collin’s,” but the book’s author is Jim (James C.) Collins. I corrected it. Outdated tool: Google Analytics’ A/B testing (Google Optimize) was shut down in 2023, so the “10 versions” feature no longer exists. The exam may still use the course answer, so I kept the card, but don’t count on it in real life.

Why is reporting on marketing performance overwhelming?

Many channels (PPC ads, website, messaging apps, social) and touchpoints (specific ads, blog posts, pages, emails) create too many data points to pull together manually.

Marketing attribution reporting

Shows the impact marketing had on a specific goal, usually a purchase or sale. It compiles customer data in one place and assigns credit to the channels and touchpoints that influenced the buyer's journey.

Davis Mastin (HubSpot) definition of attribution

"Attribution surfaces which interactions a person or group of people took along their journey toward a desired outcome or 'conversion' point."

Attribution models

The statistical models that calculate how credit is assigned to interactions. They can be customized to your organization's sales cycle.

Ultimate benefit of attribution reporting

Determining the ROI of each channel or touchpoint, so you spend budget effectively and tailor campaigns and content to personas.

Barn2Door: who they are

A Seattle software company that lets farmers sell directly to consumers online (stores, bookkeeping, taxes) and digitize record-keeping.

Barn2Door buyer's journey

Facebook ads promote eBooks, blogs, worksheets, and checklists; an email address adds the lead to the CRM; nurture with success stories, testimonials, webinars; then a video demo.

Barn2Door's most important qualifying step

The video demo. It signals the lead's interest and tells sales to reach out.

What did attribution give Barn2Door that they lacked?

They already knew how people discovered them, but needed insight into what came after: which content influenced leads before talking to sales. They cut low-ROI content, saving time and resources.

Why aren't page views enough (per Cade Midyett)?

They don't show whether people are actually warming up or will eventually sign up.

How Barn2Door uses attribution proactively

HubSpot lead scoring assigns qualification points to key content interactions based on their historic impact, signaling good-fit prospects to Marketing and Sales.

Should reporting be reactive or proactive?

Reporting needs to be more than a reactive strategy. Use it proactively, e.g., lead scoring.

How often does Barn2Door review attribution and historical data?

Monthly, so nothing catches them off guard and anomalies can be investigated.

Barn2Door's advice on analyzing attribution reports

Always analyze them in the context of all the data you track. This informs which models fit, shows the full data story, and increases confidence.

How Barn2Door got started with attribution

Started simple with one attribution report on their general marketing dashboard, gradually customized it, and compared models over months until finding the best fit.

Questions Barn2Door's attribution dashboard answers

ROI of last quarter's Facebook ads; which content drove the most leads last month; where to invest resources; which pages are viewed most before someone becomes a customer.

The "age-old question" attribution reporting answers

What is the ROI of my marketing efforts?

Attribution model (definition)

Lets you apply different credit to each interaction according to the model's rules.

What determines which attribution model to use?

Your team's goals and the expected supporting analysis. There's no one-size-fits-all; test models over time within larger dashboards and datasets.

Single-source attribution models

Assign all credit to one touchpoint, usually first touch or last touch.

First touch (first interaction) attribution

Gives 100% credit to the first interaction that led to the conversion. Best for learning which content or channel initially brings visitors to your site.

Last touch (last interaction) attribution

Gives 100% credit to the last interaction before a conversion, like a closed-won deal.

Example: Organic search visit, then a week later a Facebook ad click and purchase. Who gets credit?

First touch: organic search gets 100%. Last touch: the Facebook ad gets 100%.

Multi-touch attribution models

Give each contributing channel a slice of credit, accounting for the entire customer journey.

Linear attribution

Gives equal credit to every interaction before the conversion. Great when prospects stay in the consideration phase for extended periods.

Full-path attribution

22.5% each to: first interaction, contact creation, deal creation, and deal close. The remaining 10% is split evenly across all other interactions.

When to use full-path attribution

To understand how your marketing impacts revenue generation.

Time decay attribution

Gives more credit to interactions closer in time to the conversion. Ideal for measuring short-term touchpoints like campaigns.

Scenario: Maria (AvocadoToasted) ran Instagram ads and wants to prove ROI and see what drives NEW traffic. Which model?

First touch.

Who benefits most from first touch attribution?

Demand generation marketers highlighting campaigns that first introduced customers to the brand, regardless of outcome.

Limitations of first touch attribution

Gives little insight into which parts of the journey to optimize; hard to use to justify impact on the bottom line.

Scenario: Charlie (CleanCut) added pop-up forms to top blogs to increase blog conversions. Which model?

Last touch.

When is last touch attribution most powerful?

With a short buying cycle, where leads don't have many touchpoints before buying.

Limitation of last touch attribution

Ignores influences on the path to conversion; a lead with a dozen prior interactions gets zero visibility into them.

Scenario: Luisa (CatchAVibe) removed redundant form fields on landing pages reached via an email nurture. Which model?

Time decay.

Why time decay fits Luisa's test

It recognizes every interaction (email and landing page) while placing the most value on touches closest to conversion, like the form.

Time decay is especially important for...

Optimizing touches that drive conversions or increase near-term conversion likelihood, and analyzing conversion paths and sequences.

Limitation of time decay attribution

Can't recognize the interaction that originally introduced the customer to the brand, so it's not the full story.

Scenario: Malik (NoBones, new RevOps manager) wants a year of channels, offers, and initiatives that funneled qualified leads and customers to Sales. Best model?

Full-path. It highlights high-impact conversions without fully discounting assisting interactions, covering macro and micro conversions.

When would linear be the better choice for Malik?

If he were just getting started with attribution and needed a sense of the overall customer journey.

Limitation of linear attribution

Not every interaction equally propels a purchase. Low-value touches like email clicks get the same credit as high-value ones like demo requests.

Why weren't first/last touch best for Malik?

They show a limited scope of what influences buying behavior. Better for later reports on lead generation or areas for improvement.

Are there definite rules for picking an attribution model?

No. Some fit certain situations and questions better; it depends on the report and its purpose.

Why experiment instead of just reporting?

To move from reactive marketing to proactively solving for customers by turning data into action, e.g., finding friction in the buyer's journey (often via conversion rates).

"Fire bullets, then cannonballs"

From Jim Collins' "Great by Choice." Fire low-cost, low-risk, low-distraction experiments (bullets); once data proves success, concentrate resources in a big bet (cannonball).

Marketing experiment

A form of market research that aims to discover new strategies for future campaigns or validate existing ones.

Marketing experiment analogy

An insurance policy on future marketing efforts that minimizes risk and aligns efforts with desired results.

Most common way to start marketing experiments

A/B testing.

A/B testing (split testing)

Splitting your audience to test variations of content: two versions with a change to a single variable, shown to two similarly sized audiences over a set period.

A/B testing benefit: increased website traffic

Test blog post or webpage titles to increase clicks to your site.

A/B testing benefit: higher conversion rate

Test CTA location, color, or anchor text to increase clicks to landing pages and form submissions.

A/B testing benefit: reduced bounce rate

Test blog introductions, fonts, or featured images to retain visitors.

A/B testing benefit: reduced cart abandonment

Test product photos, checkout page designs, and where shipping costs appear. Average ecommerce cart abandonment is about 70%.

Three most common A/B tests

Landing pages, CTAs, and email.

How landing page A/B testing works

One URL, two or more page versions; visitors are randomly assigned, and a cookie ensures they see the same variation each time for statistical validity.

What to test on landing pages

Offers, copy, and form fields.

Landing page test example: both viewed 180 times; one got 20 clicks, the other 5

Keep the 20-click version; it's optimized for the desired action.

HubSpot blog search bar test: why run it?

Non-bounce desktop users who engage with search have a 163.8% higher blog lead conversion rate, but very few use the search bar.

HubSpot blog search bar test: winning variant

Variant C: more prominent, "search the blog" placeholder, blog-only search. +3.4% conversion rate, +6.46% search engagement. Primary metric: offer thank-you page views.

How CTA A/B testing works

Create two or more CTA variations on the same page, show them randomly, and see which gets the most clicks.

Which CTA rate should you ultimately optimize?

View-to-submission rate. A low click-to-submission rate may mean the problem is the landing page, not the CTA.

Can you A/B test the CTA and landing page at the same time?

No. Run one test at a time and change one variable so you know what caused the result.

What to test on CTAs

Placement, size, color, copy, and button graphics.

HubSpot mobile CTA test: background

Earlier tests showed mobile users were 44% more likely to click through and 18% more likely to convert with a single, non-exitable bottom CTA bar.

HubSpot mobile CTA test: results

A (control): unchanged. B (minimize caret): +7.9%. C (X to dismiss): -11.4%. D (no X/minimize): +14.6%, projected about 1,300 extra submissions/month.

How email A/B testing tools work

They randomize recipients into two or more groups (large enough for significance), assign variations, and can send the winner to the rest of the list.

Email A/B testing: open/click rate vs. conversions

Most tools pick winners by open rate or CTR, but also track which variation (with the right landing page) drives the most conversions. This requires integrating email with marketing analytics.

What can you A/B test in emails?

Format, layout, timing, sender, subject lines, and target group.

HubSpot review-request test: results

Plain-text email vs. templated email vs. in-app notification. Emails beat in-app by 140%: 24.9% of email openers left reviews vs. 10.3% for in-app.

Why did in-app notifications underperform?

Users often overlook or miss them.

What else can be A/B tested beyond landing pages, CTAs, and email?

Web forms and key page layouts, like your pricing page.

A/B test best practice: pick one variable

Isolate one "independent variable" to know what caused the change. Simple changes (email image, CTA wording) can drive big improvements and are easier to measure.

Independent variable

The single element you change in an A/B test.

Dependent variable

The primary metric you choose before running the test; it changes based on the independent variable.

When should you choose your primary metric?

Before running the test, even before setting up the second variation. Also consider forming a hypothesis.

Control vs. challenger

Control: the unaltered version as it exists. Challenger: the altered version tested against it.

Control/challenger example

Testing whether a testimonial boosts landing page conversions: control has no testimonial, challenger has one.

Splitting sample groups

For tests where you control the audience (like email), use two or more equal-sized, random groups.

HubSpot pro tip on email A/B splitting

HubSpot Professional and Enterprise automatically split traffic so each variation gets a random sample.

Sample size: email vs. web page

Email: send to a large-enough subset, then send the winner to the rest. Web page (no finite audience): test duration determines sample size, so run long enough for substantial views.

Statistical significance (A/B testing)

When results show the variables tested (e.g., conversion rate and landing page type) aren't random but influence each other, so you don't waste money on campaigns that won't deliver.

Common confidence threshold

Many tools use 95%, but a lower threshold may be fine if the test doesn't need to be as stringent.

Matth Rheault's statistical significance analogy

It's like placing a bet: 80% confidence is like being 80% sure and betting everything on it.

When do you need a HIGHER confidence threshold?

When testing something that only slightly improves conversion rate, since random variance plays a bigger role.

When can you use a LOWER confidence threshold?

For radical changes likely to improve conversion 10%+, like a redesigned hero section.

Rheault's takeaway on radical vs. specific changes

The more radical the change, the less scientific you need to be. The more specific (button color, microcopy), the more scientific you should be.

Significance calculators and tools named in this lesson

HubSpot's A/B testing calculator, Optimizely, SurveyMonkey.

A/B testing tools named in the course

HubSpot (emails, CTAs, landing pages); Google Analytics (up to 10 full versions of a single page for non-HubSpot customers).

Why test both variations simultaneously?

Timing (time of day, day, month) affects results. Running A and B at different times makes it unclear whether design or timing caused the change.

Exception to testing simultaneously

When you're testing timing itself, like optimal email send times.

How long should an A/B test run?

Long enough for a substantial sample. It could take hours, days, or weeks; traffic is a major determinant, and low-traffic sites need longer.

Focus during A/B test analysis

Your primary goal metric. E.g., if leads are the goal, don't get distracted by open rate or CTR.

Next step when one variation wins

Disable the losing variation in your A/B testing tool.

Next step when neither variation is statistically better

Mark the test inconclusive; stick with the original or run another test, using the data to iterate.