SEO & content marketing: real-world application

By November 2023, blog.hubspot.com was probably the largest content property in B2B software. Ahrefs, which sells the SEOSEOSearch Engine Optimization: the practice of improving your pages' natural (unpaid) rankings in search engine results pages to attract more organic traffic.View full definition → toolset those estimates come from, put it somewhere near 13 million organic visits a month. A year later the same source showed roughly three quarters of that gone. Over the same stretch HubSpot's revenue went from $2.17 billion in 2023 to $2.63 billion in 2024. Hold those two lines next to each other, because everything useful in this lesson sits in the gap between them: the most copied content engine in the industry lost most of its traffic and the business kept compounding. One company, followed from 2006 to the rebuild that is still running.

Core concept: what HubSpot actually built

HubSpot began blogging in 2006, the year Brian Halligan and Dharmesh Shah founded it, and named the thing it was doing "inbound marketinginbound marketingA strategy that attracts prospects organically via valuable content (blog, SEO, social) rather than interrupting them.View full definition →" in a 2009 book. The blog was never a brand asset. It was the top of a funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → that ended in a free tool: Website Grader in 2007, then the free CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition → tier from 2014, with gated templates and calculators in between. Every post carried an offer, and the offer matched the post rather than the company: a piece on email subject lines led to a subject line swipe file, not a demo request.

From 2017 the architecture became explicit. HubSpot published its pillar-and-cluster model, built on the clustering method the frameworks lesson sets out, and rebuilt the blog around it: one deep page per broad subject, dozens of supporting posts linking up to it, and the internal link graph doing the work of telling Google which pages mattered. The second mechanic is less copied and mattered more. In 2015 Pamela Vaughan's analysis of HubSpot's own blog found that around 76 percent of monthly views and 92 percent of monthly leads came from posts published in earlier months. HubSpot responded by staffing a permanent updating programme rather than a publishing programme. Old posts were rewritten, re-optimised and republished on a cadence. That is why the engine compounded: the library was treated as inventory with a shelf life, not as an archive.

Sub-concept 1: the drift into traffic that does not buy

The same machine that ranked "email marketing" also ranked how to type the shrug emoji, how to write a resignation letter, and lists of famous quotes. Those posts pulled enormous volume. None of them were read by someone shopping for marketing software.

Two things happen when a blog does this at scale. The first is obvious: cost per lead across the property looks fine in aggregate while a growing share of production goes to visitors with no purchase behind the query. The second is slower. The subject-level standing the frameworks lesson describes is built from what a site consistently covers, and a blog that covers resignation letters and CRM pricing with equal enthusiasm sends Google a blurrier signal about what it is for. HubSpot got away with it for years because domain strength carried the pages. When Google tightened, the generic pages had nothing else holding them up.

There is a fair counter-argument, and HubSpot's team made it internally for a long time: broad posts earn links, and links lift the commercial pages. True, up to a point. The problem is that the equity accrues to URLs you later want to retire, which turns pruning into a link-salvage exercise instead of a delete.

Sub-concept 2: the decay, in order

September 2023: Google's helpful content update. March 2024: a core update that folded those signals into the main ranking system and explicitly targeted scaled content produced for search rather than readers. May 2024: AI Overviews on general release in the US, answering exactly the definitional questions that a large part of the HubSpot library was built to catch.

The decline was not uniform, and the shape of it is the useful part. Informational pages, the how-to and what-is layer, fell hardest. Pages sitting close to a purchase decision, pricing, comparisons, integrations, held up far better. A blog that had been optimised for reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → lost the reach and kept most of the commercial surface. That is also why revenue did not follow traffic down: the traffic being removed had a per-visit value close to zero.

Sub-concept 3: the rebuild

Four moves, none of them fast. Pruning and consolidation: retire or merge posts that no longer earn their maintenance cost, redirect them into the surviving cluster page on the same subject, and accept a short-term traffic loss for a cleaner index. Re-anchoring content to the product, so that a post on lead scoring shows lead scoring being done rather than defining it. Rebuilding the commercial layer, including HubSpot's own comparison pages against Salesforce, where the query already contains a buying decision. And buying distribution that Google does not control: HubSpot acquired The Hustle newsletter in 2021 and built a podcast network the same year, which looks like a media hobby until an algorithm update takes 70 percent of your sessions.

Sub-concept 4: measuring an engine that is shrinking on purpose

Run the arithmetic before you panic. If sessions drop 70 percent and signups from content drop 8 percent, the 62 points of difference were never demand, and the honest read is that you were overpaying to serve them. If signups drop in line with sessions, you have removed the wrong pages and the redirect mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → is the first place to look.

The metrics that survive this are per-cluster: signups and pipeline by topic cluster, not by post; branded search volume, which is the closest proxy for whether the content is still creating future buyers; and whether your pages are being cited in AI answers for the queries you care about. Cohort reporting matters more than monthly reporting, because a cluster published in Q1 does most of its work in Q3 and Q4.

Real-world cases

Case 1: the cluster, in practice

Take the email marketing cluster. One long pillar page covering the subject end to end, supporting posts on segmentationsegmentationDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition →, subject lines, deliverability and automation, each linking up to the pillar with descriptive anchor text, and the pillar linking back down. HubSpot's own testing on this model found rankings for the pillar improved as the number of internal links pointing at it grew. The maintenance rule matters as much as the build: each supporting post carries an owner and a refresh date.

Case 2: the free tool at the end of the cluster

The conversion end of every HubSpot cluster is a tool or template, not a contact form. Website Grader in 2007 gave the blog something to send readers to that produced a result in thirty seconds and an email address as the price. The free CRM in 2014 turned that into a product entry point. The design constraint is unglamorous: the offer has to be worth the email on its own, without the reader intending to buy anything.

Case 3: the Salesforce comparison layer

"HubSpot vs Salesforce" is searched by people mid-evaluation, and both vendors publish their own answer to it. HubSpot maintains comparison and migration content aimed at that query rather than ceding it to review sites and affiliates. These pages are a fraction of a percent of the library's traffic and a disproportionate share of its influenced pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition →, which is why they were largely untouched by the 2024 decline.

CMO action items

• Pull your top 50 organic landing pages and mark each one with the signups or opportunities it produced last quarter. Expect a long tail of high-traffic pages with zero. That list is your pruning candidate set, not your success story
• Before deleting anything on that list, export referring domains per URL in Ahrefs or an equivalent and write the redirect map by hand, page to nearest surviving cluster page. Redirecting a retired post to the homepage is treated as a soft 404 and throws the link away
• Put a named owner and a refresh date on every pillar. If nobody is accountable for updating a cluster, assume it needs a substantive rewrite within about a year and budget the hours now rather than discovering the decay in a quarterly review

Common mistakes that kill results

Mistake 1: pruning as a delete project

Content pruning is a link-salvage exercise wearing a tidying-up costume. Teams that mass-delete underperformers lose the internal link structure holding the pillars up, then watch the commercial pages slip two months later and blame the algorithm. Consolidate into fewer, stronger pages, redirect precisely, and re-crawl to confirm the internal links still resolve.

Mistake 2: paying the team on sessions

A team measured on organic sessions will always find the cheap million, and the cheap million is the shrug emoji. This is an incentive design problem, not a talent problem. Change the target to signups or pipeline by cluster and the same writers stop pitching listicles within a month.

Mistake 3: treating technical debt as someone else's problem

A pillar page on a slow, badly linked site does not rank, and no volume of new posts fixes it. The CMO owns the organic outcome, so the CMO owns the argument with engineering about site speed, crawlability and Core Web Vitals. If nobody has looked at the technical health report in six months, run it before commissioning another quarter of content.