“Web analytics has been rebuilt three times, and each rebuild followed a shift in how software was sold: hit counters for on-premise sales, then funnel reports once growth went marketing-led, then event analytics for the mobile app economy. This piece traces those three step changes and argues that the self-serve, subscription era needs a fourth.”

Every time the tech industry goes through a step change, web analytics follows.

Tracing the changes that produced the product analytics category tells you a lot about where newer trends like product-led growth and recurring revenue models will take user analytics next.

This post is part one of a series on Priority Analytics, exploring how emerging trends in tech are creating the conditions for a new generation of the product analytics market. Part two is The Guessing Gatekeeper.

Small beginnings: the hit counter

The first web analytics tool, Analog, launched in 1995. It analysed server logs to work out which pages a user visited on a website. Analog started the era of hit counters and made page views the first north star metric in web analytics.

Hit Counter UIs graphic from the Amplitude blog

A decade of web analytics improvement followed. Clicktale, Omniture and Urchin, which Google acquired in 2005 and turned into Google Analytics, improved the usability and functionality of log-file analytics and spread it to more developers. Page views held the top spot throughout as the metric that mattered.

Step change 1: web analytics moves to the cloud

The first analytics product to make funnel reports accessible was KISSmetrics.

In late 2009, Hiten Shah and the KISSmetrics founding team had just abandoned their second failed startup idea. Looking for a better one, they decided to survey online marketers. In Shah's own 2019 account of that period, the survey results went on to inspire the feature at the centre of today's product analytics tools.

His summary of what changed is that they stopped writing code and went out to talk to customers, including people using their largest competitor, Google Analytics. What those users said was that they needed data to improve marketing campaign performance once people arrived on their websites, and that they would pay for analytics that improved conversion rates.

That discovery marked the beginning of the second generation of web analytics and set off the marketing-led growth era of the 2010s.

Why inbound marketing took off when it did

People overestimate HubSpot's role in the rise of marketing-led growth, largely because it built the brand most associated with inbound marketing. Gabriel Marguglio makes the point in his 2019 history of inbound marketing: the phrase was coined by HubSpot co-founder Brian Halligan in 2005, appeared online in small quantities from 2007, and only started growing properly around 2012.

Why 2012? Cloud computing.

In the late 2000s, cloud computing rapidly overtook on-premise as the preferred infrastructure for software, especially among new startups.

*Graphic from* *State of Cloud 2020* *by Bessemer Venture Partners*

Building on the cloud changed what a company's website was for.

On-premise software was expensive to implement and carried a high price tag. A high price tag meant the buyer was usually the CEO or CIO, and the purchase required serious due diligence beforehand. The objective was to get a knowledgeable salesperson on the phone and have every question answered upfront. That model needed more people arriving at your website each month and a high percentage of them calling or emailing. Page views was the only metric an on-premise company needed.

Cloud software had scale built in. Infrastructure moved from a fixed cost per customer to a variable cost with economies of scale spread across the whole customer base.

Cost of goods sold fell, which let cloud companies undercut on-premise competitors on price. The hourly rate for a salesperson didn't fall. No company could afford to have salespeople answering the same volume of questions and educating prospects when the customer was spending three times less than before. That gap is what drove the move from sales-led to marketing-led growth.

Cloud companies had to change their websites from a glorified contact page into something carrying blogs, FAQs and infographics, to attract, educate and qualify leads. To know which content was working, marketers needed to track a lead through each step from visitor to qualified lead to conversion.

KISSmetrics took web analytics out of the developer tools category and opened it up to digital marketers.

Building a conversion funnel in 2008 or 2009 was genuinely hard. Search for "funnel" and "google analytics" across that period and the top organic result is an article about funnel problems in Google Analytics. With a clear complexity problem in front of them, Shah's team redesigned the process with digital marketers in mind. By his account, funnel reports before KISSmetrics took too long to create and understand, and the version they built became the industry standard that other analytics tools copied.

The team shipped several strong features in that first year, including profile-focused reporting and a much better debugger. The most important factor behind their success was timing.

KISSmetrics arrived with the right answer to the cloud step change just as the first signs of it appeared. Log-file web analytics and page views were replaced in the early 2010s by event analytics, and funnel reports became the most important source of metrics. That held for the following five years.

Step change 2: web analytics meets the mobile consumer

The KISSmetrics lead in web analytics was short-lived. By 2013 Mixpanel had taken the top spot and KISSmetrics began fading into an alternative nobody chose.

What let Mixpanel overtake them? Mobile analytics.

While marketing-led growth was building steam in B2B, smartphones made the leap from corporate to consumer in the late 2000s. Apple launching the iPhone and iPod Touch alongside the App Store gave a fast-growing population of consumer smartphone users more choice than they'd ever had.

In an increasingly competitive app market, companies fought for consumer attention any way they could. The most successful tactic was moving the point of purchase deeper into the user journey. Instead of an upfront payment, the App Store economy came to depend on in-app purchases. Just over half of new App Store apps were free in early 2012. By mid-2016 the figure was above 90%.

Moving conversion deeper into the journey changed web analytics in several ways at once. Tools now had to track users and events across multiple sources, from social media or a landing page into actual product usage. They needed to work on desktop, mobile or both, and to handle increasingly complex paths, where conversion related more to key usage milestones than to a predictable sequence of steps.

Each alternative to KISSmetrics picked one of those problems. Mixpanel and Ampl itude focused on mobile-first analytics. Optimizely productised A/B testing for better downstream impact analysis. Heap's no-code approach made variable user journeys far easier to analyse retroactively.

Competitor by competitor, the web analytics category KISSmetrics had redefined in 2009 was redefined again. As data collection moved from landing pages into the deepest parts of products and apps, web analytics earned a rebrand to product analytics.

Step change 3: the empowered end-user

The consumers downloading free mobile apps at the weekend were often the same people working a nine-to-five during the week.

Through the late 2010s a behavioural gap opened up. By day these people dealt with poor software bought by their boss on annual lock-in contracts. By night they downloaded apps for free and only reached for a credit card when they hit a usage limit or an incentive worth upgrading for.

The mobile app market had done two things: conditioned people to try before buying, and made them comfortable searching for and evaluating digital products on their own. The easiest place to see it is ecommerce, where a large majority of US adults aged 18 to 34 held an active Amazon Prime membership by mid-2020.

Consumer SaaS is what caused it, though. Paying for software outside work on a subscription became normal, particularly among people with no buying power at the companies they work for.

Graphic from Consumer Subscription Software Insights, September 2020 by GP.Bullhound

Take an example. Sam the assistant brand manager in 2011 was never in a position to consider paying for a new tool at work. Jamie the junior product manager in 2021 holds premium subscriptions to Netflix, Amazon Prime, Strava, Spotify, Calm, Grammarly, Canva and Robinhood. Does Jamie know how to evaluate whether a paid tool would improve their work or fix a frustrating problem? Obviously.

So in their personal lives, people now have three advantages they didn't have before:

The knowledge and skills to find and pay for digital products that fix their biggest pains.
The expectation that they can try a product before buying it.
An understanding that subscriptions reduce the upfront burden of paying for expensive software.

The end-user who expects self-serve onboarding and a subscription by default is the source of the most recent step change in analytics.

Recurring revenue requires recurring impact

Moving from a single sale to recurring revenue means rethinking the sales funnel entirely.

A funnel is a set of steps leading to a closing event, like a prospect signing a contract. It assumes leads drop out at each step and a subset arrive at the final one.

*^ The traditional sales funnel. Source:* *Marketing Insider Group*

For a product team on recurring revenue, the one-off payment is not the objective. Their aim is to retain that customer every time the subscription renews.

The V shape stops being relevant. SaaS companies are compared on net revenue retention rate, not just total revenue. NRR measures the money you collected from a cohort of new subscribers in one month against what that same cohort pays a year later.

A worked example:

100 customers subscribe in January 2021 at $10 a month, giving you $1,000 in monthly recurring revenue.
By January 2022, 90 remain, so 10% churned. Those 90 now pay an average of $15 a month, because some added seats.
That cohort is now worth $1,350 a month, which is an NRR of 135%.

For context, Slack reported an NRR of 125% in Q2 2021.

If the focus shifts from the initial purchase to growing account value over time to compensate for churned revenue, the funnel starts looking like a bowtie.

**Sales Bowtie** graphic by Jacco van der Kooij, from Winning By Design

Bottom-up growth is emotional

Most product-led startups acquire customers bottom-up.

Picture a company hierarchy as a triangle. A top-down approach targets the executives at the top, who push the product down through the company using their authority. A bottom-up approach targets the operational layers at the bottom. Those employees share the tool with colleagues either for status, being seen as the early adopter of something good, or for function, needing a teammate to collaborate with.

Choosing between them is a fundamental strategic decision. It goes on to influence team objectives and the design of each function in the company, which is the subject of part two.

When a team goes bottom-up, the person they're acquiring is the person who will use the product. Unlike a CEO, the end-user isn't comparing products on how they might improve company profits. They want to fix an annoying problem they've been dealing with, which is why ranking customer problems predicts adoption better than any feature list.

Those two drivers of adoption are rational impact, meaning bottom-line profit, and emotional impact, meaning solving a burning pain point.

A bottom-up strategy usually acquires users through emotional impact, and the growth flywheel needs both.

Impact and growth compound together. The end-user is most likely to take actions that grow the value of the account each time they get meaningful impact from the product.

The five growth levers inside an existing account

These sit on a sliding scale from emotional to rational impact. The more emotional the lever, the less influence the traditional decision-maker has over how the account grows.

1. Status-driven internal referral. The most emotional lever. Picture an end-user close to tears because a manual task eats six hours a week, who then finds a tool that does it in six minutes. If they know ten other people with the same problem, they will tell every one of them. The referral costs nothing, needs no sign-off, and earns them a reputation as the person who knows the good tools.

2. Usage-based expansion. Two kinds exist. Infrastructure products like AWS and Stripe earn more the more you use them, which is a large growth lever but not an end-user one. The version that matters here is organic discovery of additional use cases. An account manager starts using Calendly for inbound sales calls, then shares the link with their boss for a quarterly review, then uses it for a marketing catch-up, then for scheduling the Friday social. It worked once, so it gets used everywhere.

3. Unlocking feature requirements. That same account manager now runs several use cases through Calendly. The free version allows one event type, and a blanket 30-minute block doesn't suit a 15-minute marketing catch-up or a 90-minute review. Multiple event types means upgrading. This is where the lever moves from emotional impact, meaning it makes me feel or look good, to rational impact, meaning it saves me time or money.

4. Functionally-driven internal referral. Also based on sharing with colleagues, but driven by necessity instead of social credit. On Miro you invite colleagues because you need their ideas on the board. On Segment you invite a developer to help with integration. Those additions may not create revenue immediately, and they restart the whole bowtie with a new batch of users.

5. Enterprise sales. Some companies use bottom-up to attract multiple users inside a team, then switch to a top-down outbound approach and pitch the decision-maker responsible for that group. It works best where formalising an enterprise contract unlocks additional rational impact, like features that improve how the group collaborates.

Every one of those levers depends on knowing what the end-user actually needs, at both ends of the bowtie.

Introducing priority analytics

End-users have buying power they didn't have before, they expect to self-serve, and converting them on a subscription costs almost nothing. None of that is going to reverse. Yet product analytics works almost exactly as it did ten years ago, and the web analytics assumptions underneath it are older still.

Like the move from log-file to event analytics in the late 2000s, the core data we collect needs rethinking. If the customer journey is now a self-serve experience powered by a segmented understanding of end-user needs, analytics should be focused on discovering, segmenting and measuring those needs.

Priority analytics is a quantitative form of user data, like product analytics, except it focuses on user priorities instead of user behaviour. Product analytics tells you what people are doing. Priority analytics tells you what they are thinking, and why.

That data serves every function and every step of the bowtie. Understanding highest-priority needs informs the positioning used to attract users, the messaging sales teams use, the segmentation of self-serve onboarding, and the rationale for picking referral methods. It's the same argument behind the Discovery Sandwich, which layers qualitative work around a quantitative core.

Today, identifying the most important user needs is almost entirely qualitative. The standard approach is a handful of user interviews and some messy docs full of customer quotes. Building a serious startup needs a quantitative source of data on end-user needs that is as easy and reliable to set up as a funnel report was on KISSmetrics in 2009.

That's the vision at OpinionX: helping companies discover and rank their users' priorities in minutes. Teams at Disney, Google, LinkedIn and Shopify use OpinionX to understand what their users care about most.

Methods like customer problem stack ranking, pairwise comparison and needs-based segmentation are how that data gets collected.

Frequently asked questions

What was the first web analytics tool?

Analog, launched in 1995. It read server logs to determine which pages a visitor had seen, and it started the hit counter era that made page views the first metric anyone tracked.

How did web analytics become product analytics?

Through two shifts. Cloud software moved companies from sales-led to marketing-led growth, which created demand for funnel reports and event tracking instead of page views. Then the mobile app economy moved the point of purchase deep into the user journey, so tools had to track users across sources, devices and non-linear paths. Once collection reached inside the product itself, the category name followed.

What is net revenue retention?

A measure of what a cohort of customers is worth a year after they subscribed, against what they were worth at the start. 100 customers paying $10 a month is $1,000. If 90 remain a year later paying $15 each, that is $1,350, or an NRR of 135%.

What is the sales bowtie?

A replacement for the funnel under recurring revenue. The funnel ends at the sale. The bowtie treats the sale as the midpoint and gives equal weight to everything that happens afterwards: onboarding, impact, expansion and renewal.

What is priority analytics?

Quantitative data on what users consider most important, as opposed to product analytics, which measures what they do. Where product analytics answers what happened, priority analytics answers why, and it is designed to be collected at the same scale and reliability as behavioural data.

Over 42,000 researchers and product people get one breakdown like this each week in The Full-Stack Researcher.

OpinionX helps product teams discover and rank what their users care about most. Every question type and every analysis feature is unlocked on the free tier, capped at 25 participants per survey.

Part two of this series is The Guessing Gatekeeper.

Create a free survey 👉🏻

The History of Web Analytics and Future Predictions (1990s-2020s)