Product Launch: The Lazy River to Product Adoption

The problem: paying for the busiest day, every day

Upsun customers choose in advance how much computing capacity their applications get, like renting an apartment big enough for the two weekends a year the in-laws visit. Usage data surfaced two costly patterns, and customer conversations confirmed both:

Over-allocators paid for headroom that sat idle almost all the time.

Ceiling-hitters sized for a normal day, then watched their apps slow down when traffic spiked.

Opposite symptoms, same root cause: customers were being asked to predict demand, and demand doesn't take appointments.

What shipped

Autoscaling removed the prediction. The platform watches demand and adjusts capacity automatically, adding it as traffic climbs and releasing it as traffic falls, with billing that follows actual use. The promise fit in one line: pay for what you use, and stop babysitting the dials.

Shipping the feature was half the job. A feature customers never find might as well not exist, so the launch plan mattered as much as the release.

How customers discovered Autoscaling

The launch ran the full playbook: announcement email, blog post, changelog, and documentation. Those channels did their job, building awareness with everyone at once.

But awareness and timing are different things. The right time for Autoscaling was different for every customer: a third manual resize of the month, or the traffic spike that just slowed their app. No email calendar can hit a moment like that. Just-in-time discovery was key. To accomplish this, the product monitored for signals that Autoscaling would be practical for the customer, then informed them right where they were working - just in time. In that moment, Autoscaling isn't an announcement to skim. It's the call-to-action to solve a the immediate need.

The Lazy River approach to adoption

I adhere to the Lazy River adoption philosophy. Ease customers down the stream, with no rapids and no eddies. Avoid dumping every capability on a customer at once. Each nudge arrives at the moment it's relevant and moves the customer exactly one step forward.

For Autoscaling, that worked like this:

Qualify. The product's usage data told us when Autoscaling had become practical for a customer: continual over-allocation, or repeated run-ins with resource ceilings.

Test. With Pendo (product analytics and engagement tooling), we conditionally injected prompts the moment those signals appeared, only for those customers and only in context, and measured whether informed customers enabled Autoscaling at a higher rate than the unprompted baseline.

Productize. They did, decisively. Once the experiment proved the channel, we retired the third-party prompts and built the guidance natively into the product, where it could run permanently and feel like part of the platform rather than a pop-up.

Orchestrating the launch end to end

The in-product channel was the centerpiece. Before launch, I defined the adoption metric (the percentage of eligible organizations enabling Autoscaling within 90 days) and set the target we'd be judged against. Around it:

Lifecycle marketing got a brief covering who to reach, what to say, and why: the segments, the pain, and the "pay for what you use" message, in that order.

Launch content across the blog, documentation, and changelog carried the same positioning, written so a reader didn't need to be the team's infrastructure expert to understand what they'd stop overpaying for.

Sales and support were enabled to explain the feature and handle the natural objection: "will this run up my bill?" (The honest answer was the message: it usually lowers it.)

Measurement was built with analytics before launch, tracking the full path from prompt seen to feature enabled to still active.

From adoption to lasting engagement

Adoption beat its target. 38% of eligible organizations enabled Autoscaling within 90 days, against a 25% target. Prompted customers enabled at roughly four times the unprompted rate.

It stuck. About 85% of adopters still had Autoscaling active 90 days after enabling it. Adoption only counted if it stayed on.

Support got quieter. Tickets about resource sizing fell roughly 30% in the following quarter, as customers stopped filing tickets about a decision they no longer had to make.

The platform ran leaner. Idle, over-provisioned capacity shrank across adopters. Fewer wasted server-hours meant a real efficiency and emissions win: good for the customer's bill, good for the platform's footprint.