Why we cut email volume in half overnight and grew the channel anyway

Groupon VP of Revenue and CRM Marketing Ernesto Martín Rodés shares how Groupon cut managed-channel volume by 66% year-on-year — and grew the channel anyway. The lesson: precision beats volume, and respecting your customers’ attention is a growth strategy.

We were spamming our own customers. And we knew it.

A year ago, I asked ChatGPT what came to mind when it saw the word “spam.” The answer: Groupon and Nigerian Prince scams.

Uncomfortable. But fair.

For years, we built a customer experience that even those of us working here couldn’t tolerate. Inbox rules. Unsubscribes. Silence. We compensated for low relevance by sending more, which made relevance worse, which made us send even more. A loop we understood perfectly and changed too slowly.

The reasons were real: technical constraints, short-term revenue pressure, organisational inertia. But if I’m honest, it also came down to this — we weren’t capable enough yet, and we weren’t brave enough.

We should have done this years ago.

So I made a call.

Cut volume by 50% overnight, in all managed channels. No plan B.

No soft landing, no parallel test, no fallback campaign teed up if the numbers tanked. Two reasons.

First, we needed to prove the concept — that “precision beats volume” isn’t a slogan, it’s a P&L outcome you can only see when you actually stop hiding behind volume.

Second, the savings from cutting sends were the only way to free up the budget for the external tooling we needed to do this properly. You don’t argue for a Customer Data Platform on a slide deck. You argue for it by removing the thing it’s supposed to replace and letting the business feel the gap.

The internal objection was obvious: cutting sends cuts revenue. The counter was equally obvious, just less popular — if we keep attriting customers with a painful experience, the math never compounds. Short-term volume is borrowed from long-term retention. At some point you have to stop borrowing.

Precision beats volume. Every time.

The UK was the first proof point. We ran a statistically significant A/B test over 47 days. The group receiving smarter, fewer messages generated more gross profit per message than the control, with higher email and push click rates. Sending less worked.

The customer-side numbers backed it up. In April 2025, we rolled out a global cut to new-subscriber send volume — over 80% fewer messages in the first weeks. Unsubscribe rates dropped. The customers who did eventually unsubscribe stayed meaningfully longer before doing so (median tenure improved by roughly 50% in the post-launch cohort). Revenue held. Less spam wasn’t just kinder; it was profitable.

That gave us the budget and the conviction to rebuild the infrastructure underneath.

The infrastructure problem nobody talks about

You cannot fix relevance with better copy or smarter subject lines if the systems underneath are working against you.

Before this transformation, every new campaign, every new trigger, every test required an engineering ticket. Speed to market was measured in weeks. Our recommendation engine prioritised Groupon’s deal margin over customer relevance — hard-coded rules that served the business at the expense of the user. Data lived in silos. We never had a true unified view of who our customers were.

That’s why we invested in Bloomreach as our Customer Data Platform. Not because it was fashionable. Because we had hit a wall. True user-level orchestration — right customer, right channel, right moment, content that actually reflects their behaviour — was not possible with what we had.

The work itself was carried by people I trust deeply. Kate Usova led the CDP build and the Bloomreach migration alongside Matías Coumet. Celia Sainz rebuilt our lifecycle programme from scratch. Sonja Seechurn redesigned how we use our managed channels. Great support from all engineering, app and data teams as well.

The complexity of what they delivered was, frankly, absurd — 13 countries, different platforms, different apps, a parallel rollout of new app technology running in the background, and BAU work still shipping every week. None of that fits on a slide. None of this happens without them.

The numbers

Today, Bloomreach is live in all 13 of our markets. Lifecycle journeys that used to take quarters now take hours. Campaigns once gated by engineering are launched directly by CRM teams. First-time subscribers receive roughly 80% fewer messages than they used to — and now move through nurturing journeys with signal-based cadence instead of a generic blast schedule.

Email volume is down 66% year-on-year — and the channel is growing. Not flat. Growing.

The business case held. Lower send volumes and the operational simplicity of running on a single modern platform offset most of the investment. We’re tracking to positive ROI in year one.

In 2025, Groupon as a whole returned to growth in billings and revenue, for the first time in a decade. CRM is one piece of a much broader turnaround — but the foundation we’ve rebuilt is the platform every other piece of the strategy now leans on.

And we didn’t stop there

Once the platform was stable, we started layering automation on top. We’ve now automated more than 40 recurring workflows — pulling over 100 hours of work off the team’s calendar every week. And we have agentic AI running constant journey optimisation in the background: testing, refining, pruning, escalating when humans need to weigh in.

The point isn’t the AI itself. It’s what it lets us spend our human attention on: judgement, strategy, and the calls that still require a person in the room.

What this is really about

This isn’t a story about technology. The platform we chose is excellent, but the platform is not the point.

It’s about deciding that the short-term cost of doing the right thing is worth paying. Cutting volume hurts revenue in the quarter you do it. Rebuilding infrastructure pulls focus from other priorities. Holding the line when every metric is telling you to send more requires a kind of organisational nerve that is genuinely hard to maintain.

What made it possible was clarity of priority. We named this a Company Bet — which meant teams across CRM, Data, Product, Engineering and Finance aligned their time around work that compounds, instead of work that just fills the quarter.

What’s next

We’re still early. The next chapter: pricing that adapts to the individual, not the segment. Lifecycle orchestration at the user level. A loyalty programme our customers actually want to be part of, instead of one we have to push.

And we’re not done cutting. The plan is a further 20% reduction in volume each year for the next two years — while continuing to grow the channel. Lifecycle orchestration is what makes that math possible.

We won’t stop until every message we send is one our customers would have asked for.

The lesson, if there is one, is simple.

Respecting your customers’ attention is not a soft metric. It is a growth strategy.