Every few weeks, someone running Adobe asks us a version of the same question: is there a way to spend less on this without pulling it out?

We’ve identified three independent drivers, and they're worth separating, because they're set by different decisions and reduced in different ways.

• Collection: Every event sent, and every field attached to it.
• Usage: How many customer profiles you have, multiplied by how much detail is attached to each one.
• Distribution: What it costs to move data to the systems that act on it.

‍

‍

Nothing here involves replacing What changes is how many events go in, how many profiles get created, and what it costs to move data back out.

1. Why sending less is harder than it sounds

With the Adobe Web SDK, the first network request off the browser goes to Adobe's own Edge Network — the edgeDomain setting defaults to edge.adobedc.net. Adobe receives the data, then distributes it from there.

Event Forwarding is often assumed to change this. Adobe's documentation is clear that it doesn't: it "requires event data to already be sent through Experience Platform Edge Network in order to forward it to destinations."

So the filtering happens downstream of the counting. You get real control over where data goes once Adobe has it — but not over how much Adobe receives. That's the reason the obvious lever has been unavailable.

"We're already server-side"

One follow-up question resolves this: when a visitor loads your homepage, what domain does the first network request go to? If the answer is edge.adobedc.net, that's server-side processing, not server-side collection.

2. The three areas where you can save

2.1 Collection 

Send only what Adobe's use cases need

The Web SDK's default is broad collection. The alternative is a deliberate list: every event and every parameter forwarded because a report, a segment, an activity or an audience depends on it.

This compounds across all three meters at once. Fewer events means fewer server calls. Fewer parameters means lower profile richness. Fewer incidental interactions means fewer clicks and downloads billed at the same rate as a page view.

Our own modelling puts this at 20–30% of event volume and storage. That's MetaRouter's estimate rather than an audited industry figure, and it depends heavily on how much an implementation has accumulated since it was built. 

One caveat: this assumes what goes to Adobe can be separated from everything else. If a single stream feeds Adobe and four other tools, you're splitting it rather than shrinking it — a different job.

Remove bots before they reach Adobe

Automated traffic isn't a rounding error any more: The Thales/Imperva 2026 Bad Bot Report found automated traffic made up 53% of all internet activity in 2025 — 40% bad bots, 13% benign automation. That figure is internet-wide and includes API and attack traffic, so it isn't a retail page-event mix. But it sets the scale. On retail sites we typically see bots at around a third of page traffic. Note: This is our own observation, not an audited figure.

Every one of those events costs a server call, costs storage, and then sits inside the reporting you're paying for — inflating sessions, flattering conversion rates, and feeding the audiences built on top of them.

2.2 Usage

Stop paying for duplicate profiles

Adobe prices Real-Time CDP on two things multiplied together: how many customer profiles you have, and how much detail each one carries. 

Safari deletes tracking cookies after seven days. So a customer who comes back a month later arrives looking like a stranger, and gets counted as somebody new. The same person, counted again and again. 

The result is more profiles sitting in Adobe than the business has customers. The fix is to set the ID from your own server, where it survives, rather than from the browser, where Safari clears it. 

That fixes the first half. The second; how much detail each profile carries; is fixed further upstream, by sending only the fields something actually uses. 

When this doesn't apply: if you already run software that merges duplicate records after the fact, this is probably handled. Compare the two numbers before assuming there's money here.

Don't store the same data twice 

Most enterprises running Adobe alongside a warehouse are keeping much the same data in both, and buying more Adobe storage as it grows.

Where the warehouse is already what the business queries, Adobe only needs what Adobe's own tools use. The rest lives once, at storage prices rather than platform prices.

2.3 Distribution

Free exports, expensive forwarding distribution 

Flat file exports into your own warehouse are often free, but the charge lands on the thing you would act on — forwarding events to other platforms as they happen. Ad platforms, marketing tools, anywhere a signal needs to arrive while it still matters.

So most teams have made a choice: take the free data and accept it's a day old, or pay per event for the version that arrives in time to be useful. It's also why the warehouse everyone relies on is never current, and why personalisation runs on yesterday.

And because forwarding is metered, cost ends up capping volume, and teams end up sending a fraction of what they could to the channels that perform best

When collection happens in your own cloud, those platforms connect there instead, and that means the charge doesn't shrink, it actually stops existing.

So what actually changes?

• Filtering redundant fields and duplicate events cuts volume without touching sessions, unique visitors or conversion rate. Removing bots is the exception: sessions fall and conversion rate rises, because you stop counting traffic that was never a customer. 
• Both collection paths run in parallel on the same pages until the numbers line up and everyone's satisfied.
• Route the full unfiltered stream into your own warehouse. Adobe gets the curated subset; you keep everything, permanently, in case you ever need it back.

3. With MetaRouter your Adobe tools get better too

Adobe's tools are only as good as the context that reaches them — who this person is, what they've already done, and whether it's arriving in time to matter. All three are decided upstream, before Adobe sees anything.

If one person arrives as seven separate profiles, Adobe has seven strangers instead of one returning customer. It'll personalise to each of them as a first visit, because that's what the data says. But you can fix the identity before Adobe sees it, and there'll be one profile with a history behind it. 

The same applies to automated traffic. An audience built partly on visitors who were never human is smaller and less accurate than it looks, and every model trained on it inherits the error.

Timing also matters. Data that arrives overnight can explain what happened. Data that arrives in the moment can change what happens next — which is what Target, personalisation and activation were bought for in the first place.

So the argument isn't that you're spending too much on Adobe. It's that Adobe is working with worse inputs than it should be, and fixing that happens to cost less.

4. 5 things to check on your own set up

☐ Does your first-party collection request go to a CNAME or A-RECORD endpoint?

☐ How many profiles are in Adobe, versus how many customers do you actually have?

☐ What share of your server calls aren't page views?

☐ Does your bot filtering happen before or after Adobe counts the call?

☐ How does data get out of Adobe — streaming, or batch?
If it's batch: was that an architecture decision, or a cost one?

Talk to us

Interested in what this would look like on your setup? Talk to our team.

‍

MetaRouter is the first-mile Customer Context Infrastructure that captures identity, behavior, and consent — in real time, inside the enterprise's own cloud, before the moment is lost. Trusted by global B2C enterprises including The Home Depot, Costco, Dicks Sporting Goods, and the world's largest Retail Media Networks. For more information visit: www.metarouter.io. 

‍