Open tracking is a liability dressed up as a metric

A founder asked a fair question on a call last week: if we switch off open tracking, how do we know anyone is reading the emails? It is the right instinct pointed at the wrong problem. The honest answer is that he had not known for several years, because the number on his dashboard stopped meaning what he thought it meant somewhere around 2021, and nobody told him.

The situation that produced the question is common enough that I have now seen it four times this year. A company’s domains and mailboxes sit with one provider. The outbound sequencer is a different vendor. To attribute an open or a click, the sequencer needs a CNAME record on the sending domain so it can rewrite every link through its own tracking host and serve a one-pixel image from it. The domain provider refused to add that record, because the sequencer’s tracking host carried a reputation problem it did not want attached to its own infrastructure. Campaigns stalled. Sends fell to single digits per mailbox per day, and most of those landed in spam.

The instinct in the room was to fix the tracking: buy fresh domains from a provider that permits the CNAME, warm them for three weeks, migrate everything. That is six weeks of no outbound in order to protect a metric. I would not do it.

What open tracking actually costs you

Open and click tracking is not a passive read on your campaign. It is an active rewrite of the mail you send. Every link becomes a redirect through a host you do not own, and an invisible image is loaded from that same host on render. Two things follow from that, and both of them are load-bearing.

The first is reputation coupling. The tracking domain is shared infrastructure. Its standing with the mailbox providers is the aggregate of everyone else sending through it, and you inherit that standing on every message. This is exactly why a domain host will refuse the CNAME: it is being asked to lend its own reputation to a third party’s.

The second is fragility. A security fix at a major marketing platform this January expired every tracked link generated on or before the 21st. Forrester’s read on the incident, using Validity data, put the deliverability drop at 25% across the four days around it, and the replacement encrypted links then broke DKIM signatures in some Microsoft environments through line wrapping. That is the shape of the risk. Your tracking layer is production infrastructure with its own vendors, its own outages, and its own failure modes, sitting directly in the path of every email you send.

The number you are protecting is mostly machines

Set the deliverability cost aside for a second and look at what you are buying. An open is recorded when the pixel loads. It is not recorded when a human reads.

Since Apple’s Mail Privacy Protection shipped, Apple Mail prefetches remote images for its users whether or not the message is ever opened. Corporate security gateways do the same thing, at scale, on the way in: link scanners follow every URL in the message to check it, which registers as a click nobody made. Image proxies cache the pixel once and never call home again, so a genuine second read is invisible. The result is a metric that is inflated by protective software, deflated by privacy caching, and different for every recipient domain in your list. It does not compare across segments and it does not trend honestly over time.

So the number is not a proxy for attention. It is a proxy for what mix of mail clients happen to be in your target account list this month. Optimizing subject lines against it is optimizing against your recipients’ IT policy.

Keep, edit, delete

I use the same three-way split on tracking that I use on CRM fields and pipeline stages, because the decision is the same decision.

Delete: open rate. It is corrupted at the source, it costs you the CNAME argument with your domain host, and no commercial decision downstream of it survives contact with the noise. Nothing in your forecast changes when you stop looking at it.

Edit: click tracking. The signal is real and worth keeping, but you do not need the sequencer’s redirect host to get it. Tag the destination instead. A UTM on the link means the click is recorded by your own analytics and your own CRM on arrival, with no rewrite, no borrowed reputation, and no expiring links.

Keep: reply rate, meetings booked, and per-mailbox send health. These three are measured on infrastructure you control, they are countable without a pixel, and every one of them is something a revenue leader can act on this week.

Where to measure outbound instead

Outbound gets measured at the destination, not inside the envelope. That reframing solves the problem the CNAME created and it produces a better dashboard than the one you lost.

The destination-side chain is: UTM on the link, session captured on landing, session stitched to the contact record on form fill or signup, contact associated to a deal. Every step there runs on your own domain and your own CRM, whether that is HubSpot or Salesforce. Nothing in it depends on a sequencer’s tracking host staying in good standing. When a campaign produces trials, you can name which campaign, which sequence step, and which segment, which is more than an open rate has ever told anyone.

Forrester’s August read on the state of B2B marketing foundations is that adoption is racing ahead of the plumbing: 88% of B2B marketing organizations have taken up AI tools, while poor data quality and accessibility sit at the top of the challenge list. Outbound instrumentation is that gap in miniature. Teams are sending more mail than ever through measurement wiring that never worked properly.

The volume side matters too, and it is getting harsher. Jason Lemkin’s argument on the AI SDR wave is that recipients block machine-sent outreach far more readily than human outreach, because blocking software carries no social cost. Blocks and complaints feed the reputation signal that decides whether your next campaign is delivered at all. Which puts sender reputation, not open rate, at the centre of the scoreboard.

What this looked like in practice

We recently worked with a Series A data-infrastructure company in EMEA whose campaigns were effectively dead: domains flagged in the sequencer, warm-up blocked, a handful of sends per mailbox per day, most of it filtered. The team’s plan was a full migration to a platform that would sell them domains and sending in one place.

We shipped the cheap version first. We duplicated the live campaigns with tracking disabled, which removed the CNAME requirement and unblocked sending the same afternoon. Marketing tagged every destination link with UTMs so the trial signups the campaigns were actually built to produce stayed measurable. The consolidation onto a single multichannel platform is still on the table, and it is probably the right end state, but it now runs on a three-week warm-up in the background instead of as an emergency with the pipeline switched off.

Nobody has asked about the open rate since.

The playbook

1. Clone the campaign, do not edit it. Duplicate the affected campaigns with open and click tracking switched off. Most sequencers will not let you change tracking on a live campaign.
2. Confirm the CNAME error clears and the mailboxes rejoin the sequence. This is same-day work, not a project.
3. Put a UTM on every destination link, with the campaign and the sequence step encoded in it, before the first send goes out.
4. Verify the chain end to end. Click a real link, confirm the session lands, confirm the contact record picks up the source, confirm the deal inherits it. Do not assume any of the four steps.
5. Rebuild the dashboard on reply rate, meetings booked, trials started, and per-domain send health. Delete the open-rate tile so nobody reports against it out of habit.
6. Give the sending estate an owner. Warm-up, rotation, and reputation monitoring are infrastructure, not a campaign task.

Run the first three steps this week. If the sending estate underneath is the actual problem, our deliverability infrastructure work is where that gets rebuilt properly, and the measurement chain behind it is standard revenue operations scope. If you want a second opinion on which of the two you are dealing with, book a call.

Sources

• Forrester. “When Fixing Security Vulnerabilities Breaks Your Customer Email Program.” Shar VanBoskirk and Jess Burn, March 2026. forrester.com
• Forrester. “B2B Marketing Is Moving Faster Than Its Foundations Can Handle.” Mark Ogne, August 2026. forrester.com
• SaaStr. “Could Block Be the Death of the AI SDR?” Jason Lemkin, May 2026. saastr.com