New Newsletter Ad Campaign Not Delivering? Week One Explained

A new newsletter ad campaign that barely delivers in its first week is usually not broken. In most cases one of five things is happening, and you can check each of them. The ad server has no click history for your ad yet, so it ranks you on an estimate. Your share of trial impressions depends on how established the leading campaign is. Newsletters only produce impressions when issues go out and readers open them, so your start date may have missed the sends. Even daily pacing caps every day at the same allowance, which wastes the send-day burst and never rolls over. And a dozen eligibility rules quietly remove your campaign for particular readers before any ranking happens.

This post works through each one using the actual numbers from the MailAdx ad server: the thresholds, the formulas and the defaults. It ends with a symptom-by-symptom checklist and a rule of thumb for when to wait and when to change something.

Most writing about the "learning phase" is about Meta and Google, whose own help page on the learning phase is typical, where the platform hides the mechanics and the advice comes down to "do not touch it for a week". Email ad serving works differently, and for most of the reasons above, waiting does not help.

What "delivering" means for a newsletter ad

On a website, an ad slot exists whenever someone loads a page, so a campaign that loses one auction gets another chance a second later. A newsletter ad slot exists only when a reader opens an email. With open-time ad serving, the ad in the email is an image that is decided when the reader opens it, not when the email is sent. Each open is one request to the ad server, and each request is one chance to deliver.

That has two consequences for week one. First, supply arrives in bursts: thousands of opens in the hours after a send, then a long tail of late opens, then nothing until the next issue. Second, every reader is a known subscriber, identified by a hash of their email address, so rules about individual readers, such as frequency caps and fatigue, apply from the very first impression. The open-time serving explainer covers the mechanics. For this post, what matters is that a newsletter campaign competes for a lumpy, finite set of opens, one reader at a time.

Reason 1: the ad server does not know your click rate yet

If you buy on a CPM basis, this reason barely applies. A CPM campaign is ranked on its CPM, so a $12 CPM campaign outranks a $9 one from the first request.

If you buy with a click or lead objective, the ad server has to compare your bid with CPM bids. It does that by converting everything to an expected value per thousand impressions, the effective CPM or eCPM:

CPM campaign   eCPM = CPM bid
CPC campaign   eCPM = CPC bid × predicted CTR × 1000
CPL campaign   eCPM = CPL bid × predicted CTR × predicted CVR × 1000

A $1.50 CPC bid with a predicted click-through rate of 1% competes as a $15 eCPM. The same bid with a predicted 0.4% competes as $6. The bid is the same. Your rank depends on the prediction.

For a new campaign there is nothing to predict from, so the ad server starts from a prior: what similar ads have done before. In MailAdx the prior is looked up from the most specific match available, falling back step by step:

1. publisher category × creative category × device
2. publisher category × creative category
3. creative category
4. everything (global)
5. built-in default: 1% CTR, 5% conversion rate

Then, as your campaign collects real impressions and clicks, the prior and your observed results are blended. The prior counts as a fixed number of pseudo-impressions, 200 for the built-in default:

estimated CTR = (200 × prior CTR + your clicks) / (200 + your impressions)

Here is what that does to a $1.50 CPC campaign whose real click rate turns out to be 0.6%, starting from the 1% default:

Impressions  Clicks  Estimated CTR  Competes as
          0       0      1.00%          $15.00 eCPM
        100       1      1.00%          $15.00
        250       2      0.89%          $13.33
        500       3      0.71%          $10.71
      2,000      12      0.64%           $9.55   (blended; a trained model may replace it)

Two things follow from that table. A campaign whose real click rate is below the prior gets an easy start and then slides down the ranking as evidence arrives. To an advertiser watching the dashboard, it looks like delivery started well and then faded. A campaign whose real rate is above the prior has the opposite experience: a slow start, then improving delivery as the clicks come in.

At 500 impressions a campaign leaves cold start. From then on, where a trained model is available, its prediction replaces the blended estimate. The model takes into account the reader, the hour and the device, not just the campaign's average. But 500 impressions is the soonest that can happen, and for a campaign on a single small newsletter that can take several issues.

One more rule applies only to click and lead campaigns. If a campaign reaches 500 impressions with no clicks at all, its ranking score is halved, and it is flagged so someone looks at the creative or the targeting. Zero clicks in 500 opens is rarely bad luck. More often the link is broken, the image does not render, or the offer does not match the audience.

Reason 2: you only get a share of the trial impressions

If an ad server always showed the campaign with the highest estimate, a new campaign starting below the leader would never be shown, never collect evidence, and stay below the leader forever. Every ad server that learns has to set aside some impressions for trying alternatives. The standard name for this is the explore-exploit trade-off, and the simplest rule for handling it, epsilon-greedy, is set out in chapter 2 of Sutton and Barto's Reinforcement Learning: An Introduction.

MailAdx uses an epsilon-greedy rule. On each request, with probability ε (epsilon), the server explores: it picks a random eligible campaign other than the current leader. Otherwise it shows the leader. The value of ε depends on how much evidence the leading campaign already has:

Leader's impressions   ε (share of requests explored)
fewer than 500         20%
500 to 5,000           12%
5,000 to 50,000         8%
more than 50,000        3%
00:00–06:00 UTC        +4 points (capped at 25%)

The detail that surprises people is that ε is set by the leader's maturity, not yours. If you launch into a slot where an established campaign with 80,000 impressions is winning, only 3% of requests are explored, and that 3% is shared among every other eligible campaign. With four challengers, each one gets well under 1% of the slot's requests until its own results lift it into contention.

Exploration also skips any campaign the reader is fatigued on, and it never applies to guaranteed placements. A roadblock or sponsorship always wins its slot outright, so a new campaign competing with one gets nothing from exploration on those placements. If a publisher has sold a sponsorship for the issue you are targeting, that is the whole explanation.

The exploration share is deliberately small, because each explored impression is usually worth less than the leader's. It is the cost of finding out. But it means a challenger launching into a crowded, mature slot should expect a slow first week, however good the ad.

Reason 3: the newsletter has not sent yet

This one is not about algorithms at all, and it is the most common.

A weekly newsletter produces almost all of its opens in the day or two after each send. If your campaign starts on a Wednesday and the newsletter sends on Tuesdays, your first real day of supply is six days away. For those six days the only requests are late opens of last week's issue, and many readers have already seen whatever ad they were going to see in it.

A weekly newsletter, 20,000 opens per issue, sends Tuesday morning

Tue  ████████████████████  ~11,000 opens
Wed  ███████               ~4,500
Thu  ███                   ~1,900
Fri  ██                    ~1,100
Sat  █                       ~600
Sun  █                       ~500
Mon  ▏                       ~400   (illustrative shape, not one publisher's data)

A campaign starting on Thursday sees less than a quarter of that issue's opens. Its first week will look like a failure, and it is not one. The fix is to check send days before choosing a start date. Most publishers list their schedule in their profile or media kit. Our advertiser guide to newsletter inventory explains how to read a publisher's supply before buying.

The same shape explains why open-time ads keep delivering for days after a send. Every late open is a new request, decided at that moment. A campaign that launches on Wednesday is still eligible for every Tuesday-issue reader who opens on Wednesday, Thursday and Friday. It has missed the peak, not the issue.

Reason 4: even pacing cannot absorb a send-day burst

This is the reason I would most like every advertiser to understand, because the default setting works against newsletters.

If your campaign has an impression goal and flight dates, MailAdx spreads the goal across the flight with a daily allowance:

daily allowance = (impression goal ÷ days in flight) × pacing multiplier

Pacing   Multiplier   Behaviour
High     2.0          up to twice the even daily share
Medium   1.0          the even daily share (default)
Low      0.5          half the even daily share

Once today's count reaches the allowance, the campaign stops being eligible until tomorrow. And an unused allowance does not roll over. A day that delivers less than its share is simply lost.

Now combine that with the open curve above. Take a 14-day flight with a goal of 28,000 impressions on that Tuesday newsletter. Medium pacing allows 2,000 a day.

              Opens you could win   Allowance   Delivered
Tue (send)          11,000             2,000       2,000   ← capped
Wed                  4,500             2,000       2,000   ← capped
Thu                  1,900             2,000       1,900
Fri                  1,100             2,000       1,100
Sat–Mon              1,500             6,000       1,500
Week total          20,000            14,000       8,500   (61% of the week's allowance)

The campaign could have won 20,000 impressions that week and was allowed 14,000. It delivered 8,500, because the allowance was used up on the two days that had more supply than it permitted, and wasted on the five days that had less. Over two weeks it finishes at roughly 61% of goal. The ad was fine. The pacing shape did not match the supply shape.

For a weekly newsletter, High pacing is usually the better choice: it doubles the allowance on the days the opens exist. In the example above, High pacing lifts the week from 8,500 impressions to about 12,500, and the flight from 61% of goal to roughly 89%. If your goal is a ceiling you must not exceed, rather than a target you want to reach, keep Medium. And use Low only when the impression goal is truly a maximum, because at half the even share per day, Low pacing on its own cannot reach the goal over the flight.

If what you actually care about is spend, a daily budget is often a better control than an impression goal with pacing. It caps money rather than impressions, and you can set it high enough to catch the send day.

Reason 5: the filters that run before any ranking

Ranking only happens among campaigns that are eligible for this reader, in this slot, at this moment. Before that, every campaign passes through a series of filters. In the MailAdx ad server the order is roughly this:

Is the line item active, with an approved creative?
Does the creative's format and size fit this ad unit?
Does this publisher accept network demand?
Is the advertiser's wallet funded?
Is today inside the flight dates?
Is this ad unit targeted (if the campaign targets specific units)?
Does the reader's location match the geo targeting?
Is this a Gmail open, and is the campaign allowed to serve to Gmail?
Device, operating system, language
Impression goal reached?
Today's pacing allowance used up?
Total budget or daily budget spent?
Frequency cap reached for this reader?
Fatigue: 8 impressions to this reader without a click?
Audience segment and retargeting consent
Segment floor price

Every one of those can remove your campaign for some readers, and several have newsletter-specific traps.

Gmail and location targeting

Gmail fetches every image through its own proxy, so an open from a Gmail user does not reveal the reader's location or device. MailAdx lets each line item decide whether to serve to those opens. With the setting on, which is the default, the campaign serves to Gmail users and skips geo targeting for them. With it off, Gmail users are excluded entirely. For a consumer newsletter where Gmail is often the largest single provider, turning it off can remove a large share of supply at a stroke. Our post on Gmail and Apple Mail tracking covers what each provider hides.

Tight geo or device targeting on a small list

Targeting a single city on a 15,000-reader newsletter can leave a few hundred eligible readers. They each open about once a week. Do that arithmetic before launching, and widen the campaign or add publishers if the number is small. The audience targeting guide goes through which targeting dimensions are reliable in email and which are not.

Frequency caps and fatigue

Because a newsletter's most engaged readers open nearly every issue, a frequency cap reaches them fast. A cap of three per week on a daily newsletter means the core readers are exhausted by Wednesday. Fatigue works similarly but on response: eight impressions to the same reader with no click pauses the campaign for that reader for a week, or until they click. Both are good for the campaign. Both reduce the delivery number. The ad fatigue post and the post on cross-publisher frequency capping explain how to set them.

Wallet and approvals

MailAdx is prepaid. A campaign whose advertiser wallet is empty is skipped, and so is one whose creative has not been approved. Both are simple, both are common in week one, and both show up in the dashboard rather than needing any analysis. Check them first.

Floors

Publishers set minimum prices on their ad units. A campaign bidding below a unit's floor is never eligible on that unit, however well it would perform. If delivery is concentrated on a few publishers and absent on others, compare your bid with their floors. Our floor CPM guide explains how publishers set them and the CPM benchmarks show typical ranges.

What we learned from logging every non-delivery

For a long time our ad server could tell you what it served and very little about what it did not. So when an advertiser asked why a campaign was not delivering, the honest answer was a guess informed by experience.

In September we started recording how every serving request ends, filled or not, with the specific rule that ended it, counted in 15-minute buckets. It changed how we answer this question. In the first hours, the biggest single cause of blank slots across the platform was not demand at all. It was ad tags arriving without a subscriber identifier, a publisher-side template problem described in our post on blank newsletter ad slots. Advertisers had been reading that as weak delivery.

The same habit of checking turned up a bug of our own that is directly relevant here. Our fatigue rule was meant to pause a campaign for a reader after eight impressions without a click, then reset after a week. On one storage path, the reset never happened. Eight non-click impressions excluded a campaign from that reader permanently. Over the life of a campaign, more and more of the most engaged readers were silently excluded, and delivery decayed in a way that looked exactly like a campaign running out of steam. It was fixed in late September and the one-week window now applies everywhere.

I mention it because it is the kind of cause no advertiser could diagnose from the outside, and because it argues for one practical habit: when delivery falls in a way you cannot explain, ask the platform which rule is excluding you, not just how many impressions you got. A platform that cannot answer is guessing.

A week-one checklist

Symptom                                   Most likely cause                    What to do
────────────────────────────────────────  ───────────────────────────────────  ─────────────────────────────
Zero impressions anywhere                 Wallet empty, creative not approved,  Check status, wallet, flight
                                          start date in the future              dates
Zero until a specific day, then a jump    Waiting for the newsletter's send     Nothing; next time align the
                                                                                start with send days
Delivers on send day, then stops          Medium/Low pacing allowance used      Switch to High, or use a
                                          up by mid-morning                     daily budget instead
Strong start, then fading (CPC/CPL)       Real CTR below the prior; estimate    Improve creative; or accept
                                          falling as evidence arrives           the price the CTR implies
Slow start, then rising (CPC/CPL)         Real CTR above the prior              Nothing; it is working
Delivers on some publishers, not others   Bid below their floor, or targeting   Compare bid with floors;
                                          excludes their audience               check geo and Gmail setting
Steady decline over weeks                 Frequency cap or fatigue on the core  Refresh creative; widen
                                          readers                               publishers; review caps
Trickle on a slot with a sponsor          Guaranteed placement wins outright    Choose other issues or slots

When to wait and when to act

For CPM campaigns, there is very little to wait for. Ranking is decided by your bid from the first request. If a CPM campaign is not delivering after a newsletter has sent, the cause is almost certainly a filter or pacing, and waiting will not change it.

For click and lead campaigns, the question is when you have enough evidence to judge the click rate. The answer is more than most people think. At a 0.5% click rate, the margin of error on your measured rate is roughly:

Impressions   Expected clicks   95% range of measured CTR
        500              2.5        about 0.0% – 1.1%
      2,000               10        about 0.19% – 0.81%
      5,000               25        about 0.30% – 0.70%
     20,000              100        about 0.40% – 0.60%

At 500 impressions you cannot tell a 0.3% ad from a 0.8% one. At 5,000 you usually can. So judge creative after a few thousand impressions, not after a few hundred, and judge it per issue rather than per day, because one issue's audience differs from the next. The post on tracking newsletter ad ROI covers what to measure beyond clicks, and the bot click post explains why some early clicks are security scanners rather than readers.

Unlike the social platforms, editing a campaign on MailAdx does not reset anything about its history. Its impressions and clicks are kept, so its estimate stays where the evidence put it. If you have found a real cause, such as a pacing mismatch or a Gmail exclusion, change it immediately.

What publishers can do

Most of this post is for advertisers, but publishers feel week-one under-delivery too, in the form of empty slots and advertisers who do not renew.

Publish your send schedule where advertisers will see it, so start dates line up with issues. Keep your ad tag healthy, because a slot that sends no subscriber identifier cannot be filled by anyone; the ad unit best practices cover the setup. Set floors from data rather than hope. And when a campaign under-delivers on your list, look at the cause before offering a make-good: a make-good fixes the advertiser relationship, but not a pacing setting that will cause the same shortfall next month.

The short version

A newsletter ad campaign competes for a lumpy, finite set of opens, one known reader at a time. In week one it has no click history, a small share of trial impressions, possibly no sends yet, a pacing allowance shaped for steady supply, and a long list of reader-level rules to pass. Each of those is visible and each has a fix. The fixes are mostly settings, not patience.

If you are planning a first campaign, the advertiser product page explains how buying works on MailAdx, the CPM calculator helps size a budget, and the small-budget guide covers what a first $500 can realistically test. You can create an advertiser account and set your first line item with pacing chosen for the newsletters you buy, not the default.

Frequently asked questions

Why is my new newsletter ad campaign not delivering?

Check the simple causes first: an unfunded wallet, an unapproved creative, flight dates in the future, or a newsletter that has not sent since your start date. If those are fine, the usual causes are a pacing allowance that is used up early on send days, targeting or a Gmail exclusion that leaves few eligible readers, a bid below the publisher's floor, or, for click campaigns, a predicted click rate that ranks your bid below the current leader.

Is there a learning phase for newsletter ads?

For click and lead campaigns, yes, in effect. The ad server starts from a prior estimate of your click rate and blends in your real results. On MailAdx the campaign leaves cold start at 500 impressions. CPM campaigns have no learning phase in ranking, because they compete on their bid from the first request.

Should I wait a week before changing a newsletter campaign?

Not if you have found a concrete cause. Editing a MailAdx campaign does not reset its history, so fixing a pacing, targeting or Gmail setting has no learning cost. Wait only when the question is whether the creative works, and then wait for a few thousand impressions rather than a fixed number of days.

Which pacing should I use for a weekly newsletter?

Usually High. Weekly newsletters produce most of their opens on the send day and the day after. Even pacing caps each day at the same allowance, so it stops the campaign during the burst and leaves the allowance unused on quiet days. High pacing doubles the daily allowance so the campaign can deliver when the opens exist.

What does exploration mean in ad serving?

It is the share of requests on which the ad server deliberately shows a campaign other than the current leader, so new campaigns can collect evidence. On MailAdx that share ranges from 20% when the leading campaign is new down to 3% when it has more than 50,000 impressions, with four extra points overnight UTC.

Why did my campaign deliver well at first and then slow down?

For click campaigns, the most common reason is that the real click rate is below the starting estimate, so the campaign's ranking falls as evidence arrives. For all campaigns, frequency caps and fatigue rules progressively exclude the most engaged readers. A creative refresh or more publishers usually helps more than a higher bid.

Do Gmail opens count for geo-targeted newsletter campaigns?

Gmail hides the reader's location by fetching images through its proxy. On MailAdx each line item chooses: serve to Gmail opens without geo targeting, which is the default, or exclude Gmail users entirely. Excluding them makes targeting stricter but can remove a large share of a consumer newsletter's opens.