Email and Text Campaigns: Two Sides of the Same Send

Contents11 sections

One message, two sides

Every marketing message has two sides. The customer sees one email or one text. The business sees a send event, a row of response data and a line in a campaign report. Campaigns work when both sides of that coin are designed together. The customer feels every shortcut taken on the business side.

I learned this running the data warehouse and business intelligence (BI) organization at Infogroup's Yesmail, a national email marketing and analytics provider serving large corporate clients. There, campaign, customer and response data was the product itself. The pipeline behind a send decides whether the message lands as useful or as noise. A/B testing lets customers pick the winning message. Enrichment and segmentation turn a mass send into a relevant conversation.

How an email campaign works

A campaign is a loop, and each pass produces data that makes the next one sharper.

• Build the list. Collect opted-in addresses and phone numbers through sign-ups, purchases and loyalty programs. Remove anyone who unsubscribed, replied STOP or bounced.
• Segment. Split the list into groups that should hear different messages.
• Create. Write the subject line, content, offer and call to action. Every link carries tracking codes that tie a click back to the person, the campaign and the version.
• Send. An email service provider delivers at volume, with sender authentication so inbox providers trust the mail.
• Track. Events flow back: delivered, bounced, opened (a hidden image loads), clicked (a tracked link redirects), unsubscribed, and converted (a purchase traced to the click).
• Learn and repeat. Response data feeds the next campaign's segments, offers and timing.

The customer sees only the send. Everything else is the business side of the coin, and it decides whether that one message lands as useful or as noise.

The customer's side: one message, one moment

The customer never sees a campaign. They see a subject line in a crowded inbox, or a text on a phone that's already buzzing. In that moment they weigh three things: whether they know the sender, whether the message is for them, and whether it's worth their time.

Email and text answer those questions differently. Email is patient. It waits in the inbox, carries images and detail, and can be read later. A text is intimate and immediate. It lands next to messages from family, so an irrelevant one feels like an intrusion.

That difference sets the rules. Email tolerates more frequency and more content. A text has to earn its place. It does that only with timely, short, clearly useful messages, such as an order update, a reminder, or a time-limited offer the customer asked for. On both channels, consent and an easy way out are part of the experience. An unsubscribe link that works and a STOP reply that's honored are how a brand shows respect.

The business side: a data pipeline that never stops

Behind that single message sits an operation. Client lists and feeds arrive in different shapes. Sends go out in high volume on tight schedules. Every delivery, bounce, open, click, unsubscribe and complaint comes back as an event. Those events have to be loaded, matched to the right customer and campaign, and reported to the client, often by the next morning.

At Yesmail, the hard problems lived in operations:

• Onboarding at scale. Each large client brought its own feeds and schema. We automated the construction of DDL, the data definition statements that create and change database tables, along with schema change, so onboarding a client became configuration rather than hand-written tables.
• Recoverable loads. We built automated hydration, so a target table could be emptied and refilled from the landing layer, where the raw feeds were kept. A bad load became a rerun instead of an incident.
• Trust in the numbers. Delivery service-level agreements (SLAs) and a data governance program kept reporting consistent. Clients made budget decisions on those reports, so a wrong number was a business failure.
• Delivery speed. Agile practice, automated testing and deployment automation sped our biweekly delivery 15 percent.

The lesson for any marketing organization is that the customer experience is only as good as the pipeline behind it. A duplicate record means a duplicate message. A late suppression load means texting someone who already opted out.

Reading the funnel from both sides

Each campaign metric is one event seen two ways. To the customer it's a choice they made. To the business it's a measurement with known blind spots.

Table 1. Each campaign metric from the customer's side and the business side, with its blind spot.

Conversion to sales is the final judge. Click-through is how we knew which campaign was generating the interest. Text campaigns have no open signal at all, so for SMS, the text-message channel, the click-through is the first and best measure of interest.

A/B market testing: letting customers vote

A/B testing is a controlled experiment run before the full send. Instead of guessing which message will work, the business lets a sample of customers decide.

• Pick one variable. Choose the subject line, offer, image, send time, call to action or segment, and change only that, so you know why a version won.
• Split a random sample. Take, for example, 20 percent of the audience and divide it evenly between version A and version B. Random assignment keeps the groups comparable.
• Track each version separately. Each version has its own tracked links, so every click-through is credited to the exact version that earned it.
• Read the result. After a set window, compare click-through rates and check that the gap is larger than random chance would explain.
• Roll out the winner. The winning version goes to the remaining 80 percent, and sales data confirms that the interest turned into revenue.

Click-through is the right first read because it arrives fast and measures intent: the customer chose to act on that message. Opens show who noticed; click-throughs show which message worked. Testing several variables at once, called multivariate testing, is possible. It needs a much larger audience to reach a reliable answer.

Both sides of the coin gain. The business spends its full send on the strongest message. Most customers receive the version their peers already showed they cared about.

Segmentation and enrichment: from a list to a market

A client's list usually knows little beyond a name, address and purchase history. Enrichment appends what the list lacks, and segmentation groups people so each group gets a message that fits. Infogroup was one of the national list compilers that made this practical, matching a client's records by name and address to a large reference file of households and businesses.

Table 2. Enrichment categories, with example attributes and their typical use.

Much of this data is modeled, meaning it's statistically estimated for a household from similar households rather than known for that person. That makes it useful for groups and unreliable for individuals.

Segments become test cells. A campaign can be A/B tested within a segment to find its best message, or the same message can be tested across segments to find who responds. Over time, the response data itself becomes the strongest segmentation input. What someone did last month predicts their next purchase better than any appended attribute.

The customer's side of the coin is protected by law, and the rules are strict:

• Protected traits. Federal law bars using race, national origin, sex, religion, age, familial status or disability to target or exclude people in credit, housing and employment offers. In June 2022, Meta settled a Justice Department housing discrimination case and agreed to withdraw an ad-targeting tool over this. Outside those areas, most marketers avoid race-based targeting anyway because of the legal and reputational risk.
• Credit data. Consumer credit reports can be used for marketing only in narrow cases, such as prescreened firm offers of credit or insurance. General marketing relies on aggregated or modeled credit attributes.
• Health data. HIPAA and state laws such as Washington's My Health My Data Act restrict health-related targeting.
• Consent. CAN-SPAM governs commercial email and requires a working unsubscribe. The Telephone Consumer Protection Act (TCPA) requires prior express written consent for marketing texts. The General Data Protection Regulation (GDPR) in Europe and a growing set of US state privacy laws give people the right to opt out of data sales and profiling.

Good enrichment improves relevance without making the customer feel watched. If a message reveals more than the customer expects a brand to know, the segmentation has gone too far.

Optimizing touches: fewer, better messages

A touch is any message the customer receives. Every touch spends a little of the customer's attention and a little of the brand's goodwill. Optimization means spending that budget where it returns the most, on both sides of the coin.

In practice that comes down to four levers:

• Frequency. Response data shows when added sends stop producing sales and start producing unsubscribes. That point differs by segment, so frequency caps should too.
• Channel. Email suits detail and consideration, and text suits urgency and service. A customer who never opens email but clicks texts is telling you their preference.
• Timing. Send-time patterns by segment beat a single blast hour. Triggered messages, such as an abandoned-cart or back-in-stock alert, beat both, because the customer started the conversation.
• Testing. A/B tests on subject lines, offers, creative and segments, read first on click-through and confirmed on conversion rather than opens, keep the program honest.

The result is fewer messages that each do more. Customers get less noise. The business gets lower send costs, a better sender reputation and higher revenue per touch.

Why it matters

The tools have changed since the Yesmail years. Customer data platforms, real-time event streams, and AI that picks send times and drafts content now do what we built by hand in a data warehouse. The coin still has two sides.

What changes is who pays when the sides drift apart. A campaign idea the pipeline can't measure is a guess with a budget behind it. A duplicate record or a stale suppression list reaches a real person as a second message or a text they already refused, and that person blames the brand — never the warehouse. As AI writes and times more of the messages, more of the weight lands on the data behind them. Get both sides right and the customer sees a message worth reading, while the business sees a channel worth funding.

Sources

• US Department of Justice, United States v. Meta Platforms, Inc. (S.D.N.Y.), complaint and settlement filed 21 June 2022.

Glossary