Broken personalization is worse than none

A brand that ignores your name is forgettable. A brand that gets your name slightly wrong, or asks you to review a meal you never ate, has told you something specific about how much attention it is paying.

The subscriber cannot see the bug. She sees a company that is confused about who she is.

This post covers the brands in the QSR email index that did use the data they collected, and used it badly.

About this study. The QSR email index enrolled in the email, loyalty, and lifecycle programs of 127 quick service and fast casual restaurant brands using one identity and one inbox. Every message is captured, labeled, and timestamped. No claim about a named brand appears here without a verbatim quote, a timestamp, or a directly observed screen.

The Great Greek asked her to review a meal she never ate

This is the most revealing single email in the study, and it needs to be read next to its own signup form.

The front door is the most demanding in the study. The Great Greek requires an email verification code and a phone verification code, neither optional, behind a panel advertising “Join Our Email List / Join For Free.” Two-factor authentication to join a mailing list.

The third thing that list sent was this. Subject line: “Enjoy your meal?” Body:

Hi Corrie, Thanks for your recent follow. Did you happen to dine with us? If so, we’d love to hear your thoughts on whatever dishes you tried. Just visit our raves page and write a brief review on any dish… Thanks and we appreciate your business!

Three separate failures, each quotable alone.

1. The subject asserts a meal that did not happen, and the body hedges. “If so” means somebody knew the trigger was unreliable and shipped it rather than fixing the segment. That is a decision, not an accident.
2. It solicits a product review from a person with no experience of the product. This is a review integrity problem before it is a personalization problem. Every review a brand generates this way makes its own ratings less useful.
3. “Thanks for your recent follow” is social media vocabulary sitting in an email template. Nobody followed anything. The copy was written for a different channel and never revised.

Two verification codes to get in. A request to review a meal she never ate on the way out. The gap between what a brand demands at the door and what it does with the access is the whole subject of this study, and no row states it more clearly.

Capriotti’s mangles a name it stores correctly

Capriotti’s referral module reads:

INVITE FRIENDS WITH CODE: CorreH2743

Her name is Corrie.

We narrowed this claim twice before publishing it, and the narrowing is what makes it survive. Our first version called it a misspelling with no baseline, since the email carries no greeting to compare against. That version was withdrawn. The subscriber then logged into her Capriotti’s account, and the page greets her “HI CORRIE!” correctly.

So the mechanism is located. The brand holds “Corrie.” Its referral code generator emits “Corre.” This is not a capture problem and not a database problem. It is also not a truncation, because a five-character cut of “Corrie” gives “Corri,” so a transformation rule is mangling the string rather than shortening it.

Sourcing note: the code is email evidence, the stored name is account interface evidence. This study labels those separately, and this claim depends on both.

This is the worst-placed broken merge in the study, and placement is what makes it matter. Every other broken merge here is private. This one is built to be forwarded. It is the field the brand asks her to send to friends, so a misspelling of her own name travels outward every time the referral mechanic works exactly as designed.

One further detail from the email itself: there is no greeting at all. Capriotti’s holds her name, displays it correctly on its own website, mangles it in the referral code, and omits it from the copy.

The rest of that email is competent. Clear mechanics, a real referral incentive, live text rather than images, and a decent campaign name. The one field engineered to travel is the one that is broken.

Two brands drop the email address where a first name belongs

Shipley renders “corrie You’re In!”

la Madeleine opens “Bonjour corrie!”

Both in lowercase. Both taking the portion of the email address before the @ symbol and dropping it into a slot built for a first name.

la Madeleine’s is the sadder of the two, because “Bonjour” is exactly the right instinct for a French bakery cafe. The brand had a voice, reached for it, and the merge field ruined the line.

This is a two-brand pattern rather than an anecdote, and it has a specific cause: a fallback rule that reaches for the address when the name field is empty, written by somebody who never looked at the output.

Sonic asked for a first name and shipped “Hi there”

Sonic’s form requests a first name. Its email greets her “Hi there.”

Either the field never reached the platform, or the fallback fires regardless of whether a value exists. Both are worse than not asking, because the brand spent a form field and the subscriber’s patience on a value it then declined to use.

KFC produces the most specific version of half-personalization in the study. It clearly holds her name, because it rendered “CH” as an initials avatar in the email. It then spent that name data drawing two letters inside a circle rather than saying hello. The name was available, it was processed, and the output was a monogram.

Right data, raw presentation

Golden Corral prints the restaurant the subscriber selected herself, which puts it in the small group of three brands doing something almost nobody does. The data is correct.

The presentation is not.

• The ZIP code renders as a five-digit code zero-padded to nine digits with no hyphen.
• The street address prints in all capitals, exactly as stored.
• A lone period renders on its own line above it.
• The From header and the footer name two different sending addresses.

This is a different failure from the others in this post. Nothing here is wrong. Nothing here was proofread either. It is what a database field looks like when it is piped into an email without anyone looking at the rendered result, and it is a reminder that “we personalize” and “we checked what personalization looks like” are separate claims.

Back-to-school, sent to everyone

Three brands sent back-to-school offers to a subscriber with no household signal, all within four days.

• Bojangles sent a free Kid’s Meal.
• Auntie Anne’s sent a licensed kids meal promotion.
• Dave’s Hot Chicken sent a “Class is in session” bundle.

The brands know nothing about children in this household, because nothing on any of their forms asked. They sent a family offer because the calendar said August, not because anything in their data said family.

This is the mirror image of the zero-balance problem. One acts on a default because the real value is unflattering. The other acts on a default because no real value was ever collected. Both produce a message written for a person who does not exist.

What working personalization looks like here

The counterexamples matter, because they show the failures above are choices rather than platform limits.

Seven brands merged a name into a subject line and got it right: Chili’s, Applebee’s, Subway, Panda Express, Habit, Perkins, and Sheetz. All on ordinary platforms, none of them doing anything exotic, and all of them handling the same merge field that produced “corrie” and “Hi there” elsewhere in this study.

Three of those brands went on to use the name again in a later message rather than stopping at the welcome, which is covered in the post on welcome sequences. A separate group of three printed the subscriber’s selected restaurant in the body of an email, covered in the post on collected data.

None of that is technically difficult. The difference between these brands and the ones above this section is that somebody opened the rendered email and read it.

The pattern underneath all of it

Every failure in this post has the same shape. A brand built a personalization slot, wired something into it, and then never looked at the output.

The name field, the location field, the referral code, the fallback rule, and the seasonal segment all work exactly as configured. Nobody checked what “as configured” produces when a real subscriber’s real data flows through it.

Broken personalization is not a data quality problem. It is a proofreading problem with a database attached. The fix costs one person, one test address, and ten minutes per campaign, and it is the single highest-return thing most of the brands in this study could do this quarter.

This is post five in the QSR email index series. Post six covers the brands using email to move subscribers into an app.