500,000 Christmas greetings and not a single cut-off name: What hyper-personalized videos need in large quantities

Sources

• With hundreds of thousands of videos, you can't watch every single one. The template itself must handle every data record cleanly.
• The key is text that adapts to its length, containers that adjust to the text, and images that fit into any frame.
• When every layer parameter can be controlled via data, a single template is often enough for many variations.
• Data is provided via feed (CSV, JSON, spreadsheet) or API, and the videos are rendered in advance.
• Quality assurance means: specifically searching for and reviewing edge cases.

In mid-December, Annegret Schmidt-Brockhausen opens an email from her car dealership. In the video, the team wishes her a Merry Christmas, with her name in large letters and her dealer's name underneath. Everything is perfectly aligned. The hyphenated name wraps at the right spot, nothing spills over the edge, and the animation runs just as smoothly as it does for Mr. Ott, who received the same email.

For a single video, this is a given. You build it, watch it, make corrections, and you're done. With 500,000 videos, it’s a different story. Somewhere in the database is the longest surname in the country, somewhere is a dealer with a three-line company name, and somewhere is a field that was left blank. And none of these cases can be allowed to stand out.

We built vidflare for exactly these scenarios. What a platform needs to be capable of to handle this is less obvious than it sounds.

Why volume is the real problem

Almost any tool can handle personalized videos with a handful of variations. You create a few versions, check each one individually, and approve them. That approach falls apart once you reach a few hundred data records. No one is going to watch 500,000 videos, not even by taking random samples that would guarantee anything.

The responsibility therefore shifts from the finished video to the template. It must know how to react to every conceivable data record in advance: What happens if a name is twice as long as planned? If a dealer photo arrives in portrait instead of landscape format? If the color for a variation comes from the data instead of the design?

Those who answer these questions while building the template can simply hit start at the end. Those who leave them open will get them back in the form of complaints.

We used to program these videos

vidflare is a product of MJOO. In our early years, we programmed hyper-personalized videos line by line. Every project had its own custom code that governed how text wrapped, where images were placed, and which animation started when. It worked, but it was labor-intensive, and every design change had to go back to development.

From those projects, we know exactly where templates break when dealing with large volumes of data. That knowledge is now built into vidflare. We can still program, and in exceptional cases, we do. However, about 95 percent of requirements can be handled directly in vidflare, without a single line of code.

For you, this primarily means that design changes no longer end up in a development loop, but go directly into the template.

Text that handles any length

Names are the most common personalization element and, at the same time, the most unpredictable. "Ott" and "Schmidt-Brockhausen" have to fit into the same space, and both should look good.

In vidflare, you define how each text layer behaves:

• Auto-fit automatically adjusts the font size so that the text always fits within its frame.
• Max-rows limits the number of lines. For example, a name may break onto two lines, but never three.
• Native or glyph-accurate positioning determines how vidflare positions the text: in its native text box or glyph-accurate, i.e., at the edges of the actually visible letters. Glyph-accurate is practical when a name needs to be perfectly flush with a surface or the edge of an image.

It’s the combination that makes the difference. With Auto-fit alone, a very long name would eventually become tiny. With Max-rows alone, it would be cut off. Together, you can control how a long name reacts: it wraps, shrinks, or both, without ever falling out of the frame.

Shapes that follow the text

Many designs use colored shapes behind the text, such as a bar under a name or a highlight for individual words. With static text, you simply draw the shape to fit. With 500,000 different texts, the shape must grow with the content.

In vidflare, you therefore assign shapes (solids) to a text that align with it. There are three reference points for this:

• Box: The shape encloses the entire text block, regardless of how many lines it has.
• Line index: The shape is attached to a specific line, for example, only to the first one.
• Word index: The shape highlights a single word, such as the first name in "Merry Christmas, Annegret."

This keeps a design that plays with shapes looking clean, even with unpredictable text lengths. The bar under a short name is short, the one under a long name is long, and if there are two lines, each line gets its own.

Define animations once, use them everywhere

A holiday greeting rarely consists of just one animated text element. Names, dealers, greetings, and logos usually need to appear in the same rhythm, with the same movement and timing. If you build that for each layer individually, you'll have a lot of work to do the moment you need to make a change.

vidflare uses fx-layers for this. An fx-layer describes an animation—for example, a smooth fade-in from the bottom—and can be assigned to any number of layers. If you change the animation in one place, it updates everywhere it is used.

This saves time during production. More importantly, it keeps the video consistent, even if the content varies significantly from one data record to the next.

Every parameter can come from the data

Personalization doesn't stop at the name. In vidflare, every layer parameter can be controlled via input data, from position (x/y) and colors to alignment.

That sounds technical, but it has a very practical effect: you need fewer templates. Here are a few examples of how that can look:

• An insurance company can color-code a video based on the specific plan, all using a single template.
• A car dealership with multiple brands can place the appropriate brand logo in the position required by that specific logo format.
• An online shop can adjust text alignment for customers who use right-to-left languages.

Instead of maintaining ten templates, you maintain one and add one extra data column. For large campaigns, this is often the difference between a manageable project and one that requires ten times the effort for every single change.

Images and videos that fit into any frame

As soon as visual assets come from data, things get messy. Dealers send photos of their showrooms in both portrait and landscape, logos arrive as squares or wide banners, and product images come in every conceivable resolution.

In vidflare, you define a frame for every image and video layer and determine how the material behaves within it. Options include:

• Native: The material is used at its original size.
• Fit to fill: The material completely fills the frame, with any overflow being cropped.
• Fit to width: The material is scaled to the width of the frame.
• Fit to height: The material is scaled to the height of the frame.

For showroom photos, Fit to fill is usually the right choice because the image will always appear borderless. For logos, a mode that doesn't crop anything is better. You make this decision once per layer, not for every image individually.

Data in, videos out

The best template is of little use if getting the data into it is a struggle. vidflare accepts data in two ways:

• As a feed in CSV, JSON, or spreadsheet format. This is the easiest way if you are setting up a campaign from an export.
• Via API, if you want to connect vidflare directly to your systems.

You can also connect your CRM using these two methods. In practice, this usually means the CRM exports the recipient list with the fields that need to be personalized, and vidflare handles the rest.

The videos are rendered in advance. When the email is sent, every video is already finished and ready, so nothing needs to be calculated upon retrieval. This is particularly convenient when sending to hundreds of thousands of recipients at once, as no recipient has to wait for their video.

There is no general rule for how long rendering takes. It depends heavily on the smart layout—specifically, how long the video is and how many personalized elements, images, and animations it contains. Therefore, plan for render time early on and get an estimate for your specific layout before setting your send date.

An example: Christmas greetings for Mercedes-Benz

For Mercedes-Benz, we used vidflare to create around 500,000 personalized Christmas greetings. Two things were personalized: the recipient's name and their dealership.

That might sound simple, but it is exactly the kind of combination where templates often fail. Customer names range from three letters to hyphenated names with titles. Dealership names are often even longer because they include the location, legal entity, or branch name. Both texts must look good in every single video without anyone having to make manual adjustments.

The dealership in the video is more than just a detail. For many customers, the local car dealership is their primary point of contact. A greeting that appears to come from there feels more personal than one from the corporate office, even if it was produced centrally. In our experience, it is precisely this connection between the brand and the local contact that strikes the right chord in such campaigns.

Quality assurance: looking for the edge cases

If you cannot watch every single video, watch the ones where something might go wrong. Random spot checks are of little help because most data sets are unremarkable. The problems lie at the edges of the data.

Before a large campaign goes out, it is worth taking a targeted look at these cases:

1. The longest and shortest values in every personalized text field.
2. Names with hyphens, titles, or name affixes like "von" or "zu".
3. Special characters and umlauts, including those in spellings the font might not be designed for.
4. Empty fields, such as when no dealer is assigned to a customer.
5. Images in unusual formats or very low resolution.

These data records can be filtered out and checked as a preview before the main rendering process. If they look good, it is highly likely that the rest will look good too. For empty fields, you should decide in advance what should happen: use neutral phrasing, a default value, or exclude the record entirely.

Consider data protection from the start

Personalized videos use personal data, and with a campaign involving hundreds of thousands of recipients, this quickly adds up to a large amount. Therefore, GDPR is a factor from the very beginning.

Think about which fields you really need beforehand. For a Christmas greeting, a name and dealer are sufficient; purchase history or vehicle data should not be in the feed if they do not appear in the video. Also, clarify the legal basis for contacting the recipients, where the data is processed, and how long the data and finished videos will be stored.

This is not legal advice. Before launching a campaign of this scale, speak with your data protection officer.