The way to keep video ads current when inventory turns over every few days is to stop treating each ad as a project and start treating it as a view of a data row. When the row changes, the ad is regenerated. When the row disappears, the ad is retired. Templater handles the regenerating part from inside After Effects, and the rest depends on how well you wire your inventory to it. The render is the easy part. Knowing what changed is the hard part.
This article is for the person who runs ads for a business with fast-moving stock, such as a used-car lot, a boat dealer, a furniture clearance operation or a rental manager. Items arrive, prices change, and a few days later half the list is different. If your video ads are made by hand, they go stale before the budget is spent, and a stale ad is worse than no ad because it advertises something you no longer have. Here is how to build a pipeline that keeps up, and where its limits are.
Three kinds of change, and why each needs its own answer
This section is general practice, and it is the part most teams skip. Inventory does not just “update.” It changes in three different ways, and each one asks something different of your ads.
An item can arrive. That needs a new video and a new ad. An item can change, with a new price, a new photo, a status such as pending, or a corrected detail. That needs the existing video regenerated, and the ad that uses it replaced or updated. An item can leave, because it sold. That needs the video and the ad retired, and fast, because showing a sold car to a buyer is the most expensive mistake on the list.
Most teams build for the first, half build for the second, and discover the third when a customer calls about something that left on Tuesday. So design for all three from the start, and decide who owns each. Arrivals are a production question. Changes are a data question. Departures are a control question.
Where the video itself is concerned, Templater is a production tool. It makes a video from a template and a row of data. It does not know whether an item has sold unless your data says so, and it does not remove an ad from a platform. Our product pages list Vimeo, JW Player, Frame.io and AWS S3 as destinations for Projector. They do not list advertising platforms. That is not a flaw to hide. It tells you where your own scripts or your ad tool need to do the work, and you should plan that part before you build the template.
Wire the inventory to a render so a change becomes a new video
Start with the source of truth. Your inventory lives somewhere, in a dealer management system, a spreadsheet, a database or a feed. Templater reads Google Sheets in the cloud, tab-separated files exported from Excel, Numbers or Sheets, and JSON files or feeds. It cannot read an Excel file directly, and it maps to one worksheet and one data source. If your system produces a feed in JSON, use it. If it does not, the realistic bridge is a sheet that something writes to, whether a script, an export you schedule, or an automation tool.
QUE is the route to consider if several systems feed the same pipeline. It is Dataclay’s JSON-based data manager, with an API that receives data from web, mobile or other apps, and a Zapier integration that connects it to more than five thousand apps without code. Items arrive as jobs, jobs are grouped in campaigns, and the computers running After Effects and Templater, which QUE calls satellites, process them. The QUE Dash shows each job’s status and how busy the machines are, so a person can look at one screen and see what is waiting. QUE does not balance load across satellites. You assign jobs to machines and make sure each has the fonts and plugins the job needs.
Then there is the render itself. In the Bot edition, Templater checks periodically for incoming data and starts the customization and rendering by itself. It runs as a background process and handles very large data sets, and you can also push data in from the command line. The Pro edition renders when someone clicks. For stock that turns over every few days, Pro may be enough, since a person can run the batch each morning. If the lot is large or the turnover is faster, Bot saves the click, and the person.
The status column deserves attention here. Bot reads a render-status column in the data source to decide whether to render a given row, and it will stop with an error if the column is missing. That column is your control surface. A new arrival gets a row with the status ready. A changed item needs its status set so that Bot treats it as work to do. We have not documented every way Bot treats a row that is rewritten after it was rendered, so test that in rehearsal before you rely on it, and write down who resets the status when a price changes. Our post on flags in data sources covers the reserved column names.
Name the outputs from the data. Templater can set a custom output file name for each job from the data source, so a file can carry the item’s own identifier, such as a stock number. That detail sounds small and it is the one that makes retirement possible. If each video file and each ad carries the same identifier as the inventory row, a sold item can be matched to everything that has to come down. If files are named final and final-2, you will chase them by hand.
Here is a concrete case, described as an illustration and not a customer. Picture a used-car lot that changes a good part of its stock in a few days. The office keeps one sheet, one row per vehicle, with a stock number, the year, make and model, the photos folder, a short feature line, and a status column. When a car comes in, a person fills the row and sets it to ready. Bot renders a short video from a template built in After Effects, named from the stock number. When a car sells, the person who closes the deal marks the row sold. The morning routine is to render what is ready and to retire what is sold. The template never changes. Only the rows do.
Close the loop after the render
A rendered file is not an ad. Something has to deliver it, and something has to take it down. This is where most pipelines leak, so we will spend time on it.
Delivery first. Projector renders several customized videos at once, one per core on the machine, transcodes as part of each job, and sends finished videos to Vimeo, JW Player or Frame.io, with archival to AWS S3. It reads data from Google Sheets or QUE, and it retrieves metadata from the streaming platform and writes it back to your data source. That last step is useful for ads, because it can put a platform-assigned identifier or link in the row, which your ad process can use. Check what your platform returns, since we cannot list that for you. If your ads live on a platform we do not name, the general route is a script. Our FAQ says you can write a script that uploads videos when they finish rendering, and that a script can return the URL of the video to the data source, which email marketers use to embed the right link in each campaign. The same trick works for an ad system. The script uploads the file, writes back the address, and your ad tool reads it from the row.
Templater supports event scripts, using shell scripts and ExtendScripts that hook into its processes, and our GitHub repository carries samples for Windows, Mac and Node.js. Our custom development team can write them for you, and previous engagements include sending an email or SMS notification when a render completes. That last one is a handy safeguard. A message to the person who owns the lot when a batch finishes, or fails, tells them to look. Engagements are quoted individually.
Retirement second. Because Templater makes files and does not manage ads, the retirement step belongs to your ad tool or your script. The rule that works is simple. A row marked sold triggers two things, the video is removed or unlisted where it was posted, and the ad that points to it is paused. If you cannot automate both, make the manual step part of the same routine that marks the row, so that sold and retired happen together. A separate cleanup job that someone remembers on Fridays will always be a few days late.
Quality third. Fast turnover is not an excuse for skipping the check, because a wrong price in a template runs at the same scale as a right one. Run a spot check across the day’s rows before you let the batch ship. Pro and Bot can export still images or short clips of chosen frames from the merged compositions, which shows you the car with the long model name or the row with no photo while it is still cheap to fix. Our post on spot checks explains how to choose the frames. And keep the template light, because Templater evaluates every layer, and unused layers cost render time on every video. Our guide to speeding up After Effects exports shows how.
If a morning batch is big, estimate it. Render one representative video on a machine like the production one, multiply by the count, and compare to the time you have. Running Bot on several machines divides the total, with two machines cutting it in half, and the data source needs to be set up for multiple Bots. A batch that finishes at noon when the ads were meant to be fresh by eight is a design problem you can see a week early.
It helps to write the routine down as a short list that anyone on the team could follow, because turnover this fast means the person who built the pipeline will not always be the one running it. A good routine has a morning half and an evening half. In the morning, someone confirms that the sheet matches the lot, runs the spot check on the new and changed rows, and starts or confirms the batch. In the evening, someone confirms that everything marked sold is down, and that the day’s finished videos are posted where the ads expect them. Each step has a name beside it. The list is short on purpose, because a long one gets skipped when the lot is busy.
Give the pipeline a way to tell you when it is unhealthy. Templater can log its activity, so you can inspect what it did and when, and troubleshooting is far easier when the log exists. Add a notification when a batch finishes or fails, as described above, and a count of rows rendered against rows marked ready. If those two numbers disagree, something was skipped, and you want to know that before a customer does. A pipeline that fails quietly is worse than a manual one, because nobody is watching it.
Finally, practice the bad day. Once, before you rely on the system, run a rehearsal in which three things go wrong on purpose. A car is marked sold after its video was posted. A price is corrected after the render. A photo is missing for one row. Watch what your pipeline does with each, and fix the gaps while nothing is at stake. Rehearsal is cheap, and the first real bad day is not.
What this will not do for you
A few honest limits. Templater is not a real-time system. It renders files, so there is always a gap between a change in the inventory and a new video, made of the time Bot takes to notice, the render, and your delivery. If your stock can sell within minutes and the ad cannot be retired as quickly, the pipeline will lag, and you need a faster retirement path than any rendering tool provides.
It will not decide what to advertise. The template shows whatever the row holds. If a photo is poor, the video is poor. If a price in the sheet is wrong, the ad is wrong at the speed of automation. The sheet needs an owner, and the owner needs to be someone who knows what is on the lot.
It will not manage your ad accounts. Budgets, audiences and placements belong to your ad platform, and we do not claim to touch them. We make the video and, with scripts, we can hand it on.
Last, a template is a fixed design. If your ads need to look different for different kinds of stock, such as a truck and a sedan, build more than one template. Templater can switch between template projects per row from the same data source, so a column can pick the right design for each item. Keep the count small. Every template you add is something to keep correct when your branding changes. Our post on making hundreds of videos from data goes into building a pipeline like this.
The bottom line is that fast-turnover inventory needs the ad to follow the data, not the other way around. Make the inventory row the single source of truth, let a template turn each row into a video, carry the item’s identifier through the file and the ad, and make retirement as automatic as creation. Test the awkward cases in rehearsal, especially the item that changes after it was rendered and the one that sells at ten in the morning. A pipeline that handles those will keep your ads honest, and an honest ad is the only kind worth paying to show.
Teams with fast-moving stock ask us a similar set of practical questions. Here are the short answers.
Frequently Asked Questions
How do I get a new inventory item into a video automatically?
Put the item in a data source Templater can read, such as a Google Sheet, a tab-separated file or a JSON file or feed. In the Bot edition, Templater checks the data source periodically and renders rows marked ready. QUE accepts data through its API and a Zapier integration if several systems feed the pipeline.
What happens to an ad when an item sells?
Templater makes video files and does not manage ad accounts, so retirement belongs to your ad tool or a script. Carry the item’s identifier through the output file name and the ad, mark the row sold, and make that trigger removing the video and pausing the ad. Do not leave it to a weekly cleanup.
Can Templater push videos straight into an ad platform?
Our product pages list Vimeo, JW Player, Frame.io and AWS S3 as Projector destinations and do not list advertising platforms. You can write a script that uploads a finished video and returns its URL to the data source, and our custom development team can write scripts for you, quoted individually.
How do I handle an item whose price changes after the video was rendered?
Bot reads a render-status column to decide whether to render a row. Make the status change part of the same step that edits the price, and test in rehearsal how Bot treats a rewritten row before relying on it. Decide in advance who resets the status when something changes.
How long will a morning batch take to render?
It depends on video length, resolution, hardware, how much of the timeline changes, effects and template build. Render one representative video on a machine like your production one and multiply by the count. Bot on two machines cuts total time in half, and Projector renders one video per core on a single machine.
How do I stop a wrong price or a missing photo from going out at scale?
Run a spot check across the day’s rows before the batch ships. Pro and Bot export stills or short clips of chosen frames from the merged compositions, so you can see long names, blank fields and missing photos. The sheet still needs one owner, because a template prints what the row says.
Do I need a different template for each kind of item?
Only if the designs need to differ. Templater can switch between template projects per row from the same data source, so a column can pick a truck design or a sedan design. Keep the number of templates small, since each one has to stay correct when your branding changes.
