Humans are still the best writers

Sometimes people ask us, “Since Textio has such a huge data set, can’t it just write job posts for you? Wouldn’t it save time just to automate the whole process?”

The answer is simple: Automation might be faster, but your writing wouldn’t perform nearly as well.

Whatever you’re writing, you have a unique voice. When that voice comes through, you see more engagement, better response rates, and happier readers. Your writing performs better with Textio’s data and guidance helping you along the way. But it only works because you’re in charge of your writing decisions, not some algorithm.

In fact, your point of view is so important that, as with all the best learning loop products, Textio can’t work as well without that point of view included. Across Textio’s augmented writing experience, your decisions make the engine better.

Put another way, no one else’s voice works as well as your own. In the case of job posts, it is especially easy to see. If every job sounded exactly the same, no job would stand out.

Last year I wrote a piece on how learning loops are defining the next generation of enterprise software. Now I want to walk through learning loop design across Textio, and point out a few places where your opinion will shape (or in many cases, is already shaping) Textio’s augmented writing experience.

The best business writing rarely starts from scratch

If you’re like most people, much of the serious writing that you do at work begins from work that you already have around. To write this week’s status mail, you pull up the one you sent last week and edit it. Next quarter’s product roadmap deck starts from the same template you used this quarter.

In particular, you rarely start a new job post from scratch. If you use Textio, you typically begin in your team’s job library, where you can see all the jobs that have been shared across your team, along with their current Textio Scores.

Let’s say you’re working on a new Software Engineering role. You search for other similar roles in your team job library. When you find one with a good score that’s similar to the one you need to post, you select it as a starting foundation for your own job.

Like any good learning loop, Textio learns from your preferences and actions. You haven’t started writing yet, but just by searching for a job, you’ve made your subsequent experience better.

The job titles that you search for most often give Textio an idea of where hiring is hottest for your team — and where the outcomes feeding the learning loop are changing the most rapidly. The person whose job posts are most frequently selected by coworkers is a writer whose patterns might get extra attention from Textio over time. Are you consistently choosing starter job templates that score below 90? That’s a good indication that you are looking for factors other than those weighted most heavily by Textio today.

All of this taken together makes it possible for Textio to improve the selection of jobs that are highlighted for you as possible starting points, and which writing guidance patterns may be particularly interesting for your company. That adds up to better predictions and guidance over time, and you haven’t even started writing yet!

The language that works is local and personal, not universal

Language is not one-size-fits-all. People speak and write differently depending where they’re from. Different industries and communities use different jargon. The language that resonates with you may vary depending on your particular background. Any good augmented writing experience needs to take this into account, or its guidance just won’t work.

In the case of job posts, your job type and location have significant impact on the language that works to attract the people you want. Whether you’re starting from an existing foundation or writing a job from scratch, Textio learns from the job title, category, and location that you specify while you’re working on the listing.

The hiring language that works to hire an engineer in Seattle is different from what works to hire an accountant in London. Textio has global customer data to provide writing guidance across industries and locations, and this creates powerful results. But sometimes job classification is complicated.

For instance, teaching Textio the difference between Engineering roles and IT roles is not straightforward. This is because the people writing the job descriptions are actually shifting the classification of roles over time.

Every time you tag a job post in Textio, you provide your opinion about the classification of the job you’re working on. That helps Textio build a more robust data set of job classifications. With your input, the predictive engine produces more and more detailed models that make sure that you get exactly the right guidance for your job, whether it’s IT, Engineering, or something else entirely.

Even though you haven’t started writing yet, Textio is already learning from your opinions.

Where you look changes what you see

The language that resonates with readers depends on their industry, geography, and other demographics. Just as important, the language that works best for you as a writer depends on what you’re trying to achieve.

Once you’ve selected your job template and updated it with the right supporting information, Textio scores your job for the first time. Chances are your next move will be to hover on highlighted phrases or click on Textio’s sidebar to see guidance. But how do you decide where to hover or click first?

Because you’re writing with a point of view, not every suggestion is equally relevant to you. Perhaps your biggest concern is time-to-hire, in which case you want to get rid of as many orange phrases as possible. Or maybe you’re more focused on removing gender bias from your job, and you begin by hovering over blue and purple phrases to understand your existing tone. Or maybe you just want to finish the editing process as quickly as possible, so you click on the top sidebar item to see what you need to change first.

After you’ve used Textio for a few days, the platform gets a good sense of the guidance you tend to care about. Over time Textio can surface the pieces that you will be most interested in — the guidance that is most relevant for your particular style and goals.

And you still haven’t even started writing yet.

Time to write!

It’s one thing to know who will respond to your writing ahead of time. It is a harder and much more valuable thing to know why, and how to fix it. Even once you’ve found the patterns that work in the market today, you can be confident that they’ll change rapidly as the market shifts. Any good augmented writing product must be supported by constant learning loops.

In Textio, once you’ve clicked and hovered around the screen, you have a sense of how people will respond to your job as it’s currently written. Now comes the fun part. Let’s actually make your job post compelling enough that great people will apply!

Textio’s editor learns from how you respond to guidance. Every time someone chooses one suggestion from among a set of possibilities, Textio counts it as a vote in favor of that alternative. In order for a suggestion to stick around in the product, it’s not enough for it to perform well with readers of job posts. It also has to get selected by writers using Textio. People reject suggestions that aren’t appropriate for the context where they show up.

Sometimes Textio shows you an experimental suggestion. This is where Textio has reason to believe that the suggestion will perform better than what you’ve written, but the data isn’t yet strong enough to be definitive. After you’ve selected the suggestion, Textio asks you how it sounds and how to improve it.

Textio also learns from the suggestions you don’t take, and the phrases you never hover over. When a particular pattern is rejected by a lot of people in the platform, Textio needs additional statistical data to continue recommending it.

It’s not just the words you choose, but how you put them together

Individual word choice can be impactful, but except in really egregious cases, no one word or phrase makes or breaks someone’s response to your writing. (In fact, it turns out that people may not even care about your grammar mistakes at all!)

However, the patterns across your collection of words matter a lot. Not just linguistic patterns, but also visual ones.

How people respond to your job post depends partly on what it says, but it’s also important what the post looks like visually. Textio offers guidance on your formatting as well as your content, and the platform learns from the formatting choices you make.

In your job posts, for instance, the way you use bulleted content, how you style section headings, and the length and format of your equal opportunity statement all change who responds to the post. Textio anticipates the formats that will get the best engagement for what you’re writing, so the editor proposes formats right as you’re typing.

As a writer, you may have a different point of view on how you want your post to look. Sometimes you will even intentionally break with Textio’s guidance for a particular effect. The places that you reject Textio’s guidance are vital for the engine’s learning loop, especially if your job ends up performing well after you post it.

These cases are rare, because writers that follow Textio’s guidance create job posts that perform better. But while uncommon, the places that writers break with Textio’s guidance are as valuable as gold. When you reject platform guidance, it creates a unique opportunity to learn from whatever language you’ve chosen to use instead.

This is especially true if you are a strong writer, which in Textio terms means that your jobs typically score above 90. The strongest writers of any marketing content — and let’s face it, what’s a job post if not marketing for your company’s recruiting efforts? — are often ahead of the curve in how they write. They play with language and experiment, which makes them critical partners for Textio’s learning loop.

When a group of strong writers starts trending on a new pattern that Textio isn’t recommending, that’s an opportunity for the platform to measure and learn.

Learning loops need communities

To build a strong learning loop in software, you need all kinds of people using your product. More than that, you need to design your experience so that just using the product is tantamount to sharing an opinion about it.

When your platform collects enough opinions, then you’re in a position to see patterns within them. No one person’s opinion makes a learning loop. But a whole community’s worth of opinions is more powerful than any one perspective on its own.

In Textio’s case, the most important part of the learning loop is your voice — how you write, and how real people respond to you. You have clear goals when you write, and you are creative in service of them. Textio has a lot to learn from you!