Mad Libs madness

The original Mad Libs® were a series of books by Roger Price and Leonard Stern, published in the 1970s [1]. The books contained sentences where children had to ‘fill in the blanks’, creating funny stories (and maybe also learning a little grammar as they go).

More recently, the term ‘mad libs’ has been appropriated to refer to an alternative, prose-like method for designing forms. Rather than the standard, vertical, ‘question: answer’ approach, with mad libs forms fields are placed wherever the answer fits within a piece of narrative text.

In this article, we’ll show that most mad libs forms result in a much poorer user experience than standard forms. The main reasons for this are:

  • with mad libs forms, users often have to expend greater effort to read, interpret and answer questions;
  • mad libs forms are harder — and thus slower — to move through (especially for those using assistive technology); and
  • in a mad libs form, it is easier for users to miss questions altogether, and/or lose their position.

These findings are based on research data, including some Formulate testing results being released for the first time. But before we get into the detail, it’s worth having a little background.

The popularising of mad libs forms on the web

A couple of leading lights in the web industry really got people excited about mad libs forms. First, respected web developer Jeremy Keith used it for the sign-up form for his podcast-creation service, Huffduffer (Figure 1).

Figure 1: Huffduffer sign-up form. Click on this (and all following images) to see a larger version.

Then, Luke Wroblewski redesigned a car dealer contact form for (Figure 2), and shared the resulting improvement in conversion of 25-40% across the board.

Figure 2: Before and after screenshots for the dealer contact form.

Soon enough, the idea of fill-in-the-blanks was the next big thing for online forms. Somehow, it made forms seem more ‘human’. But is the mad libs approach really a good idea?

Mad libs form fails an A/B test

Patrick McKenzie, owner of software development business Kalzumeus, was being asked about mad libs forms so much, he decided to A/B test them using the registration form for his own Bingo Card Creator service. You can see his original form in Figure 3 and the mad libs version immediately below it in Figure 4.

Figure 3: Screenshot of the original Bingo Card Creator sign-up form.

Figure 4: Screenshot of the mad libs version of the Bingo Card Creator sign-up form.

In Patrick’s case, the mad libs approach led to a 22% decrease in conversion. While it is true that the mad libs approach wasn’t the only change between the two versions, it is the most significant difference. Therefore, it’s likely that a fair proportion of the negative result is attributable to using the mad libs style.

Patrick’s conclusion is a good one: test the mad libs approach before adopting it. As they say “individual results may vary”! But wouldn’t it be good to know whether the mad libs approach is worth considering, even before any design work commences?

Formulate’s research shows standard approach usually better, and why

In 2009, we were invited to redesign an online car insurance quoting form, from scratch. Such license is pleasurable but rare, so we took the opportunity to think as laterally and openly as possible. Through this process the mad libs idea — which hadn’t even surfaced yet in the web world — came around [2]. We designed a mad libs-style prototype of the [multiple screen] form, as well as a version that used the standard vertical approach. Snippets of the two prototypes can be seen in Figures 5 and 6.

Figure 5: Partial screenshot of the car insurance form using the standard approach.

Figure 6: Partial screenshot of the car insurance form using the mad libs approach.

The research participants were professionally recruited and together were representative of the target market, which was essentially the Australian adult population. In the test sessions, participants were asked to use the prototypes to get a quote for car insurance, with the facilitators randomly assigning either the mad libs or the standard approach. While observing this attempted usage, we gathered plenty of information about how mad libs forms generally compare to standard forms.

Overall users preferred, and were more successful with, the standard approach. For us this wasn’t a massive surprise, as we also found mad libs forms more difficult to design (more on this later).

Successful mad libs forms depend entirely on narrative

Mad libs forms aren’t just about turning questions into sentences within a paragraph: in order to fill in the blanks, the surrounding text must provide sufficient context and clarity. After all, the narrative text is all that users have to tell them what sort of answers are required.

For example, the sentence

“I am looking for ___ insurance”

suggests that a type of insurance is what will go in the blank. Contrast this with

“I am looking for ___”

where any number of different answers could plausibly complete the sentence.

Similarly, with a standard form, the user gets to read the whole question before having to answer. On a mad libs form, however, the blank often has to go mid-sentence, meaning the user encounters the form field before they have all the information to be able to answer it. In turn this means at least part of the question has to be read twice.

The following statement is a good illustration of this problem:

“The car ___ been insured for the whole time between when I bought it and today.”

The text before the form field (“The car”) is not enough for the user to know what sort of answer to put in the form field. They must read the whole sentence, then go back to the start again, in order to be able to answer.

Sometimes it’s just downright hard to turn a prose sentence into something where a user can fill in the blank. Take the following question about insured value, from the standard form (Figure 7).

Figure 7: Question asking about the insured value, when the form uses the standard approach. This is a particularly difficult question to ask in a mad libs context, as Figure 8 shows.

How would you reword this question so that the answer options are included within the mad libs sentence? It’s pretty difficult to find a solution that doesn’t completely mangle the English language. In the end, we went into testing with the sentence shown in Figure 8:

In the case of a total payout (e.g. if the car is written off), I want the payout amount to be ___ value.

The prose version isn’t nearly as easy to read and understand as the standard version.

Figure 8: Collecting the same information as the question in Figure 7, when the form has a mad libs approach.

The narrative approach also doesn’t really cater for questions that have multiple answers. This quote process included collecting information about all of the car’s non-standard accessories and modifications. On a standard form, this would mean a list of multiple-choice (i.e. checkbox) options. There isn’t really room for checkbox answers on a mad libs form, nor do they really fit within the model. We could use a multi-select drop-down, but these aren’t as widely used as the more basic form controls and consequently, many members of the general public struggle to interact with them.

Overall, the visual design of the standard layout delivers more to users

Mad libs forms often require much less screen real estate — because questions appear next to each other — but this can be at the cost of usability.

Mad libs fields are hard to find

The standard approach — where fields are aligned vertically — provides a linear path to completion. Moving from one field to the next is very fast because the user can predict where the next field will be. With the mad libs approach, the user has to search, and navigate to a different place, for each field.

Easy to get lost in a mad libs form

A user of a standard form can see, at a glance, where they are up to. It’s much more difficult to visually distinguish completed fields from surrounding text in a mad libs form, making it much harder for users to keep track of where they are.

Easy to miss things in a mad libs form

With our mad libs form we noticed an interesting phenomenon related to the fact that fields can appear anywhere along the horizontal. The question in Figure 8 was often missed by research participants, presumably because they didn’t see the field out toward the right hand edge of the screen. The vertical alignment of a standard form alleviates this problem altogether.

Difficult to include help in a mad libs form

The flowing layout of mad libs forms also makes it harder to incorporate help, tips and other supplementary information.

With the standard layout, we have many options available for incorporating text besides the questions. We can put tips, links and help icons under or beside the question and/or the answer field. One way or another it is easy to make sure that users get the information they need, ‘just in time’.

With the mad libs style, we have to either interrupt the flow or put supplementary information away from the question that it relates to. Both options significantly compromise form usability.

Making accessible mad libs forms is a bit of a nightmare

Finally, the flowing layout of mad libs forms also makes it hard to develop the form accessibly.

This has been demonstrated by accessibility experts Simply Accessible, who have done extensive research into screen reader behaviour with mad libs forms. If you’re really keen, you can check out all their mad libs forms variations and the resulting treatment by popular screen readers.

There are many reasons to avoid mad libs forms

In summary, the disadvantages of mad libs forms are that:

  • users often have to expend greater effort to read, interpret and answer questions;
  • mad libs forms are harder, and thus slower, to move through;
  • it is easier for users to miss questions altogether and/or lose their position;
  • the topic has to lend itself to data collection via narrative text;
  • questions have to be very carefully worded to provide sufficient context and clarity;
  • there isn’t an ideal way to handle multiple choice questions;
  • it’s difficult to suitably incorporate help, tips and other supplementary information; and
  • mad libs forms are likely to be inaccessible for some users.

Mad libs suited to certain situations

These problems do not mean mad libs forms should be ruled out altogether. In fact, we went for a final design for the car insurance quote form that used a combination of mad libs approach for the first screen (Figure 9) and a standard approach for the remaining screens.

Figure 9: The first screen of the live car insurance quoting form, using a mad libs approach.

We adopted the fill-in-the-blanks approach for the first screen because it met the conditions under which a mad libs approach can work:

  • it was short (small number of straightforward questions);
  • the questions themselves suited a narrative wording;
  • minimal need for help, tips or other supplementary information; and
  • all questions require text or single choice answers only.

We also believe mad libs forms are suited to contexts where a sense of novelty, fun or playfulness is valuable. A great example of this is the Bureau of Communication website, which allows users to create correspondence by filling in the blanks (see Figure 10). This really wouldn’t be much fun at all if the standard form approach was used.

Figure 10: One of the fun Bureau of Communication mad libs forms.

Coming back to our car insurance example, we wanted potential customers to feel comfortable starting a process that is often all too daunting, and thought mad libs might achieve this. For then switched to the standard layout for the remaining screens, providing optimal usability for the more complex content while also conveying a sense of professionalism more appropriate for later stages of a quoting process.

The ‘combination’ design tested well in later rounds of research and has been performing well live. Having said that, the first screen does still suffer from the problems that beset mad libs forms, like questions being missed and the wrong type of data being entered into open text fields. This shows just how hard it can be to create successful mad libs forms.

Some people have incorrectly attributed the term to a short-lived television game-show by the same name. The show was from the United States and aired in the late 1990s. In it, children solved puzzles by finding the right words to fill in the blanks (see The books pre-date the game show.

The credit for thinking of the mad libs approach goes to our client at the time.

Mad Libs book cover picture courtesy of and copyright Price Stern Sloan.