"Should we show product thumbnails in our typeahead search widget?" "How many suggestions are too many?" “Should we pre-populate our widget with trending products and searches?” If you've ever debated these questions with your team, you're asking very good questions.
Visitors who search are typically closer to conversion — they know what they want. Autocomplete can help them achieve their mission faster and with less friction.
How you design and deliver your autocomplete experience makes or breaks its revenue-boosting powers. Let's explore five autocomplete A/B testing strategies that can help you quantify conversion lift and settle any internal debates with data.
Pre-Merchandising
Many merchants choose to pre-populate their autocomplete widget with trending terms and products to inspire shoppers.
These suggestions disappear once the user starts typing a query and are replaced with search matches.
Autosuggest widget with pre-populated suggestions
While this is a popular practice, it’s important to validate whether this tactic is the best pattern for your customers.
Visitors that touch the search box already have a query in mind and a task to complete — they are in ‘search’ versus ‘browse’ mode. Showing unrelated suggestions may be distracting enough to interrupt their task and lead to premature abandonment.
An A/B test that compares measures clicks, conversion, and revenue can validate whether showing suggestions or keeping search clean works better for your business and customers.
Should pre-merchandising win this experiment, consider testing different recipes within your widget UI.
Widget Content and UI
There are infinite ways you can style your autocomplete UI and populate it with suggestions. Many widgets contain a mix of dimensions (suggested queries, categories, brand and product hits). Some even include banners and other messages.
It’s a great idea to treat your widget like any landing page and test strategies against each other.
Show / hide product hits
For example, showing product hits provides instant gratification, but may be perceived as your full selection. Consider testing a widget with and without product hits (and it’s always a good idea to include a ‘Show all’ link with your product results).
Autocomplete widget with search terms and product hits
A widget without product hits steers shoppers to larger sets of results versus a single product, and may enable a richer browse-to-click experience with positive impact on revenue.
Autocomplete widget without product hits
Number of results per dimension
How many results you display per dimension may also impact user experience. Showing fewer search terms, category matches or featured products may not provide enough inspiration. On the flip side, a simpler menu is easier to scan with less clutter — especially on mobile.
Consider testing different recipes of suggestion settings.
Autocomplete widget with a 2 x 2 x 4 result setting across dimensions
Category thumbnails
You can also test showing image thumbnails with category or brand suggestions to provide visual context.
Autocomplete widget with category thumbnail images
Facet scopes
Facet scopes serve as a pre-filter to narrow results before the user submits their query or selects an autosuggestion.
Department scopes
If your catalog spans a few top-level departments such as Men, Women, Kids and Beauty, giving searchers a way to select a department or facet before submitting their search will return a more focused and relevant results set.
Autocomplete widget with department scopes
Unfortunately, as with all tabbed elements, they can get lost in the clutter. In addition to testing with-and-without scopes, you may want to test them in combination with simple versus complex widget population.
An autocomplete widget with department scoping and a large list of suggestions
Sale and price scopes
You may also try sale and price scopes in your widget. For example, SSENSE provides a ‘Sale only’ checkbox, and CVS includes price filter chips.
Autocomplete widget with Sale only checkbox
Autocomplete widget with price filter chips
Next Steps
Keep in mind that your autocomplete experience is completely different on mobile versus desktop. This is one touchpoint where it’s valuable to run A/B tests separately across devices to deliver the best experiences for all visitors.
And autocomplete is only a single touchpoint within your search and discovery experience. For more A/B testing insights and strategies from the Constructor experimentation team, check out the Experiments Blog.