I have a large pool of telephone agents using a custom built piece of software to track the calls. Due to the high pressure of having a caller on the line agents often rely on old methods for completing tasks and it can greatly increase the length of the call.

The problem

Our new piece of software includes one key search tool based on about 6 different fields. They will often populate 2 to 4 of these fields, hit search, and get no results. Instead of manipulating the fields to do things like a partial name search etc they jump to an older terminal system where they worked for years. Ironically when there they do the same partial search I just described. In fact, they will try many search combos until they find one that works.

When on a call with a client the agent may misspell complicated names, or type in an incorrect birth date etc. The whole human to human to computer interaction is prone to data entry errors. As such mistakes are common and this situation occurs frequently.

Question How might the interface coach users into trying partial searches. Most people want to blame the users but I feel the interface should better guide their behavior. Ideas I have include:

  1. Provide a message describing what to do such as: "Try searching without a birth date"
  2. Have the system automatically do partial searches and provide a list of "possible matches" if no exact match was found
  3. Retrain the users demonstrating how to "best" use the tool

I prefer #2 but it is by far the most technically complex.

So how can one encourage users to manipulate their search input to find matching records?

Edit The more I have considered this I came up with two additional options:

  1. Have the search be real time as they enter search criteria. This would enable the user to stop when it is visible that their search is successful.
  2. Have the system automatically perform alternate searches based on portions of the data in an effort to find matches. Display them as "possible matches".

I am now leaning towards option 4 as it improves performance time beyond simply stopping users from jumping systems. It would enable them to stop entering data when the match is found.

  • 1
    This is unfortunately too broad a question for this site, because it's impossible to answer without first understanding why users are flipping to the old system, and how they use that system. I would start by interviewing and observing users to figure out why they are switching, working out how you can provide a superior interface to provide positive value/incentive to switching, and then looking at change management process (e.g. mandatory training) to help ease the change.
    – tohster
    May 11, 2015 at 17:30
  • 1
    Actually I have done extensive observation. Users try the new system, but after a single try that nets zero results they abandon ship and go back to the old tool. It is simply a matter of preference and bias towards the older system they know. The question is how to entice them into trying the same search patterns on the new system.
    – PMcNeil
    May 11, 2015 at 21:01
  • 2
    Could you provide a mockup of the current search form? May 12, 2015 at 4:08
  • If the new systems performs the same tasks in the same way as the old systems then the simple answer is to remove the old system. Alternatively, run a 'training' session for managers and/or team leaders and let them pass on the knowledge May 12, 2015 at 14:19
  • Andrew that was indeed our very first idea, but it is impossible as it is a component of a larger system that we simply can not phase out.
    – PMcNeil
    May 12, 2015 at 15:02

2 Answers 2


I don't totally understand the difference between old and new here. I do understand that it's not always acceptable to post visual reference of proprietary systems. Here's a shot in the dark:

Providing cues

In app cues are critical with this kind of change. Providing some form of contextual help when results fall below a reasonable level will go a long way. Something as simple as "Try using less specific terms to expand your search" in a clear way can be a big encouragement to users.

If you can provide real-time results while the user types without a big performance hit, that will help as well. You could also try to expand your search algorithm to find related searches to solve common pitfalls.

Here's Google's solution with reasonable language. You'd want to integrate it in a scan-able, prominent notice at the top of your search results area.

Your search - "Give me exactly what I want including these astrophysics misspellings of kwark and ... - did not match any documents.


  • Make sure all words are spelled correctly.
  • Try different keywords.
  • Try more general keywords.
  • Try fewer keywords.

Bridge the learning gap

Is there a reason you can't provide a switch to change the search tools to something more familiar in the new system? Your new way may be functionally better, but that's irrelevant if users don't take to it. There's something about the old tool that works for them, whether it's just learning or something more nuanced.

As the opportunity presents itself, your support or education team should provide workshops, knowledge base articles, and/or blog posts on the value of the new search format. Along the way, you'll probably refine the feature and you can reveal that as well to generate interest.

The bottom line is, training people on internal tools is often slow and painful. Just because you design a better system, doesn't mean it will gain adoption right away.


Groundwork first:

To streamline the process and increase efficiency you need to do some groundwork first to address the root issues, search mechanics will follow. You can see the below as either as a three step process or separate work streams that you might need to focus on:

1. Focus on the data:

You need to assess the quality of your data as this is crucial to any operational and transactional processes you may have. You should document and benchmark your results and go through a cleaning exercise, removing duplicate records if they exist, flag outdated or incomplete records and standardise the data sets. This exercise should in principal be focused on prioritised list of variables that are most likely to affect the accuracy of you search results.

2. Understanding where things went wrong:

Based on the above try to understand which factors actually led to poor quality data, Most likely this will be another area where you will need to dedicate some effort and time to understand how your users operate and where the system failed them. Improving data capture processes inline with your users behaviour and providing opportunities to complete and verify the records would be an optimal outcome here.

3. Search mechanics

You can now turn your attention to search mechanics. Having real time feedback will definitely improve the UX and increase user confidence in the system, particularly given the legacy issues you stated.

An idea worth exploring here is the use of some form of faceted search, users can search for a keyword using incremental/typeahead search and scan a list of results, a left side navigation panel incorporating all the facets and their count is displayed. This is particularly useful because it does not only give users useful search results but also provides insights into data structure at the same time.

Good luck!

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.