How to avoid these research problems
By Angela Sinickas, ABC, Diva of Diagnostics
Ready to measure your communication success? To avoid rookie measurement mistakes, don’t:
1. Forget to measure until after communicating. One of the best ways to prove a campaign or channel was effective is to compare baselines before and after the communication. But it’s easy to forget to measure the “before” baseline. Here’s how to fix:
- Check to see whether any locations did not use your intended campaign as it was supposed to be used. Use these as your “control groups” against your “pilot groups” (where it was used correctly).
- Ask the audience to gauge what impact your communication had on their knowledge, attitudes or behaviors.
- Include demographic questions in your existing audience attitude research that identify which communications your audience was exposed to, and the difference they had in impact.
2. Think only surveys can provide hard data. There are many ways to get data:
- Tap into operational measures your colleagues are already tracking (e.g., health and safety, quality assurance.)
- Quantify how you spend your time. How much is spent being strategic or tactical? Proactive or reactively fixing problems? How much is spent in supporting various units of your organization and how does that compare with their overall contribution to the bottom line?
- Measure the reading grade level of your writing to compare it with the reading ability of your target audience (online tools can help).
- Quantify the volume of different types of communication. Check that there’s not too much from different departments and that it doesn’t come all at once.
- Track the communications’ content against your key messages. If company goals aren’t mentioned in the flagship communication vehicles, employees won’t understand or believe in those goals.
3. Use focus groups to track change. If you can’t get the budget for a survey, don’t waste your time using focus groups as “proof” of improvements — the results won’t stand up to scrutiny. Here’s why:
- Not enough people are involved for project results to represent the entire population (see mistake No. 8).
- Participants aren’t selected randomly, so can’t accurately represent the whole organization.
- Focus group facilitators are human and, even if scripted, cannot repeat the exact same way of asking questions from session to session.
4. Hand-pick focus group participants. If you pick people because they’re vocal or are opinion leaders, you’ll only hear what you’ve heard before. For more representative research results, listen to a cross-section of your audience to hear from people who don’t usually cross your path.
Beware of volunteers; that will generally result in bias. Also, if supervisors are allowed to make replacements among invited participants, yet another kind of bias will be introduced.
I recommend selecting participants at random from the subgroups you have identified as requiring their own focus groups. For example, let’s say you need a focus group of manufacturing employees working a third shift in a particular location. If you have 100 employees who fit that description, you would select every sixth name on an alphabetical listing, providing a list of about 16.
5. Ask survey questions that don’t lead to actionable results. Think about the aspects of your communication program that could be changed, and ask questions that will let you know if your audience thinks they should be. Work backwards to phrase questions in a way that will suggest actionable answers.
For example, let’s say you ask people to agree/disagree with the following statement: “The newsletter should continue to be published once a month.” If they disagree, you don’t know whether to increase or decrease the frequency, or to eliminate it altogether. It would be better to ask people to select their ideal frequency from a list of options (weekly, monthly, quarterly, never, etc.).
Be sure each question asks about only a single item. For example, don’t ask people how well they understand their pay and benefits program. They may understand one better than the other, but this question doesn’t allow them to tell you which one. Then you won’t know which problem to fix.
6. Automatically include a “No opinion” option in surveys. Before you offer a neutral, “No opinion” response for a question, be sure that it’s a meaningful option. Often, “Not applicable” may be a more appropriate choice, especially for questions that are not even asking for an opinion, but rather a description of how often something occurs.
Be aware that respondents may choose “No opinion” on a sensitive issue when in fact they hold a negative opinion but are fearful of being identified individually and punished by their managers.
Remember that “Neither agree nor disagree” can be a better option than “No opinion.” This is particularly true when you ask respondents to pick one option from a scale of responses running from “Strongly agree” to “Strongly disagree.” A better midpoint between the two is “Neither agree nor disagree,” rather than “No opinion,” which can be misleading.
7. Over-value online “pulse” surveys. It’s very easy to post a survey question on a website or intranet home page and get replies almost instantaneously. It’s more difficult to interpret what those results really mean.
Most communicators realize that the results apply only to the segment of their audience with easy online access. But they don’t realize that different subgroups within that group are more likely to see and reply to the question because of their more regular use of online resources. This isn’t obvious because most online “pulse” questions aren’t followed by a series of demographic questions.
Yet, the percentage of employees accessing the intranet at least weekly varies by department and job level, which would result in some groups being under- or over-represented in a weeklong pulse survey.
Once you realize there’s a bias in your results, variations in pulse survey results over time can indicate real changes — but only among the groups as represented in your non-random sample.
8. Fail to use a statistician to select a random sample. Deciding a sample size is best done with the help of a statistician who knows not only how to sample an audience, but also how to pick a random sample. If you don’t do this carefully, some executive with even a little background in statistics may invalidate all the results of your survey when you’re finished. Most organizations have in-house resources to do this in the organizational development or market research groups.
The number of surveys to distribute depends on how many people are in your total audience and what rate of response you are likely to achieve. If you have a relatively small number of target respondents, you might need to send out surveys to everyone.
If you have over 2,000 people in the audience, you would need only 500-600 completed surveys, assuming the respondents accurately reflect the demographics of the entire group. If you expect to have a 100% response rate, send out surveys to a random sample of 600 people. More realistically, if you typically have a response rate of 50%, you’d need to survey 1,200 people.
Most organizations also want statistical validity in comparing organizational subgroups against each other (locations, business units, etc.). This typically requires a much larger number of responses than 600, so that you have a sufficient proportion of each subgroup participating. For example, if you have two locations of several thousand employees each, you’ll need 600 responses for each location. Also, smaller subgroups may need a larger proportion of the group responding for statistical reliability than larger ones.
It takes a statistician to know how many people to sample, and then how to “weight” the data for a true company average if some groups needed to be over-sampled.
9. Accept a low survey response. The response rate on voluntary surveys generally varies from 5% for external audiences to over 80% for employees. Your response rate will depend on a number of factors:
- The length of the survey. The longer it is, the lower the response rate. To avoid this problem, you could break a long survey into two separate surveys, each administered to one half of your audience.
- Demographic questions. If there are too many, or they are all at the beginning, the response rate plummets. It’s better to start with questions you know your audience cares about so they’re engaged from the start.
- Previous experience. If your company has never reported back survey results, or made changes based on surveys, the response rate will continue to go down with each survey.
- Management support. If senior management makes it clear that they want to see the results and want good participation in all units, managers make sure employees are given time to complete it.
- Incentives. If every location with at least a threshold return rate receives a reward, that makes a big difference. Peer pressure gets results.
- How and where it’s administered. Surveys sent to employees’ homes get the lowest response rates, those via web, phone or email the quickest (although not necessarily in proportion to your audience subgroups). Group administration in small meetings will get the best rate. If you’re asking sensitive questions, such as intention to stay with the company, you might get a lower response rate electronically than on paper.
10. Misinterpret survey findings. People who are uncomfortable with numbers don’t always take enough time to see if the percentages kicked out by a software program make sense, or are reporting the full picture. Here are some typical misinterpretations:
- Accepting unqualified respondents. If you let people without knowledge of a topic evaluate it, the results will be more negative than the reality. Allow a “Not applicable” option.
- Using the wrong base for percentages. If a large number of respondents choose “Not applicable,” recalculate the percentages based on only those to whom it applied.
- Reporting percentages of percentages. If a question is asked only of those who responded a certain way to a previous question, be sure not to report the follow-up question’s results as applying to the entire audience.
Angela Sinickas, ABC, is president of Sinickas Communications Inc., an international communication consultancy specializing in helping corporations achieve business results through targeted diagnostics and practical solutions.