Setup Considerations: anonymity and segmentation

What this achieves

Anonymity, targeting, and how narrowly you report are settled when a survey is created, and they cannot be renegotiated once people have answered. This page covers those decisions while they are still decisions.

Decision 1: anonymous or not

Anonymous is a switch on the survey itself, set when you create it. The interface describes it as stripping respondent identity from stored responses. It is on when the panel opens, so leaving it alone gives you an anonymous survey.

Option	Description
Anonymous	Respondent identity is not kept with the stored response. The respondent sees an Anonymous badge on the survey and is told on submitting that their response was recorded anonymously.
Not anonymous	The survey carries no such badge, and the respondent is told their response has been recorded.

Be precise about what anonymity covers here, because employees will ask. It covers the stored answers. It does not mean nobody knows the survey went to you: every invitation is personal and single-use, which is what lets the product report a response rate at all and chase the people who have not answered. A survey where nobody could be told apart from a non-respondent could not do either.

Example: an anonymous quarterly pulse at HC Corp UK Ltd can show that 24 of 34 people in Engineering responded, without any answer being attached to a name.

Decide this per survey and then say the same thing to employees that the product says to them. The one thing that destroys a survey programme is a promise the tool does not keep.

Decision 2: how widely to target

Target Scope offers three shapes.

Option	Description
Whole Company	Everyone. The widest sample and the least steerable.
By Department	One or more departments. Use it when the question is about a specific part of the business.
Custom (pick people)	A list of individuals you choose.

Targeting narrowly and reporting narrowly compound. A survey sent only to one small department, then read at department level, is a small number of identifiable people twice over.

Example: a survey at HC Corp UK Ltd sent only to the People department reaches six people. However the results are presented, everyone in that department can work out roughly who said what.

Decision 3: sampling

Sampling surveys only a random subset of the target population, at a percentage you set.

Option	Description
Off	Everyone in the target scope is invited.
On	A random share of them is invited.

Sampling reduces survey fatigue on a frequent pulse. It also shrinks every segment you later try to read, and it makes two waves less comparable because they did not ask the same people. Use it on a high-frequency company-wide pulse; avoid it on anything you intend to segment.

Decision 4: recurrence

Recurring launches new waves automatically on a schedule of weekly, bi-weekly, monthly, or quarterly.

The value of a survey programme is the second and third run, so recurrence is usually right. The risk is the opposite of the usual one: a recurring survey keeps arriving whether or not anyone has acted on the last round, and a workforce that has answered four times with nothing visibly changing stops answering.

Example: a quarterly recurring pulse at HC Corp UK Ltd gives four comparable points a year. A weekly one gives fifty-two, of which the last forty will have a response rate nobody wants to show anyone.

Match the frequency to how fast you can actually act, not to how often you would like a number.

Decision 5: the smallest group you will report

This is the decision the product does not make for you. It will decline to break results down when there are too few responses to do so, and it says so in plain terms on the segment view — but the threshold is not shown on screen, and the sizes you would actually be comfortable publishing are a policy question, not a technical one.

Settle three things before the first survey closes:

Decision	Why it has to be now
The minimum responses you will report on for a segment	Deciding it after you have seen the results means deciding it around a number you already want to publish.
Which dimensions you will report at all	Department, grade, tenure, and gender are all available. Some of them are far more identifying than others at your sizes.
Who sees results before anyone else	Results have an audience order. Decide it once rather than per survey.

Example: a rule at HC Corp UK Ltd of not publishing any segment with fewer than a stated number of responses is a rule you can apply to Engineering at 34 and to Finance at 9 identically, without arguing about it each quarter.

Write the rule down and apply it to the segments you like as well as the ones you do not.

Decision 6: question categories

Questions on a template carry a Category, and categories are what group results into an engagement score you can read by theme rather than as one number.

Set them when you build the template. Retrofitting a category onto a template that already has responses behind it changes what the score for that theme means across waves.

What is expensive to change later

Decision	Cost of changing later
Anonymity, once responses exist	Cannot be changed. It governs how the responses were stored.
Target scope or sampling, once launched	Cannot be changed for that survey. Create another and lose comparability.
The question set, between waves	High. Two waves with different questions are two surveys, not a trend.
Question categories, with history	High. The category is what the engagement score rolls up by.
Recurrence frequency	Low. It changes when the next wave goes out.
Your minimum reporting size	Low to set, high to change once you have published under it.

What happens next

Build the template first, because it is the reusable part and the survey is the disposable one. Then create a survey from it and target it. Once responses exist, the decisions above are settled for that survey.