B2B Strategy

Read a Small B2B Customer Survey Without Turning Respondents Into the Whole Market

Six people report difficulty finding a document. The slide becomes “customers struggle with our delivery process.” The next version becomes “the market needs a new portal.”

Each rewrite adds a claim the responses did not contain.

Keep the unit, denominator, and recruitment method beside the observation. A small survey can produce a useful question for the next decision. Give that question the reach its evidence supports.

The response record below is entirely fictional. The reporting table is a proposed working method. It describes no customer base, company result, or measured market demand.

Start With the Invitation Record

Imagine a document-delivery team considering a test of its contents page. It asks existing customer contacts about locating a checklist in a recent delivery pack.

Account coordinators select contacts they can readily reach. The survey is voluntary. There is no random selection from a defined market population. This is a convenience sample of reachable customer contacts.

Keep these fictional counts separate:

Stage Accounts People or submissions Meaning
Invitation list 20 30 distinct people Every person receives one invitation; delivery is confirmed in this example
Any return 10 14 people Each returns one submission
Complete submissions 9 12 people Each reaches the final page; optional items may still be blank
Partial submissions 2 2 people One belongs to an account with complete submissions; one belongs to the tenth responding account
No return 10 accounts with no return 16 people No submission from those people

The complete and partial account counts overlap. Nine plus two does not mean eleven responding accounts. One account appears in both groups.

Fourteen of thirty invited people return something. Ten of twenty invited accounts have at least one return. Those statements answer different participation questions.

Call them invitation-return counts in this internal worksheet. Do not label the first fraction an AAPOR response rate without applying the relevant disposition definitions.

AAPOR's reporting guidance calls for transparency about the population, recruitment, survey mode, questions, and analysis. Preserve the invitation record so a reader can inspect those boundaries.

Define Complete Before Counting It

For this example, “complete” means reaching the final page after the eligibility check. It does not mean answering every optional question.

The primary item analysis below uses complete submissions only. That is an explicit reporting choice, not a rule that partial answers are worthless. Keep the two partial submissions in a separate record with their available fields.

Here, both partial submissions stop before the difficulty item. Neither supplies an answer to that question. Excluding them from that item removes no recorded difficulty rating.

One complete submission also leaves the difficulty item blank. Keep that blank visible. Do not assign it to the middle of the scale.

AAPOR distinguishes missing answers from explicit options such as uncertainty or refusal. In this worksheet, preserve separate values for blank, cannot recall, prefer not to answer, and not shown.

Before computing a percentage, name the eligible group and the actual answering group. A routed question can have a different base from the opening question. The export needs enough information to explain that difference.

Keep a Small Response Record Inspectable

The fictional complete submissions come from nine accounts, labeled A through I. The codes exist only for this example. They are not deidentified copies of customer records.

The item asks about difficulty finding the checklist on the respondent's most recent attempt. The original scale has five ordered choices plus cannot recall. For this summary, combine the two difficulty choices as “difficult” and the two ease choices as “easy.” Preserve the original choices in the working record.

Account Respondent role Summary answer
A Daily user Difficult
A Sponsor Easy
B Daily user Difficult
B Sponsor Cannot recall
C Daily user Difficult
C Sponsor Difficult
D Daily user Difficult
E Daily user Difficult
F Daily user Easy
G Daily user Cannot recall
H Sponsor Easy
I Daily user Blank

The twelve complete submissions contain six difficult answers, three easy answers, two cannot-recall answers, and one blank. Nobody selects the middle scale option in this invented record.

Write that reconciliation before making a slide. Six plus three plus two plus one equals twelve. If the totals fail, inspect the records before deciding what they mean.

Do not remove uncertainty because it complicates the chart. Two people completed the form without recalling the event well enough to rate it.

Put the Denominator in the Sentence

There are several valid descriptions of these same fictional counts. Each needs its base:

  • Six of twelve people with complete submissions report difficulty. One of the twelve skips the item.
  • Six of eleven people answering the item select a difficulty option. Eleven includes two cannot-recall answers.
  • Six of nine people giving a directional ease-or-difficulty rating select difficulty. This excludes two cannot-recall answers and one blank.

The first fraction is 50%. The second is approximately 55%. The third is approximately 67%. All three calculations use the same six difficult answers.

For this small internal report, lead with counts. If a percentage helps, attach the fraction in the same sentence. The denominator change deserves more attention than an extra decimal place.

Do not describe six of thirty invited people as the share experiencing difficulty. Six of thirty records the share of invitees who returned a complete submission containing that answer. The sixteen people with no return have no observed answer.

Use one primary base consistently in the main table. Put alternate bases in a note with their exclusions. Selecting whichever denominator makes the problem look largest would turn a reporting choice into an undisclosed argument.

Separate People From Accounts

Accounts A, B, and C each contribute two complete submissions. Every other account contributes one. Counting all people gives those three accounts more voice in the person-level summary.

That does not make their answers accidental duplicates. A daily user and a sponsor can report different experiences. Inspect both before choosing the unit for a business decision.

Account A contains one difficult answer and one easy answer. Calling the whole account dissatisfied would erase the disagreement. Account B contains difficulty and unavailable recall. Account C contains two difficult answers.

Five of nine accounts with complete submissions contain at least one difficult answer: A through E. That is an account-level flag. It is not equivalent to five accounts holding a shared opinion.

If the proposed analysis requires one response per account, define which perspective is relevant before selecting records. For a daily-use task, the eligible role might be the daily user. For a purchasing question, that rule could be wrong.

Do not select whichever response arrived first because it makes the spreadsheet easier. Record a deliberate rule and explain which answers it excludes. Keep the original person-level record for checking.

Read Role Differences as a Follow-Up Question

Eight complete submissions come from daily users. Four come from sponsors.

Among the eight daily users, five report difficulty, one reports ease, one cannot recall, and one leaves the item blank. Among the four sponsors, one reports difficulty, two report ease, and one cannot recall.

Those are descriptions of these respondents. They do not establish that daily users across the customer base experience more difficulty than sponsors.

The roles were not sampled to support that comparison. Their task exposure may also differ. This record does not measure it. Keep the apparent difference as a question to investigate.

For the fictional team, the next action is to observe eligible daily users locating the checklist. Ask which pack they opened and where they looked first. Keep that research separate from asking whether they want a portal.

Customer interviews can investigate context that the standardized answers leave unresolved. Use the interview to examine a task. Do not ask participants to validate a conclusion already written in a survey slide.

For a buying decision, a separate win-loss interview addresses a different event. Difficulty locating a document does not establish why an opportunity was won or lost.

Build an Evidence Table Before Recommending Work

Use one row per observation. State the population the row describes, the limit, the claim it cannot carry, and the next action.

Observation Population described Limitation Unsupported extrapolation Next research action
Six difficult answers among twelve complete submissions The twelve complete respondents One blank; two cannot recall; selected contacts Half the market struggles with document delivery Observe the checklist task with eligible users
Five of nine complete-response accounts contain difficulty Those nine accounts, under an any-difficulty rule Three accounts contribute two people; disagreement remains Five accounts reject the current service Examine the conflicting experiences within A
Five of eight daily users report difficulty The eight daily users with complete submissions One blank and one cannot recall; no representative role sample Daily users generally need a portal Investigate starting locations and task exposure
Sixteen invitees return nothing Those sixteen invitation records Their experiences and reasons are unknown Silent invitees are satisfied or dissatisfied Record the evidence gap; consider authorized alternate research

This table makes a proposed next step reviewable. It does not prescribe which project deserves funding.

Put the table beside a B2B business-case decision memo. Keep observed counts, interpretations, and spending assumptions in separate fields. A proposed prototype can be justified as an investigation without claiming proven demand.

Name Selection and Nonresponse Limits Without Inventing Their Direction

The invitation list contains readily reachable existing contacts. It omits unknown prospects, former customers outside the list, and people the coordinators did not select.

Voluntary participation creates another boundary. The survey records the experiences of people who chose to answer. The record does not explain why the others declined or stayed silent.

Do not assume unhappy customers are more likely to respond. Do not assume busy, successful customers are less likely to respond. Either story requires evidence absent from this example.

AAPOR distinguishes probability from nonprobability samples and cautions about reporting the latter. These opt-in customer responses do not establish a market-wide demand percentage.

Inviting more convenient contacts would increase the number of observations. It would not, by itself, repair the missing selection design. There is no arbitrary minimum count at which this worksheet becomes representative.

The question also measures task difficulty, not willingness to purchase a replacement. Pew's question-writing guidance emphasizes what wording asks respondents to evaluate. A demand claim would require evidence about demand.

Keep confidence intervals out of this descriptive example. No probability design or justified model has been specified for estimating a broader population.

Close the Report With a Bounded Decision

The fictional report can close this way:

Twelve people from nine accounts complete the questionnaire. Six report difficulty finding the checklist. One skips that item. Contacts were selected for reachability, and participation was voluntary. These results describe the respondents. They do not estimate market demand. The proposed next step is a task observation before deciding whether to test a revised contents page.

Attach the questionnaire version, field period, invitation counts, routing rules, and analysis choices to that paragraph. Retain the two partial submissions separately.

If the next research contradicts the survey interpretation, preserve the disagreement. A later observation can change the proposed action without making the earlier counts disappear.

Keep the claim inside the sample.

Sources

Topic map: Customer and usability research

Articles in this cluster, then the 3 clusters that share the most topics with it.

This cluster: Customer and usability research

Neighbor: Topical authority and keyword strategy

Neighbor: Content production with AI

Neighbor: Technical SEO, crawling, and migrations