Commercial insurance broker for small and mid-sized firms across Shropshire, Herefordshire and mid-Wales, from offices in Shrewsbury and Ludlow. 58 staff.
Verdict: 689 responses from 14 Sep to 9 Oct 2026; Customers 521 (25% responded); Lapsed customers 46 (13% responded, low response); Lost prospects 34 (14% responded, low response); Non-customers 88 (5% responded, low response). Non-customers: heard of us 34% (give or take 10), named us unprompted 8%; Rival A 76%; Rival B 65% (low response). Would consider us: Customers: 84% ±3; Lapsed customers (low response): 20% ±12; Lost prospects (low response): 38% ±17; Non-customers (low response): 8% ±6. Known for: Customers: Personal service (74%); Lapsed customers (low response): Personal service (52%); Lost prospects (low response): Personal service (68%); Non-customers (low response): Personal service (50%). Knows my type of business: Customers 51% ±4, rivals 28%; Lapsed customers (low response) 41% ±15, rivals 29%; Lost prospects (low response) 47% ±17; Non-customers (low response) 43% ±18, rivals 21% (for customers and lapsed customers this sets their experience of us against their impression of the rivals). Recommend: Customers: average 7.6 (±0.2), net +15 (±7) from 505; Lapsed customers (low response): average 6.4 (±0.6), net -41 (±21) from 44. Reasons not to buy: Lapsed customers (low response): The price (39%); Lost prospects (low response): The price (62%); Non-customers (low response): Happy with my current broker (57%). Response below 20% in 3 audiences: those who answered may differ from those who didn't, and their figures are marked low response; early and late responders no different beyond chance on considering us (this check can only catch differences of about Customers 9 points; Lapsed customers 32 points; Lost prospects 50 points; Non-customers 15 points); respondents' profile against the frame: differs beyond chance in Customers: 4 categories; Lapsed customers: 2 categories; Lost prospects: 2 categories. How the value is shown: Measure later for 3 measures that can move, with Unlocks; the other 14 are a snapshot. Can move: Would consider us (customers); Link us with the phrase to be known for (customers); Average likelihood to recommend (0-10) (customers). Repeat 14 Sep 2027 to 9 Oct 2027 (the same months a year later); owner: Helen R, managing director; Not yet measured: run the same survey 14 Sep to 9 Oct 2027 and fill in the Value Case.
| Audience | In the frame | Expected response | Responses if all invited | Measures that can move | Recommendation |
|---|---|---|---|---|---|
| Customers | 2,140 | 25% | 535 | 3 | Invited enough for the measures that can move |
| Lapsed customers | 410 | 15% | 62 | 0 | Snapshot only: no measure can move from this frame; invite for the picture and the products it feeds |
| Lost prospects | 260 | 15% | 39 | 0 | Snapshot only: no measure can move from this frame; invite for the picture and the products it feeds |
| Non-customers | 2,050 | 5% | 103 | 0 | Snapshot only: no measure can move from this frame; invite for the picture and the products it feeds |
| Measure (audiences where asked) | Needed each round | Most the frame can give | Verdict |
|---|---|---|---|
| Unprompted awareness of us Non-customers | 392 | 103 | Snapshot only even inviting all 2,050 gives about 103 of the 392 needed |
| Heard of us Non-customers | 392 | 103 | Snapshot only even inviting all 2,050 gives about 103 of the 392 needed |
| Would consider us Customers | 392 | 535 | Can move invite at least 1,568 of 2,140 |
| Would consider us Lapsed customers | 392 | 62 | Snapshot only even inviting all 410 gives about 62 of the 392 needed |
| Would consider us Lost prospects | 392 | 39 | Snapshot only even inviting all 260 gives about 39 of the 392 needed |
| Would consider us Non-customers | 392 | 103 | Snapshot only even inviting all 2,050 gives about 103 of the 392 needed |
| Link us with the phrase to be known for Customers | 392 | 535 | Can move invite at least 1,568 of 2,140 |
| Link us with the phrase to be known for Lapsed customers | 392 | 62 | Snapshot only even inviting all 410 gives about 62 of the 392 needed |
| Link us with the phrase to be known for Lost prospects | 392 | 39 | Snapshot only even inviting all 260 gives about 39 of the 392 needed |
| Link us with the phrase to be known for Non-customers | 392 | 31 | Snapshot only even inviting all 2,050 gives about 31 of the 392 needed |
| Average likelihood to recommend (0-10) Customers | 392 | 535 | Can move invite at least 1,568 of 2,140 |
| Average likelihood to recommend (0-10) Lapsed customers | 392 | 62 | Snapshot only even inviting all 410 gives about 62 of the 392 needed |
| Net recommend score Customers | 879 | 535 | Snapshot only even inviting all 2,140 gives about 535 of the 879 needed |
| Net recommend score Lapsed customers | 879 | 62 | Snapshot only even inviting all 410 gives about 62 of the 879 needed |
| Top reason not to buy Lapsed customers | 392 | 31 | Snapshot only even inviting all 410 gives about 31 of the 392 needed |
| Top reason not to buy Lost prospects | 392 | 20 | Snapshot only even inviting all 260 gives about 20 of the 392 needed |
| Top reason not to buy Non-customers | 392 | 15 | Snapshot only even inviting all 2,050 gives about 15 of the 392 needed |
Each measure needs enough responses, both now and at the repeat, to show the owner's realistic change (10 points on a share, 0.5 on the 0-10 average, 10 on the net score) at 2.8 standard errors, sized at a cautious 50% share. The plan compares that with the people in each frame times the expected response rate. Every client was invited, so three customer measures can move. Non-customer awareness can never get there from a chamber of 2,050 members answering at about 5%: it is a snapshot, not a measure later.
| Audience and sample frame | Delivered | Responses | Response rate | Before / after the reminder | Would consider us: before / after | Smallest difference this check can catch |
|---|---|---|---|---|---|---|
| Customers Clients with a commercial policy live on 1 September 2026, one contact a client | 2,085 | 521 | 25% | 327 / 194 | 84% / 84% | 9 points |
| Lapsed customers low response Clients who didn't renew between 1 September 2024 and 31 August 2026 | 369 | 46 | 13% | 27 / 19 | 22% / 16% | 32 points |
| Lost prospects low response Businesses quoted for new business between 1 September 2025 and 31 August 2026 that didn't go ahead | 246 | 34 | 14% | 23 / 11 | 39% / 36% | 50 points |
| Non-customers low response All members of the local chamber of commerce in Shropshire and Herefordshire; current clients screened out by a question after the awareness questions | 1,950 | 88 | 5% | 61 / 27 | 10% / 4% | 15 points |
Response below 20% in 3 audiences: those who answered may differ from those who didn't, and their figures are marked low response; early and late responders no different beyond chance on considering us (this check can only catch differences of about Customers 9 points; Lapsed customers 32 points; Lost prospects 50 points; Non-customers 15 points); respondents' profile against the frame: differs beyond chance in Customers: 4 categories; Lapsed customers: 2 categories; Lost prospects: 2 categories. The response rate counts people screened out by the screening questions as having answered. Each audience was sampled separately, so results are never added together, and the low-response bias isn't in any give or take.
Main activity (question 8): respondents' share, with the frame's known share beneath.
| Category | Customers | Lapsed customers | Lost prospects | Non-customers |
|---|---|---|---|---|
| Construction and trades | 29% frame 30% | 38% frame 32% | 50% frame 33% More than the frame | 24% frame 22% |
| Manufacturing | 13% frame 15% | 18% frame 14% | 19% frame 15% | 17% frame 12% |
| Farming and land | 26% frame 20% More than the frame | 9% frame 16% | 0% frame 12% Fewer than the frame | 10% frame 10% |
| Retail and hospitality | 12% frame 15% | 18% frame 18% | 19% frame 18% | 24% frame 22% |
| Professional services | 7% frame 10% Fewer than the frame | 7% frame 10% | 6% frame 12% | 20% frame 22% |
| Transport and haulage | 8% frame 7% | 4% frame 7% | 3% frame 7% | 2% frame 7% |
| Other | 4% frame 3% | 7% frame 3% | 3% frame 3% | 2% frame 5% |
Farming clients answered more readily than other clients, so the customer figures lean a little towards farms. With about 40 categories compared, one or two can differ by chance; the differences among lapsed customers and lost prospects rest on few responses.
heard of us 34% (give or take 10), named us unprompted 8%; Rival A 76%; Rival B 65% (low response). Every share has a give or take of two standard errors; rivals are shown only where at least 30 answered.
Customers: their experience of us against their impression of the rivals (Us: 521 had heard of them · Rival A: 359 had heard of them · Rival B: 229 had heard of them)
| Phrase | Us | Rival A | Rival B |
|---|---|---|---|
| Knows my type of business Ahead | 51% ±4 | 31% ±5 | 24% ±6 |
| Good value for money Could be chance | 42% ±4 | 37% ±5 | 20% ±5 |
| Handles claims well Ahead | 50% ±4 | 18% ±4 | 16% ±5 |
| Easy to get hold of Ahead | 49% ±4 | 26% ±5 | 34% ±6 |
| Personal service Ahead | 74% ±4 | 9% ±3 | 41% ±7 |
| Easy to deal with online Behind | 20% ±4 | 41% ±5 | 19% ±5 |
| Expensive (negative) Could be chance | 23% ±4 | 28% ±5 | 22% ±5 |
Non-customers low response: impression against impression (Us: 30 had heard of them · Rival A: 67 had heard of them · Rival B: 57 had heard of them)
| Phrase | Us | Rival A | Rival B |
|---|---|---|---|
| Knows my type of business Could be chance | 43% ±18 | 27% ±11 | 16% ±10 |
| Good value for money Could be chance | 33% ±17 | 37% ±12 | 26% ±12 |
| Handles claims well Could be chance | 10% ±11 | 21% ±10 | 25% ±11 |
| Easy to get hold of Could be chance | 20% ±15 | 27% ±11 | 30% ±12 |
| Personal service Could be chance | 50% ±18 | 7% ±6 | 46% ±13 |
| Easy to deal with online Behind | 13% ±12 | 42% ±12 | 12% ±9 |
| Expensive (negative) Could be chance | 30% ±17 | 27% ±11 | 23% ±11 |
"Ahead" or "Behind" only where the gap to the best-placed rival is bigger than the two give or takes combined. A customer's view of us comes from experience, but their view of a rival is an impression, so customer comparisons flatter the business; read them that way. One phrase is negative ("Expensive"), so the list isn't a set of compliments. Knows my type of business: Customers 51% ±4, rivals 28%; Lapsed customers (low response) 41% ±15, rivals 29%; Lost prospects (low response) 47% ±17; Non-customers (low response) 43% ±18, rivals 21% (for customers and lapsed customers this sets their experience of us against their impression of the rivals).
| Audience | Answered | 0-6 / 7-8 / 9-10 | Average (0-10) | Net score |
|---|---|---|---|---|
| Customers | 505 | 126 / 177 / 202 | 7.6 ±0.2 | +15 ±7 |
| Lapsed customers low response | 44 | 23 / 16 / 5 | 6.4 ±0.6 | -41 ±21 |
The net score is the share scoring 9 or 10 less the share scoring 0 to 6. It throws away the middle of the scale, so it is noisier than the average: with 505 customers its give or take is 7 points, against 0.2 on a 0-10 average. Reasons not to buy were asked only of people who had heard of the business (46 lapsed customers, 34 lost prospects, 30 non-customers), all three low response.
| Reason not to buy | Lapsed customers | Lost prospects | Non-customers |
|---|---|---|---|
| The price | 39% | 62% | 27% |
| Don't know enough about them | 0% | 15% | 33% |
| Happy with my current broker | 30% | 32% | 57% |
| Slow to come back to me | 33% | 12% | 0% |
| Not specialists in my trade | 22% | 21% | 7% |
| Too small for my needs | 11% | 9% | 10% |
| Theme | Customers | Lapsed customers | Lost prospects | Non-customers |
|---|---|---|---|---|
| Local and personal Local, friendly, knows them personally, same contact | 33% | 19% | 19% | 4% |
| Know their stuff Expertise, advice, knows the trade or the cover | 22% | 12% | 22% | 7% |
| Help with claims Support when making a claim | 13% | 5% | 0% | 1% |
| Good value Fair price, value, saved money | 11% | 7% | 4% | 1% |
| Expensive Premiums or fees seen as high | 8% | 26% | 30% | 4% |
| Slow or hard to reach Slow replies, hard to get hold of, chasing | 7% | 29% | 4% | 7% |
| Old-fashioned Paper-based, not online, dated | 6% | 2% | 22% | 4% |
| Don't know them Never heard of them, nothing comes to mind | 0% | 0% | 0% | 70% |
Customers: Local and personal (33%); Lapsed customers: Slow or hard to reach (29%); Lost prospects: Expensive (30%); Non-customers: Don't know them (70%). 117 of 586 answers coded by a second person, blind; agreement 85%, allowing for chance (kappa) 0.83.
Unlocks. The results feed four catalogue products, each ready now:
| Product | What it needs | Status |
|---|---|---|
| Competitive Positioning Map Positions the business against Rival A and Rival B on the phrases buyers actually link with each firm | Phrases linked with us and a rival from 30 or more aware respondents in one audience | Ready now |
| Value Proposition and Messaging Answers the reasons not to buy that lost prospects and non-customers gave | Reasons not to buy from 30 or more lost prospects or non-customers | Ready now |
| Brand Voice Guide Uses customers' own words for what the business is known for | Open answers coded by two people with agreement at or above the minimum | Ready now |
| Price Rise Plan How exposed a price rise is: how warmly customers recommend, and how many already consider a rival | Recommend score and consideration of rivals from 30 or more customers | Ready now |
Measure later, where a measure can move. The repeat runs 14 Sep 2027 to 9 Oct 2027 (the same months a year later); owner: Helen R, managing director, with the same questionnaire, audiences, sampling, senders and incentive. For each measure the workbook shows the smallest change the repeat can show (2.8 standard errors, so a real change of that size is seen four times in five) against the owner's assumed realistic change. Inviting every client means customers' "would consider us", their link with "knows my type of business" and their average likelihood to recommend can move; everything else is a snapshot. The top reason not to buy is chosen on the first half of the responses and measured on the second. Where the rivals are measured, our change is read net of theirs, with the two give or takes combined.
| Measure | Now | Give or take | Smallest change the repeat can show | Can move? |
|---|---|---|---|---|
| Unprompted awareness of us Non-customers · 88 responses low response | 8% | ±6 | 11 points | Snapshot only needs about 166 responses a round |
| Heard of us Non-customers · 88 responses low response | 34% | ±10 | 20 points | Snapshot only not possible with this baseline |
| Would consider us Customers · 521 responses | 84% | ±3 | 6 points | Can move already enough |
| Would consider us Lapsed customers · 46 responses low response | 20% | ±12 | 23 points | Snapshot only not possible with this baseline |
| Would consider us Lost prospects · 34 responses low response | 38% | ±17 | 33 points | Snapshot only not possible with this baseline |
| Would consider us Non-customers · 88 responses low response | 8% | ±6 | 11 points | Snapshot only needs about 166 responses a round |
| Link us with the phrase to be known for Customers · 521 responses | 51% | ±4 | 9 points | Can move already enough |
| Link us with the phrase to be known for Lapsed customers · 46 responses low response | 41% | ±15 | 29 points | Snapshot only not possible with this baseline |
| Link us with the phrase to be known for Lost prospects · 34 responses low response | 47% | ±17 | 34 points | Snapshot only not possible with this baseline |
| Link us with the phrase to be known for Non-customers · 30 responses low response | 43% | ±18 | 36 points | Snapshot only not possible with this baseline |
| Average likelihood to recommend (0-10) Customers · 505 responses | 7.6 | ±0.2 | 0.4 | Can move already enough |
| Average likelihood to recommend (0-10) Lapsed customers · 44 responses low response | 6.4 | ±0.6 | 1.2 | Snapshot only not possible with this baseline |
| Net recommend score Customers · 505 responses | +15 | ±7 | 14 points | Snapshot only needs about 18,324 responses a round |
| Net recommend score Lapsed customers · 44 responses low response | -41 | ±21 | 41 points | Snapshot only not possible with this baseline |
| Top reason not to buy: The price Lapsed customers · 24 responses low response | 29% | ±19 | 32 points | Snapshot only not possible with this baseline |
| Top reason not to buy: The price Lost prospects · 17 responses low response | 65% | ±23 | 40 points | Snapshot only not possible with this baseline |
| Top reason not to buy: Happy with my current broker Non-customers · 18 responses low response | 61% | ±23 | 41 points | Snapshot only not possible with this baseline |
Illustration only, not a forecast: the new-business win rate from the records is 31.2% (118 won of 378 quotations in 12 months; give or take 4.8 points; the smallest change a re-read can show is 9.4 points). If it reached the owner's assumed 36%, that would be worth £20,792 a year in first-year gross profit (£10,396 if it moved half as far); the gap is smaller than a year's re-read can show, so it would be direction only, and the survey can't prove what moved it. Cost in full: £6,418 cash (fee, the £1-a-response donation the business pays, the survey tool, both rounds) and £960 of team time (30 hours, built up from the questionnaire, lists, coding and reading the results).
If the business uses the results in marketing: an FCA-regulated broker's claims must be clear, fair and not misleading (ICOBS 2.2.2R), and a survey of those who chose to answer is weak evidence for "most local businesses prefer us". Check any claim with the compliance lead first.
The results rest on 689 responses to a survey of 14 questions, from audiences sampled separately and never pooled: Customers 521 (25% responded); Lapsed customers 46 (13% responded, low response); Lost prospects 34 (14% responded, low response); Non-customers 88 (5% responded, low response). Every share has a give or take of two standard errors; a share from fewer than 10 responses isn't shown, and a rival only where at least 30 answered. Where few responded, those who answered may differ from those who didn't, and the figures describe them (marked low response); that bias isn't in any give or take. Early and late responders and the respondents' profile against the frame are compared as checks, and they can only catch large differences. Open answers were coded into themes by one person and about 1 in 5 again by a second, blind (agreement 85%). For customers and lapsed customers, comparisons with the rivals set their experience of us against their impression of the rivals. The realistic changes, the planning figures and the target win rate are the owner's assumptions; the win rate, quotations and gross profit come from the records, and any £ is an illustration, not a forecast. The sample was sized before fieldwork; 3 of 17 measures can move at the repeat and the rest are a snapshot, so the value rests on those measures and on the products the results feed; a change in the survey shows that perception moved, not why.
Fictional and illustrative. This continues the worked example from the fieldwork in September 2026 to the preparation of the repeat in August 2027. Every name and figure is invented.
Merriscote set up the survey agent in its own accounts at the working session that agreed the questions, before any invitation went out: its AI assistant, its SharePoint and its survey tool. Its three rules are at the foot of Inputs, and it reads the settings the survey already had. Every message was approved by Helen R, managing director. The raw responses were held by Priya K, the consultant, in a folder only she could open; she checked every theme the agent proposed. Nothing was reported for a group smaller than 10. The agent drafted nothing until the research and data questions on Inputs were answered, research only was Yes and the client list stayed in the business.
| Date | What the agent found | Approved or decided by | What happened |
|---|---|---|---|
| 10 Sep 2026 | Invitations drafted for all four audiences, and a reminder for each, from the pack (section 3) Pack template; Audiences; Checks: research and data questions answered; Inputs: research only = Yes, client list shared = No | Helen R | Helen read all eight and approved them on 11 Sep. The consultant sent three audiences' invitations on 14 Sep; Helen's office passed the chamber its text and link, and the chamber sent to its own members. The client list never left the business. |
| 25 Sep 2026 | Responses before the reminders: customers 327 (16% of 2,085 delivered), lapsed 27 (7%), lost prospects 23 (9%), non-customers 61 (3%) Survey tool totals per audience link; Inputs warning level 20% | Helen R | All four below the 20% warning level. The reminders go as planned on 28 Sep, one each and no more. Any audience still below 20% at the close carries the low-response mark. |
| 9 Oct 2026 | Fieldwork closed: 689 responses (customers 521, lapsed 46, lost prospects 34, non-customers 88) Survey tool export, 9 Oct, pasted on Responses in Priya's restricted copy | Priya K | Customers reached 25%; the other three stayed below 20% and are marked low response. Helen was sent totals only. |
| 12 Oct 2026 | Theme proposed for each of the 586 open answers; none named a person Responses, open question; the coding rules on Themes | Priya K | Priya accepted or changed every proposal herself. Tom coded his 117 without seeing the agent's or Priya's codes; they gave the same theme to 100. |
| 12 Nov 2026 | Test run of the repeat's preparation, as if 14 Sep 2027 were 30 days away: a test copy of the workbook with the questionnaire word for word, the same four audiences, senders, incentive and reminder timing, 8 draft invitations and reminders marked TEST, and the list of what to refresh Value Case: repeat first day; Inputs: days before the repeat; this year's pack and Audiences sheet | Helen R | Read with the fieldwork log at the 30-day check-in on 16 Nov; nothing differed from the baseline and nothing was sent. Helen asked for the chamber's contact to be on the refresh list; the test copy and drafts were then deleted. |
| 1 Dec 2026 | Delete-by date in a month: contact lists, survey links and the export with contact details Inputs: delete by 31 Dec 2026 | Priya K | Lists, links and the identifiable export deleted on 15 Dec. The results without contact details, the questionnaire and the pack are kept for the repeat. |
| 16 Aug 2027 | Repeat prepared for 14 Sep 2027: the same questionnaire word for word, the same four audiences, senders, incentive and reminder timing; 8 drafts Value Case: repeat first day; last year's pack and Audiences sheet | Helen R | Helen had the client and lapsed lists refreshed from the broking system and the suppression list checked; the chamber confirmed it will send again. Drafts approved on 1 Sep. |
The 30-day check-in. The agent's main work was the fieldwork, during the project, so the check-in came at day 30 after the 16 October 2026 handover: 16 November 2026, the Monday, as day 30 fell on a Sunday. On 12 November the agent did a test run of the repeat's preparation, as if the repeat's first day, 14 September 2027, were 30 days away: a test copy of the workbook with the questionnaire word for word and the same four audiences, senders, incentive and reminder timing, 8 draft invitations and reminders marked TEST, and the list of what to refresh. Nothing was sent. At the check-in Priya, the consultant, and Helen read the fieldwork log, the 4 rows from 10 September to 12 October, and the plan for the repeat (preparation from 15 August 2027, lists to refresh, Helen approving) with the test run beside it. Nothing differed from the baseline; Helen added the chamber's contact to the refresh list, and the test copy and drafts were deleted. Priya's upkeep of the agent ended there; after it the agent is Merriscote's.
Helen spent about 30 minutes reading each round of drafts, and Priya about 3 hours checking the proposed themes (estimates). The agent read totals per audience during fieldwork and the export only at the close, in Priya's restricted copy; Helen saw totals. It reminded Priya a month before the delete-by date, and Priya deleted the lists and links herself. A year on, it prepared the repeat with the same questionnaire word for word, the same audiences, senders, incentive and reminder timing, and flagged the lists for refreshing; the repeat is then read the same way, with part B of the Value Case filled from the repeat's own workbook. The baseline itself is unchanged.