Office fit-out, refurbishment and furniture for businesses, schools and practices across Worcestershire, Herefordshire and the West Midlands. 38 staff. 224 quotations over 12 months, and 10 short interviews with customers who said yes and no.
Referral wins 65% of the time; tender portal 9%. Every quotation takes estimating time, so the sources with low win rates cost the most per job won.
Days from enquiry to quotation. This shows association, not cause: some slow quotations are slow because the job is harder.
Your records put most losses down to price. Customers who said no gave price as the main reason far less often; speed, follow-up and how the proposal read came up as often. Price is the easy reason to record; it is not always the real one. Six lost interviews is a small number: a strong hint, not proof.
“You were the only ones who walked the building with us before pricing anything. The others just sent a number.”
Understood our needs · Referral
“We've used you twice before and nothing went wrong. That's worth more to us than a few per cent.”
Trust and track record · Repeat customer
“Your quotation came back in two days with three options. Everyone else took a fortnight.”
Speed of response · Website enquiry
“You explained why your price was higher and what we'd get for it. That made it easy to defend to the partners.”
Value explained · Architect or agent
“We'd already chosen someone by the time your quotation arrived. Nobody rang to see how we were getting on.”
Speed of response · Website enquiry
“It was twelve pages of line items. I couldn't see what we were actually getting, so I went with the one I understood.”
The quotation or proposal · Website enquiry
“Honestly you were about fifteen per cent dearer and nobody told us why.”
Value explained · Architect or agent
“We had a question about the programme and it took a week to get an answer. By then we'd lost confidence.”
Follow-up · Referral
“It went to the lowest compliant bid. That's how the portal works.”
Price · Tender portal
1 interviewee asked not to be quoted; their themes are counted but their words are not shown.
| # | Change | Likely effect | Effort | Owner | By when |
|---|---|---|---|---|---|
| 1 | Open every proposal with a one-page summary: what you get, three options, and why it costs what it does The quotation or proposal | High | Low | Mark Ellison | November 2026 |
| 2 | Send every quotation within 3 days of the enquiry, with a call to check it arrived Speed of response | High | Medium | Joanna Pike | December 2026 |
| 3 | Ring every customer five working days after sending, and again before their decision date Follow-up | Medium | Low | Ravi Mistry | November 2026 |
| 4 | Bid on portal tenders only where we meet the framework's experience criteria in full Fit | Medium | Low | Mark Ellison | January 2027 |
| 5 | Record the real reason for every loss by asking the customer, not guessing Other | Low | Low | Joanna Pike | November 2026 |
Ranked by likely effect on the win rate, then by effort. Effect is a judgement from the evidence; measure it by rerunning the review in six months.
Fictional and illustrative. This continues the worked example from November 2026 to January 2027. Every name, figure and interview is invented.
Severn Vale set up the win-loss agent at handover in late October 2026, in its own accounts: its AI assistant, email and CRM. Helen Marsh, the office manager, approves every invitation and runs the calls. She didn't quote for any of the work, so customers can be candid; one person doing both suits a firm of this size. The rules are on Inputs, rows 31 to 40: invite 7 days after a decision and no later than 60; one reminder after 7 days; nobody asked twice within 6 months; at most 8 invitations a month; and a flag when lost customers give the same main reason 2 times. The signed-off review is unchanged. The agent fills a working copy for the April 2027 rerun.
Between November and January, 49 quotations were decided. Within the monthly cap, the agent drafted 19 invitations, alternating losses and wins. It skipped 2: one contact marked do-not-email in the CRM, and one customer already asked in June. Helen held 2 of the drafts herself: one customer had an open complaint about a past job, and one tender portal buyer's rules bar contact after the award. Nothing went to a customer that Helen hadn't read and sent.
| Date | What the agent found | Decided by | What they decided |
|---|---|---|---|
| 2 Nov 2026 | Q2809 Elm Software marked Won CRM quotation record, 26 Oct 2026 | Helen Marsh | Invitation drafted in Helen's mailbox. Helen read it and sent it on 3 Nov. |
| 2 Nov 2026 | Q2807 Foregate Insurance Brokers marked Lost; reason recorded: Price CRM quotation record, 27 Oct 2026 | Helen Marsh | Not invited: the contact is marked do-not-email in the CRM. Reason left as recorded. |
| 23 Nov 2026 | Pershore Road Dental Practice (Q2816) asked not to be contacted Reply in Helen's mailbox, 18 Nov 2026 | Helen Marsh | Helen added them to Do not contact on 18 Nov. No reminder drafted. Never to be invited again. |
| 30 Nov 2026 | Losses with no reason recorded, November Working copy, Quotations: 3 of 8 November losses | Joanna Pike | Estimators skipped the reason when the customer didn't reply. Joanna asked them to ring and ask; two reasons added as the customers gave them. |
| 7 Dec 2026 | Follow-up is the main reason in 2 lost interviews since the review Interviews for Q2808 Kingsway Insurance Brokers and Q2833 Malvern Surveyors; Inputs B39 = 2 | Mark Ellison | Ravi found the five-day call was made only on quotations over £25,000. From 14 Dec it applies to every quotation. Look again in January. |
| 11 Jan 2027 | Interviews since the review reached the minimum for wins and for losses Working copy, Interviews: 3 wins, 3 losses; Inputs B28 = 3 | Mark Ellison | Themes from the new interviews now count as a finding, not a hint. Full review stays booked for April 2027. |
Both customers said their words could be used.
“The summary on the front page told us what we'd get under each of the three options. We didn't need to read the rest.”
The quotation or proposal · Won · Website enquiry
“We were waiting for a call back that never came, so we went with the firm that kept in touch.”
Follow-up · Lost · Referral
None of the 4 lost customers interviewed since the review gave price as the main reason. Two gave follow-up, which led to the December decision. Two of the 3 winners mentioned the new one-page summary, an early sign that change 1 is holding. Seven interviews are a strong hint, not proof. The win rate is up on the review, but three months is short and the work differs from month to month: this is association, not cause. Helen spent about 20 minutes a week on drafts and coding, plus about 20 minutes for each call (estimate). The April rerun starts with the quotations and interviews already gathered.
Win rates come from 198 decided quotations over 12 months; open and withdrawn quotations are left out, and groups with fewer than 10 decided are not compared. The themes come from 10 interviews (4 wins, 6 losses). Differences in win rate show association, not cause: a fast quotation may win because the job was simpler. The likely effect of each change is a judgement.