Roasts and supplies coffee, equipment servicing and barista training to cafés, restaurants and offices across West and North Yorkshire. 24 staff. 214 customers in the export; a customer has lapsed after 4 months without an order.
Most lapsed customers won't come back, and the plan doesn't pretend otherwise. But the approach is cheap: a first round of 30 customers costs about £503 and carries £4,535 of the expected gain. And the biggest single reason they left is unknown: 19 customers stopped and nobody knows why, so the first round includes calls to the 10 most valuable of them simply to ask.
Learning as you go. Price: 12 approached in a spring trial and 2 came back, so the plan blends that in: 16% instead of the assumed 15%. Twelve approaches is still mostly luck; record every approach on the Reasons sheet and your own rate will carry more weight as they build up.
| Reason | Approach | Owner | Cost to approach | Offer per win |
|---|---|---|---|---|
| Unknown | Short call to ask why they stopped; offer a free bag with the next order | Account manager | £8 | £25 |
| Went to a competitor | Free tasting visit with a new seasonal blend; ask what the other roaster does better | Sales manager | £25 | £80 |
| Price | Call with a like-for-like price check and a three-month price hold | Sales manager | £12 | £60 |
| Service problem | Owner calls: say what went wrong and what has changed (second van, delivery window) | Managing director | £15 | £40 |
| Needs changed | Email the new decaf and oat-milk range; no offer | Marketing assistant | £2 | – |
Customers who closed or moved away aren't approached. 3 lapsed customers are left out at their own or our request.
| # | Customer and reason | Last order | Gross profit a year they used to bring | Chance | Expected gross profit | Less the cost |
|---|---|---|---|---|---|---|
| 1 | Ossett Lane Hotel Price | Mar 2026 | £4,175 | 16% | £399 | £377 |
| 2 | Wolden Hotel Service problem | Apr 2026 | £3,020 | 20% | £362 | £339 |
| 3 | Garsby Coffee House Price | Dec 2024 | £5,762 | 10% | £330 | £312 |
| 4 | Cragside Coffee House Service problem | Sep 2025 | £3,788 | 12% | £273 | £253 |
| 5 | Ossett Lane Coffee House Service problem | Jan 2024 | £6,485 | 6% | £233 | £216 |
| 6 | Moorgate Deli Price | Nov 2025 | £2,344 | 16% | £224 | £202 |
| 7 | Netherby Coffee House Went to a competitor | Jun 2025 | £5,720 | 6% | £206 | £176 |
| 8 | Lumbfoot Coffee House Went to a competitor | Feb 2025 | £5,582 | 6% | £201 | £171 |
| 9 | Pickersgill Bakery Price | Mar 2026 | £1,840 | 16% | £176 | £154 |
| 10 | Netherby Hotel Service problem | Sep 2024 | £4,410 | 6% | £159 | £141 |
| 11 | Cragside Hotel Went to a competitor | Jan 2025 | £4,746 | 6% | £171 | £141 |
| 12 | Ainsty Deli Went to a competitor | Jan 2026 | £2,764 | 10% | £166 | £133 |
| 13 | Wexcombe Café Service problem | Nov 2023 | £3,982 | 6% | £143 | £126 |
| 14 | Hebble Bistro Service problem | Nov 2024 | £1,961 | 12% | £141 | £121 |
| 15 | Holme Bank Hotel Needs changed | Feb 2026 | £4,103 | 5% | £123 | £121 |
| 16 | Hesketh Café Went to a competitor | Mar 2026 | £2,369 | 10% | £142 | £109 |
| 17 | Skelbrook Deli Price | Mar 2026 | £1,365 | 16% | £130 | £109 |
| 18 | Heptonby Hotel Service problem | Dec 2023 | £3,238 | 6% | £117 | £99 |
| 19 | Wolden Bistro Price | Dec 2024 | £1,886 | 10% | £108 | £90 |
| 20 | Garsby Café Needs changed | Aug 2025 | £4,771 | 3% | £86 | £84 |
Ranked by expected gross profit less the cost of the approach and offer. The chance falls the longer a customer has been gone.
Ask why first (no known reason; gross profit a year they used to bring): Dunnock Coffee House (£4,469), Ellerdyke Café (£4,444), Garsby Hotel (£3,616), Pickersgill Hotel (£3,276), Thrushcross Coffee House (£2,688), Netherby Bistro (£2,512), Calverly Bakery (£2,423), Hebble Bakery (£2,113), Kelda Offices (£1,966), Ashdene Hotel (£1,865).
Customers, dates and sales come from your records; a customer has lapsed after 4 months without an order, and those gone more than 36 months are left out. Why each stopped comes from your team; where nobody knows, the reason is Unknown. The chance of winning each back is assumed by reason, scaled down the longer they have been gone, and blended with your own results as approaches are recorded (after 10 approaches, half and half). A customer won back is assumed to return with 60% of their former spend; expected gross profit is a year of that at your 42% margin, less the cost of approaches and offers; about 70% of those won back are assumed to be buying a year later. It is a plan to test, not a forecast: approach the first batch, record what happens, and let the results take over.
Fictional, continuing the example: every customer, person and date is invented, and Brackenmoor's people are named by role. The log is on the Agent sheet of the worked-example workbook (a sample of its rows).
Brackenmoor chose the guided route and set up the agent in the final session, in an AI assistant the business already uses, connected read-only to its sales system and CRM (its IT confirmed this before it committed), and updated its privacy notice first, because several of its customers are sole traders. Each Monday it reloads last orders and sales into the workbook, spots customers newly past 4 months without an order, and drafts the approach for each reason from the plan, or an ask-why call where nobody knows, at most 5 for one owner a week. It flags an approach with no reply after 3 weeks and leaves a customer alone for 3 months after an approach: the managing director's settings on Inputs. It drafts an email only where the person's consent is recorded. Each reason's owner made every approach from their own phone or email. Nothing left the business, and no decision changed, without one of its own people approving it.
| Date | What the agent drafted or flagged | Decided by | What happened |
|---|---|---|---|
| 19 Oct No flag | Drafted the first round's approaches by reason, at most 5 for one owner that week: 17 of the 30, the rest over the next weeks. | Each owner | All 17 made that week; the sales manager rewrote two openings. |
| 19 Oct Flag | Garsby Café is a sole trader with no marketing consent recorded: held the email about the new range and drafted a letter instead. | Marketing assistant | Letter approved and posted on 20 October. |
| 26 Oct No flag | Price: 4 more approached and Moorgate Deli ordered again on 23 October. Contacted and won back updated on Reasons: 16 and 3. | Sales manager | Confirmed against the call log. |
| 2 Nov Flag | Ellerdyke Café asked not to be contacted again, on the ask-why call. | Managing director | Marked do not contact, with the date in Notes; nothing more drafted. |
| 9 Nov Flag | Newly lapsed after 4 months without an order: Scarcroft Bistro, Oldacre Hotel and Stainby Coffee House. No reason known, so ask-why call notes drafted. | Account manager | Called on 10 and 11 November: a cheaper roaster, a new head chef trying another, a refit until March. Reasons entered. |
| 9 Nov Flag | Wexcombe Café: no reply 3 weeks after the managing director's letter. | Managing director | One follow-up call, then leave it alone for 3 months. |
At the results review on 18 November, a month after the agent's first weekly run, Geoff and the managing director read the log and the first round's results together in Brackenmoor's own account, and Geoff's upkeep of the set-up ended. Three customers who lapsed in early November were asked why within days rather than at the next quarterly rerun, one customer who asked to be left alone was, and Price now rests on 16 approaches rather than 12. Because the agent reviews named customers' buying every week, Brackenmoor also wrote a short data protection impact assessment (not legal advice). Reading the note and approving drafts took the owners about 30 minutes a week between them (estimate). The months after add nothing to the figures above: the expected gains stay what the plan worked out, and the next rerun blends in the results recorded.