Northwind’s growth is already inside its customer base.
We asked AiCopilot a series of plain-English questions about Northwind’s sales, margins and customer accounts. Every finding below comes from Northwind’s own Opera SE records for January to September 2026, and each one ends with what to do about it.
Every figure comes from Northwind’s own Opera records and reconciles to the sales ledger within 0.02%.
Disclaimer. Company and customer names are fictional and figures are illustrative; based on a demonstration Pegasus Opera dataset.
Customers who order online buy twice as much per order.
Northwind has always felt web orders were bigger. The data confirms it, and shows why: online, customers see the range and add more products to the basket.
The same customers spend more per order online
Order value: the 418 customers who use both channels. Lines per order: all orders over £50. Shares: all invoiced and open orders. Both August–September 2026. Checked
Almost 3 in 10 web orders arrive when the office is closed
Web orders by time placed, August–September 2026 (invoiced and open orders only). Highlighted: before 09:00 or after 18:00. Checked
Restaurants order after service. None of these orders needed anyone to answer a phone or key anything in.
20 busy restaurants still order only by phone or email
They placed 496 orders worth £194k in two months, all keyed in by hand, at 6.4 lines per order against 12.2 on web. Three are chains, and the chains go much further than these sites: 70 of Mizu Group’s 78 sites and 20 of Nori & Broth’s 21 order only by phone or email (735 and 364 orders in the two months).
| Customer | Orders (2 months) | Value (2 months) | Web orders |
|---|---|---|---|
| Mizu Group Twickenham, Reigate, Leeds and Staines | 98 | £37,980 | 0 |
| Nori & Broth Fountain Bridge, Renfield, Gateshead, Manchester (M9) and Soho | 112 | £63,940 | 0 |
| J9 Modern Japanese (Fetter Lane and Coleman Street) | 59 | £24,210 | 0 |
| Tsuki | 28 | £11,480 | 0 |
| Blue Lotus | 25 | £7,690 | 0 |
| Harbour Izakaya Bolton | 24 | £10,980 | 0 |
| Sora (Retail) | 23 | £14,110 | 0 |
| 5 further restaurants | 126 | £24,880 | 0 |
- Set each one up on the web shop and take their next order online with them.
- About 2,950 manually keyed orders a year (£1.16m) move online. If their spend rises by even 10%, that is about £116k more sales a year. Across the whole Mizu Group and Nori & Broth chains the opportunity is several times larger. Scenario
Many restaurants buy their side dishes from Northwind and their core ingredients elsewhere.
Fifteen products are the most widely bought across Northwind’s Japanese restaurants (customer type R02). The ones most often missing from a customer’s basket are Northwind’s best-margin lines.
228 Japanese restaurants buy less than half the core range
Average spend per customer, January–September 2026. Checked Larger restaurants naturally buy a wider range, so we size the opportunity conservatively below.
| Core product | R02 customers buying it | Gross margin |
|---|---|---|
| Kaede Premium Rice 20kg | 244 | 48.3% |
| Shirakiku Premium Unagi Eel | 228 | 46.1% |
| Shirakiku Horseradish (wasabi) Powder | 192 | 53.4% |
| Shirakiku Goma Wakame | 268 | 49.2% |
| Shirakiku Edamame | 286 | 43.1% |
| Shirakiku Soy Sauce Sachets | 241 | 38.2% |
| Shirakiku Tobiko Orange | 213 | 21.4% |
Margin on each sale, across all customers, January–September 2026, before any customer rebates. Northwind’s overall margin is 30.4% after rebates. Checked
- If each of the 228 added just two core products at typical spend, that is about £292k more sales and £129k more gross profit a year. Scenario at £452 per product per 9 months and 45% margin.
25 established accounts, worth £872k, buy core products elsewhere
| Customer | Spend, Jan–Sep | Core range bought |
|---|---|---|
| Haru Kitchen Borehamwood | £106,480 | 6 of 15 |
| Yuki Kitchen | £73,120 | 7 of 15 |
| Akari Restaurant | £69,210 | 8 of 15 |
| Hoshi Restaurant | £52,140 | 8 of 15 |
| Kaze (Old Compton St and South Kensington) | £51,290 | 3–4 of 15 |
| Fox & Crane (head office) | £30,870 | 4 of 15 |
| Ume (Regents Park) | £30,560 | 2 of 15 |
Checked Akari Cheltenham (closed) removed from the total.
Haru Kitchen is the clearest example. They buy the sauce from Northwind but the main ingredient from someone else. They order Northwind’s unagi (eel) sauce every week, but not Northwind’s unagi eel. They order Northwind’s rice vinegar every week, but not Northwind’s premium rice. They have kept buying everything they ordered regularly earlier in the year, so this is a loyal customer who is buying their eel and rice elsewhere.
- Give each account manager this list with the specific missing products per account, and quote on rice and unagi first.
Mizu Group is two businesses under one name.
On paper Mizu Group is one customer with one margin. Look underneath and there are two very different businesses for Northwind: the frozen chicken, and everything else.
Every £1 Mizu Group spends with Northwind earns one of two very different amounts
Gross profit per £1 of sales, Mizu Group sites, January–September 2026. Checked Chicken and other products add up exactly to the total in the same query. The chicken price is the same across the chain (£31–£33 a case), so this is about what each site buys, not what it pays. Mizu Group also has a rebate account: −£154,900 in January–September. After it, Mizu Group as a whole earns 20.3% on £2.28m of sales.
A tale of two Mizu Group sites
Two sites, same brand, same prices, almost exactly the same spend with Northwind. One buys mostly chicken; the other buys none.
The pattern holds right across the chain: every dot below is a Mizu Group site, and the more of its spend goes on chicken, the lower its margin.
Gross margin against the share of each site’s spend that is frozen chicken, 40 largest Mizu Group sites, January–September 2026. Hover a dot for the site. Checked
Try it: what if Mizu Group bought more of everything else?
Scenario Mizu Group’s £1.42m of non-chicken sales after its rebate (January–September), grown by the chosen percentage, at their 28.9% margin after rebate, shown as a full year. No price changes.
- Grow the rest of the basket at the chicken-heavy sites, using Basingstoke and the other high-margin Mizu Group sites as the example of what the chain already buys.
- Several Mizu Group sites already buy little or no chicken from Northwind and buy the wider range instead, so the chain is clearly happy to buy it from Northwind.
Oko Kitchen buys like a model customer. One question decides what it is worth.
Checked 33 trading accounts plus one rebate account. Northwind’s overall 30.4% is after all rebates.
What the data shows
Oko Kitchen buys a broad, premium range with no single product above 10% of its spend, and its invoices earn Northwind 47.8%. A separate rebate account then receives one credit at the end of each month (“JUL REBATE”, “AUG REBATE”), which is paid back to the group about two weeks later, some marked “retro”. The monthly amount varies from £58k to £102k, between 28% and 40% of that month’s sales. After it, the group earns 25.1%. Around nine other customer groups have rebate accounts on a smaller scale.
What the data cannot show
How the monthly amount is worked out. It could bring invoices back to an agreed contract price, apply to particular ranges, or cover wider terms such as marketing or franchise support. Only Northwind knows, and the answer changes the picture: on its invoices Oko Kitchen is Northwind’s best-margin group; after the rebate it is close to average.
- How is Oko Kitchen’s monthly rebate calculated, and does it still reflect what the group buys?
- Whatever the answer, it helps to manage Oko Kitchen as one account with one named owner and to report on it as a group, after rebates.
Opera shows when a customer’s buying pattern changes, as it happens.
Riverbank and Fresh Larder were the 3rd and 4th largest trading customers in the first half of 2026. Both have changed how they buy since the summer. AiCopilot picked this up by reading sales and the sales ledger together, which gives the account team a reason to call while the conversation is easy.
Fresh Larder Production: buying has changed since July
Worth a conversationNet sales by month, 2026 (£). August and September highlighted. Checked totals agree exactly with the sales ledger.
Every one of their main products fell at the same time (nori, rice seasoning, rice, ginger, wasabi), so the change is across the whole account rather than one product. Fresh Larder normally pays by BACS around the 6th to 8th of the month after next. August’s £27.0k would normally have arrived by 8 October and, as at 14 October, is about a week behind that pattern.
- A call from the account manager to understand what has changed for them.
- An account worth about £500k a year in sales at their January–July rate. Estimate
Riverbank Foods: growing fast, and due its next bulk order
GrowingNet sales by month, January 2025 to September 2026 (£). Checked
Riverbank is a growing customer that buys in large, infrequent loads at a margin of around 45% (from our earlier analysis), so each missed load is roughly £40k of gross profit. Estimate Their buying runs in a clear rhythm, and on that rhythm the next load is due now.
- A timely call to line up the next load.
- About £365k a year in sales at their 2026 rate. Estimate
One way to put this to work: give every account one clear job.
This is a suggestion, not a prescription. Northwind’s team knows its customers best and will decide what fits. As an example, the findings above sort customers into four groups, each needing a different kind of attention.
Protect
Accounts whose usual buying pattern has changed
- Now
- Fresh Larder, Riverbank
- Action
- A prompt call to understand what has changed
- Measure
- Spend back to normal
Grow
Large accounts missing core products, and chicken-heavy Mizu Group sites
- Now
- Haru Kitchen, Ume, Fox & Crane, Kaze, Yuki, Mizu Group
- Action
- Quote on the two or three biggest missing products
- Measure
- Core products bought (of 15)
Convert
Restaurants that only order by phone or email
- Now
- Mizu Group (76 sites), Nori & Broth (22 sites), J9 Modern Japanese, Tsuki, Blue Lotus
- Action
- Set up web ordering and take the next order online together
- Measure
- Share of orders via web
Automate
Smaller, steady accounts
- Now
- The long tail of restaurants
- Action
- No account manager time; the web shop prompts “you usually order…”
- Measure
- Lines per order
What a weekly rhythm could look like
- Monday list. Each account manager gets three short lists from AiCopilot: customers at risk, customers due to order, and gaps to fill.
- Before each call. A one-page brief: what this customer usually orders, what is overdue, and what they buy elsewhere.
- Monthly review. Four numbers per account: spend against normal, core range bought, web share and margin.
This is a taster. The next question is Northwind’s.
Everything above came from a handful of plain-English questions. Northwind knows its business better than anyone, and the questions that matter most will be Northwind’s own. A few examples of what the team could ask next:
Each answer arrives in about a minute, worked out from Northwind’s own Opera records and checked a second way where possible.
- A short call to walk through the findings and hear which questions matter most to Northwind.
How this was produced
Every finding came from a plain-English question typed into AiCopilot, which reads Northwind’s Opera SE data and never changes it. We checked each answer before using it, re-asked where results looked wrong, and left out anything we could not confirm.
Limits. Opera keeps how an order was entered (web or manual) only for August and September 2026, so the web findings cover those two months and we could not measure customers before and after they moved online. Branch delivery accounts invoice through head-office accounts, so customers are compared at invoice-account level. Scenarios use the assumptions stated beside them. Northwind overall: £27.86m sales, 30.4% gross margin and 1,840 active trading customers, January–September 2026, excluding customer type EX (2 accounts, £303,900 of sales at nil cost; £28.16m and 31.1% with them). Customer groups include their rebate accounts. Order counts and values include invoiced and open orders only, not credit notes, quotes, proformas or unposted documents.
Web ordering. Every web figure in this report comes from Northwind’s Opera sales orders, not from a web platform.