eCatering

Board report · Maitland Agency and eCatering

eCatering and Maitland: the first nine months

1 November 2025 to 31 July 2026, measured against the same nine months a year earlier

All figures from Shopify order data, Google Analytics 4, Google Search Console, Google Ads and Microsoft Advertising. Revenue basis: Shopify, paid orders, Online Store, including VAT. Figures as at 19 August 2026 and subject to late refunds.

Section 1

The headline

Revenue

£4,839,528

£5,480,024

+£640,496

+13.2%

Orders

10,824

12,430

+1,606

+14.8%

Average order value

£447.11

£440.87

-£6.24

-1.4%

Work delivered

438 completed items

1,671.65 hours

Every person at the agency

The cleanest comparison we have: 1 to 18 August, both years

Revenue

£289,049

£347,899

+20.4%

Orders

620

828

+33.6%

The same eighteen calendar days in each year, measured the same way on both sides.

Revenue is up 13.2 per cent and orders up 14.8 per cent. Average order value is down 1.4 per cent, so the growth is coming from more orders rather than bigger ones. This report sets out where that came from, what did not work, and what we cannot yet prove either way.

Section 2

Up on last year. Behind a target set twelve months ago.

Two different things are being measured, and they are being confused with each other.

Against the same nine months last year, online sales are up 13.2 per cent and orders are up 14.8 per cent. That is a comparison against what the business actually did.

Against the target set twelve months ago, the business is behind. That is a comparison against what somebody hoped it would do, decided before the year began and before most of this work started.

Both statements are true at the same time. Being behind a forecast is not the same as going backwards, and the two need to be argued separately. This report answers the first question, because it is the one the data can answer. The second is a question about how the target was built.

The honest summary in one line: the business is trading better than it was and it is not yet where anyone wants it to be.

Section 3

Monthly revenue and orders, November 2024 to August 2026

Seven of the nine months are up. Two are down.

Nov

+£75,708

+12.4%

Dec

+£47,874

+12.3%

Jan

-£88,228

-16.9%

Feb

+£97,358

+25.2%

Mar

+£140,659

+28.7%

Apr

+£139,418

+25.0%

May

+£40,650

+6.7%

Jun

+£241,490

+38.1%

Jul

-£54,433

-8.4%

Before Maitland (Nov 2024 to Jul 2025)
With Maitland (Nov 2025 to Jul 2026)

What this means: each bar is one month of sales. In the default view the grey bars are last year and the green bars are the same months with Maitland. The full timeline view shows every month from November 2024 onwards.

January and July are the two months below last year. January is explained in section 8. July 2026 restated downwards by £7,422 after late refunds, which is normal for this business and will move again.

August, September and October 2025 sit between the two comparison periods and are shown for continuity. August 2026 is only eighteen days and is not included in any total.

Section 4

What Maitland actually did

Five hundred and eleven pieces of work were raised on the shared project board between 3 November 2025 and 18 August 2026. Four hundred and thirty-eight were completed. Every person at the agency booked time to this account.

Completed work by type of task

What this means: the longer the bar, the more separate pieces of work were finished in that area. The largest single area is campaign creative, followed by work on the website itself.

The 1,671.65 recorded hours, by type of work

Development457.78 hrs · 27.4%
Consultancy398.93 hrs · 23.9%
Creative328.14 hrs · 19.6%
Advertising150.67 hrs · 9.0%
Project management108.96 hrs · 6.5%
Email marketing93.25 hrs · 5.6%
Social40.47 hrs · 2.4%
Other38.36 hrs · 2.3%
Design30.54 hrs · 1.8%
Search engine work20.12 hrs · 1.2%
Strategy3.18 hrs · 0.2%
Content1.25 hrs · 0.1%

What this means: this is where the time went. The two largest slices are building the website and advising the business, not advertising.

Hours recorded each month

What this means: the height of each bar is the number of hours the agency booked to this account in that month.

November is understated because time tracking was not pointed at the project until 10 November. August covers only twelve working days.

Section 5

The work that is not marketing

A marketing agency is normally hired to run advertising. Over nine months this contract has gone considerably further into how the business runs. Each item below is dated and sits on the shared project board.

7 to 13 November 2025
Northern Ireland customs capture built into checkout
Conditional fields now appear at checkout when a delivery address is detected as Northern Ireland, collecting the Economic Operators Registration and Identification number and the Value Added Tax number automatically. Before this the office chased them by hand after the order was placed.
7 November 2025
Bank reconciliation traced for the finance team
The accountant could not balance the bank. One order checked out on 3 November but landed as 31 October. The cause was traced to an approval lag in the finance provider over a weekend. It affects two or three orders a month.
24 November to 10 December 2025
Marketplace order reconciliation
Orders from the OnBuy marketplace were reconciled for the finance function and the process closed out on 10 December.
10 December 2025
Warehouse dead stock analysis
182 slow and non-moving product lines were identified, holding 3,920 units and an estimated £663,792 of value. Twenty-eight of those lines had sold nothing at all in eleven months. Clearing half at a 40 to 50 per cent discount would free around £330,000 of working capital. This is Maitland's own analysis of the stock file.
23 January 2026
Bestseller stock position analysed
Fill rate on the fifty best-selling lines was 77.3 per cent. Ten lines were completely out of stock, including the number three bestseller. Nine more were down to between one and ten units. Across nineteen lines this put roughly £80,150 a month of sales at risk.
26 June to 8 July 2026
Product bundles built from the data
Bundles were constructed from twelve months of frequently-bought-together data, ninety days of line-level sales, seasonality, margin and purchase order dates, rather than from opinion.
2 July 2026
Demand measured on products that cannot be bought
Two out-of-stock product pages were still taking 26,700 visits in ninety days and around 1,300 search appearances a day, with 120 people on the notify-me list. One was the sixth most visited page on the site in July.
26 to 27 July 2026
Stock exposure quantified and taken to the buying team
Thirty-five product lines with over £5,000 of sales this year were sitting at zero or negative stock, carrying £780,522 of demand. Fourteen of those had no purchase order raised at all. Across the whole catalogue, 116 lines were at or below zero, representing 19.6 per cent of this year's demand. This is demand that has already happened on lines that cannot currently be sold. It is not a forecast of recoverable revenue, because a customer who cannot buy one model often buys another.
26 July 2026
A test designed that could prove our own advertising does not pay
A geographic holdout was designed: the United Kingdom split in half on matched trading history, advertising withheld from one half for six weeks. The design states openly that it can detect an eight per cent effect but will not resolve five per cent or smaller. The cost of running it is roughly £8,354 of withheld spend and around £60,068 of foregone revenue.
30 to 31 July 2026
New business prospect list built from Companies House
A feed was built from the public company register. It produced 3,350 addresses of food businesses that registered a limited company in July 2026, segmented into takeaways and mobile food stands at about 45 per cent, unlicensed restaurants and cafes at 25 per cent, licensed restaurants at 18 per cent and event catering at 12 per cent. It refreshes fortnightly on a fourteen-day mailing window from the date of registration. Two artwork versions were produced because roughly one in six registered offices is the founder's home address.
7 August 2026
Six-month advertising review, £539,515 of spend examined
Three findings, all against the account we manage ourselves. A campaign named Exact contained £11,251 of broad and phrase matching. The Mid Margin band fails its own definition, so part of £78,573 is being spent below break-even by the account's own logic. And £230,820, 42.8 per cent of the budget, sits in campaigns with no margin band at all and is therefore optimised to revenue rather than profit.

Section 6

Channel by channel

Each row is one route by which customers reach the shop. Before means the nine months to 31 July 2025. With Maitland means the nine months to 31 July 2026.

Revenue and spend by marketing channel, before and during the Maitland engagement
Channel	Before	With Maitland	Change	Note
Google advertising	£555,484 spent	£761,698 spent	+37.1% spend	Spend rose by more than a third on an approved budget. Clicks rose 13.9 per cent and impressions 21.1 per cent, so each click cost more: £1.51 to £1.82.
Google-sourced revenue	£960,932	£1,275,597	+32.7%	Up, but part of this is Shopify reclassifying where orders came from between the two years rather than genuine growth. Treat the size with caution.
Microsoft advertising revenue	£135,456	£149,832	+10.6%	Revenue up, orders down 9.9 per cent, clicks down 32.2 per cent. The channel was switched off for six months spanning the start of the engagement and relaunched in February with a tracking fault.
Paid social	£3,508	£137,257	2.03 times return on the shop's own order data	Effectively a new channel. See the paid social section below.
Email	£112,594	£122,954	+9.2%	Up on an attributed basis. Attributed means the email was the last thing someone clicked, not that the email caused the sale. Proving cause needs a holdout test that has not been built.
Direct	£749,704	£832,649	+11.1%	Revenue up 11.1 per cent but the number of orders is flat at plus 0.5 per cent. Growth is coming from larger orders, not more of them.
Organic search revenue	£396,123	£353,348	-10.8%	Down. This figure excludes June, which contained an exceptional one-off event described in section 7.
Affiliate and cashback partners	£134,962	£127,224	-5.7%	Down about six per cent. The main affiliate network stopped being tracked in February 2025, eight months before this engagement began, so the exact figure is uncertain.
Artificial intelligence assistants (ChatGPT and similar)	£8,285	£29,743	3.6 times	Real growth, and still only 0.46 per cent of all revenue. Forty-eight orders out of 14,883. It existed before this engagement; the analytics tool simply had no category to put it in until June 2026. Over 96 per cent of it is ChatGPT alone.

What this means: most routes into the shop are carrying more revenue than a year ago, and paid social is effectively new. Two are down: unpaid search results and the affiliate and cashback partners. The Google figures are the least reliable in the table, for the reasons given in the notes.

Section 7

Paid social: from a rounding error to a working channel

Spent on Meta advertising

£4,310 → £67,592

Revenue on the shop's own order data

£3,508 → £137,257

The same channel, measured three ways

Paid social measured three ways, before and during the engagement
Measure	Before	With Maitland
Spend	£4,309.58	£67,591.55
Impressions	746,497	8,591,729
Orders on the shop's own data	7	371
Revenue on the shop's own data	£3,508.32	£137,257.02
Return on the shop's own data	0.81 times	2.03 times
What Meta claims it made	£480,679	£792,535

Before this engagement, paid social was not being run as a channel. Three campaigns spent £4,310 across nine and a half months, which is about £15 a day, and two of those campaigns belonged to the previous agency. On the shop's own order data it returned £3,508 against £4,310 of spend. It lost money.

It is now a channel. £67,592 spent, 8.6 million impressions against 746,000, and 371 orders against seven. On the shop's own order data it returned £137,257, which is £2.03 for every £1 spent.

One warning the board should hear from us rather than from anyone else. Meta's own reporting claims it produced £480,679 in the first period and £792,535 in the second. Neither figure is real. Against the shop's till, the first is overstated by a factor of 137 and the second by a factor of six. This is because Meta counts a sale if a person saw or clicked an advert at any point beforehand, and the setting controlling that was much looser on the old campaigns than on the new ones. If anyone opens the Meta dashboard they will see the return apparently falling from 111 times to 12 times. That is a change in how it is counted, not a change in what happened. We use the shop's own order data throughout, on both sides.

Section 8

Search visibility

Distinct search terms the site appears for

35,090

44,937

+28.1%

Search terms actually earning a visit

3,794

2,700

-28.8%

The site is being shown for far more searches than a year ago, and earning visits from fewer of them. Put simply, we have built reach without depth: the new pages appear, but too far down the results page to be clicked.

Visits from searches that do not include the company name, which is the part search work owns, are down between 25 and 40 per cent depending on how much of Google's withheld data is allocated. The generic category terms have moved together: chafing dish from 94 visits to 18, commercial microwave 82 to 13, commercial dishwasher 80 to 22, commercial fridge 63 to 14. That uniformity across unrelated categories points to something affecting the whole site rather than category-by-category competition. It has not yet been tested, and it is the defined piece of work for the next term.

The one section that grew is the Content Hub, the site's guide and advice section. It produced 707 visits and 139,101 appearances in search across the nine months, against effectively nothing for the twelve months before. That is 1.3 per cent of all search visits, so it is a small, real and growing base rather than a headline.

Visits each month from unpaid search results

What this means: this line is the number of people arriving each month by clicking a normal, unpaid Google result. It has drifted down, with one exceptional spike in June 2026.

Search Console only holds sixteen months of history, so no figure exists before May 2025. June 2026 is the exceptional month described in the next section. August 2026 is eighteen days.

Section 9

June 2026, and what it tells the board

The three days in June that made £245,277, and what happened next

21 to 23 June, the three days of peak demand
All other days

What this means: each bar is one day of sales of cooling equipment. Sales rose sharply for three days, then stopped almost completely because there was nothing left in stock to sell.

On 21 June the number of people searching for cooling equipment rose roughly twenty-five fold, from about 130 searches a day to 3,155. Nobody in this room caused that. It was the weather.

What the business did with it is the point. Conversion rates rose across every channel at once on 21, 22 and 23 June. Over those three days the cooling category took £245,277 on 501 orders. Across the whole of June it took £345,219 on 712 orders, against £32,340 in June 2025.

On 23 June the category sold out. Revenue went from £63,732 on the 23rd to £442 on the 24th. That is not demand collapsing, it is a shelf emptying: order counts fell from 251 to 158 to 92 to one, while revenue per order rose, which is what happens when a catalogue sells down to its last few variants.

Demand did not stop. Search visits to those pages peaked on 24 June, the day after the stock ran out, and stayed elevated for a fortnight. Those visits earned roughly £3,900 between them, because there was nothing to sell. The equipment did not return.

Two conclusions the board should take from this. First, the marketing responded to a demand event inside 48 hours and converted it hard. Second, the constraint was not marketing, it was stock: the season ran at roughly two and a half times the previous year and the category was exhausted in three days. That is a buying and forecasting question, and it is worth asking now rather than next June.

Section 10

What has gone backwards

This section is here because a report that only contains good news is not worth reading. Every item below is measured on the same basis on both sides of the comparison.

Advertising is costing more per pound of revenue.

9.80p to 11.82p, 20.6 per cent worse.

For every pound the business took last year, Google advertising cost 9.80 pence. This year it costs 11.82 pence. Cost per conversion rose from £54.24 to £82.56. Three pieces of context, and the board should have all three. First, the whole of the movement is the price of a click. The number of clicks needed to produce each pound of shop revenue was effectively unchanged, moving 0.17 per cent. Advertising delivered visits in line with sales growth. It simply paid more for each one. Second, and this is the part that matters: the price rise happened before this engagement, not during it. Cost per click on this account went from £1.31 in November 2024 to £1.87 in July 2025, a rise of 43 per cent across the year before Maitland started. Since November 2025 it has been flat: £1.90 in November, £1.89 in July. The year-on-year gap narrows every single month, from 46 per cent in November down to 1 per cent in July. A market-wide price rise arriving this year would look like a step up across all nine months. This looks like a rise that had already happened. Third, we are not going to tell you the market explains it, because we cannot evidence that. Two credible industry studies of click prices over the same period disagree sharply, one saying about 1 per cent and the other about 14 per cent, and neither covers the United Kingdom specifically. What we can say is narrower and true: within this year, the price we paid was flat while one of those two studies recorded the market rising. What is ours to answer: the cost of advertising on the company's own name rose from 42 pence a click to £1.99. That is £40,129 more than it would have cost at last year's rate, and it is about a third of the whole efficiency movement. Industry data says brand click prices did not rise. We are investigating a change in February 2026 where that campaign started matching to searches it should not have, and we will have the answer before the September audit.

Cost per click, month by month

The grey line climbs 43 per cent across the year. The green line is flat. The two nearly meet by July.

Fewer people who add to basket go on to checkout.

77.1 per cent to 69.2 per cent.

This is the most serious finding in the report. It is measured in a way that cannot be affected by robot traffic, so it is real. The entire deterioration sits in a five-month window from November 2025 to March 2026, bottoming at 60.6 per cent in February, and it has recovered to between 73 and 77 per cent since April. That window is the first five months of this engagement. The cause has not been established. Candidates under investigation are an inconsistent warranty prompt on some product pages, shipping cost appearing late at checkout, and business customers going away for internal sign-off. We are not picking one before we can prove it.

Search visits from terms that do not include the company name are down.

Between 25 and 40 per cent.

Covered in section 6. The site is shown for more searches and clicked from fewer.

Average order value fell.

£447.11 to £440.87 over nine months, and £466.21 to £420.17 on the matched August window.

Discounting is not the cause: discounts and returns actually fell as a share of sales, from 2.25 to 1.55 per cent. It is a change in what people are buying. On the August window, orders rose 33.6 per cent while revenue rose 20.4 per cent, so volume is being won at a smaller basket.

Microsoft advertising went dark and came back with a fault.

Six months off, then a tracking failure.

The channel was effectively switched off from August 2025 to January 2026 on £399.71 of spend, three of those months inside this engagement. It was relaunched in February 2026 with a tracking fault that recorded zero orders against £2,107 of spend and 1,652 clicks. Across the full nine months of each period it ran at 4.49 times return then 4.19 times: worse, not better.

Our own measurement is slightly worse than when we started.

93.2 per cent of revenue arrives with no campaign label attached, against 92.8 per cent before.

Google advertising links carry no manual tracking tags, so nine months of paid Google revenue cannot be isolated in the shop's own order data at all. We have held this account for the whole period and have not fixed it. The fix is approved work, not a discovery.

Section 11

What we cannot prove

There is one thing this report cannot tell you, and no report from any supplier could.

Nothing in this data separates revenue that advertising caused from revenue that would have happened anyway. Every figure here is an attribution: it records what someone last clicked before buying, not what made them buy. That is true of our figures, of Google's own figures, and of any figure produced by any agency, any software, or any artificial intelligence.

The only instrument that answers the question is a holdout: withhold advertising from a matched half of the country for six weeks and measure the difference. We have designed one. It costs roughly £8,354 of withheld spend and around £60,068 of foregone revenue, it takes six to eight weeks, and it will not exist by November unless it starts now.

We are recommending you commission it, in full knowledge that it could show our advertising is worth less than we believe.

Section 12

What we would do next

1
Commission the advertising holdout test, starting now.
It is the only way to answer whether the advertising pays for itself.
2
Find and fix the basket-to-checkout fault.
It broke for five months and recovered, and we still do not know why. Establishing the cause protects against a repeat.
3
Rebuild the category pages.
This is where the generic search terms live and where the visits have been lost.
4
Fix the advertising tracking tags.
It closes the measurement gap in section 8 and makes Google reconcilable against the shop's own orders.
5
Move the advertising to a margin basis rather than a revenue basis.
£230,820 of spend currently has no margin band attached.
6
Take the stock findings to the buying conversation.
£780,522 of demand this year sits on lines that cannot currently be sold, and 14 of those 35 lines have no purchase order raised.

Section 13

What we need from eCatering

Open questions and their owners
#	What we need	Owner
1	Why did the affiliate network stop tracking in February 2025?	Andrew / Andrea
2	Did anything run on or around 23 and 24 June 2026, such as press, radio or a partner post?	Michael / Andrew
3	When did the warranty prompt on product pages go live?	Mark / Tony
4	Is there a purchase order position on the two cooling lines, and is the 2027 date a decision or a supplier constraint?	Chris / Duncan
5	Definitions for GP1 and GP2 (the two gross profit measures). Every margin figure is directional until they exist.	Chris
6	Does Google's reported conversion value include Value Added Tax? It moves every figure by twenty per cent.	Andrew, with Google
7	Approval to start the holdout test.	Andrew
8	Add shipping cost and delivery time to the product data feed.	Chris / Michael

The cleanest comparison we have: 1 to 18 August, both years

Completed work by type of task

The 1,671.65 recorded hours, by type of work

Hours recorded each month

Northern Ireland customs capture built into checkout

Bank reconciliation traced for the finance team

Marketplace order reconciliation

Warehouse dead stock analysis

Bestseller stock position analysed

Product bundles built from the data

Demand measured on products that cannot be bought

Stock exposure quantified and taken to the buying team

A test designed that could prove our own advertising does not pay

New business prospect list built from Companies House

Six-month advertising review, £539,515 of spend examined