← workwithyaro.com

What I found, and what I built

Three businesses, anonymised. The numbers are real.

Each case is written the way I deliver work: what was actually wrong, what I built, what it changed, and where I could be wrong. The first one is a full review, exactly the document you get for $2,000. The other two are shorter write-ups of builds.

Written up afterwards rather than published at the time, which means you also get to see whether the diagnosis held.

Case 1 - full review
AI Business Review

Findings and recommendations

Prepared for
R., founder (anonymised)
Business
DTC supplements, fertility
Revenue at review
$12,000 / year, flat
Prepared by
Yaroslav Zhyvylo

Worked jointly with a design partner who owned brand and creative; I owned the catalogue analysis, the tracking, the offer structure and paid acquisition.

1Summary

#FindingCost of the problemEffort
0Nothing is being measuredHides everything below2 days
1Sold as items, not as a courseUp to $160 per one-bottle buyer1 week
2Ads cannot scaleCapped at $9,600 gross profit a yearWith 0
3Reorders left to memory$64 a month per lapsed customer2 weeks

Do not spend money on: a redesign, an app, or more products.

2Where you are

You run a direct-to-consumer supplements brand in the fertility category: one product line, sold online. Revenue is $12,000 a year and has been flat.

Your brief to me was "we need better advertising." This report explains why that is not the first problem.

3Findings

Finding 0 - precondition

Nothing is being measured

This is not a finding. It is the reason the other findings were invisible to you.

You could not tell me which products drive your revenue. Not "weren't sure" - there was no way to know. Orders exist; the link between a product, its margin and its acquisition cost does not.

So every decision in your business is being made on impression. Including the decision to spend more on advertising.

Cost of the problem
Not directly measurable - that is the problem. It is the reason findings 1-3 were invisible, and every advertising dollar you have spent so far was spent blind.
Fix
Basic product-level and channel-level tracking.
Effort
Two days. Nothing else in this document could be written without it.
How I could be wrong
If the numbers exist somewhere I was not shown - a spreadsheet, your payment processor's reports - this is a process problem, not a tracking one. The fix is still the same two days.
Finding 1

You sell items. Your product is a course.

Fertility supplements are taken over months. Your customer's actual need is "give me what I should be taking", not "here is a bottle".

Your store sells single units. Every purchase decision is made from scratch, every time, by a customer who does not know what combination they need.

With tracking in place, the pattern is immediate: a small number of SKUs carry your revenue, and customers who buy more than one buy the same combinations.

One bottle                  $80
Three-month course          $240
Lost per one-bottle buyer   up to $160
  of which gross profit     up to $128
Cost of the problem
Every order is smaller than it should be, and every repeat purchase requires the customer to re-decide. Up to $160 of revenue, $128 of gross profit per customer who buys one bottle and stops.
Fix
Restructure your catalogue into bundles that match how the product is actually used.
Effort
One week, mostly merchandising, not engineering.
How I could be wrong
If the combinations are an artefact of how your store is laid out rather than of real customer need, bundling will just re-label existing behaviour and change nothing. I checked this by looking at whether multi-item orders are consistent across traffic sources. They are.
Finding 2

Your advertising cannot scale because it cannot be measured

With no attribution, there is no way to know whether an advertising dollar returns two or zero. So you kept spend small - which was the correct decision given the information you had.

The constraint is not budget. It is that nobody can see the result.

Your unit economics
Average order$80
Gross margin80%$64 contribution before ads
Breakeven ROAS at 80% margin125%

That last line matters more than it looks. At an 80% gross margin, advertising only has to return $1.25 for every $1 spent to break even. That is a very forgiving threshold, and it means you can afford to buy customers aggressively - as soon as you can see what you are buying.

Cost of the problem
At $12K a year of revenue, your gross profit is at most $9,600 a year - that is the ceiling you are stuck under. Once measured, ads returned 360%: every ad-acquired order contributes about $42 after ad spend.
Fix
None of its own - it came with finding 0. From there, the cost is the ad spend itself, which returned $3.60 per $1.
Effort
Included in finding 0.
How I could be wrong
360% was measured at the spend levels we actually ran. Returns usually fall as spend grows. The number to watch as the budget scales is the 125% breakeven, not the 360%.
Finding 3

Your reorders are left entirely to memory

Your product is consumed monthly. Reordering depends on your customer remembering, returning to the site and deciding again.

For a consumable, that is the single most expensive thing a store can leave to chance.

Cost of the problem
A returning customer brings $64 of gross profit a month with no ad spend attached. Every one who forgets to reorder costs you that $64, every month.
Fix
A subscription tier.
Effort
Two weeks.
How I could be wrong
I was. Take-up came in well below what I expected - see section 7.

4In what order

  1. Tracking first (two days). Without it, nothing below can be checked, including whether it worked.
  2. Bundles next (one week). The cheapest fix and mostly merchandising - no engineering risk.
  3. Scale acquisition only once tracking proves ads clear the 125% breakeven.
  4. Subscription last (two weeks). The most engineering, and it only pays once there are customers coming in to offer it to.

5What not to spend money on

Worth as much as the list above.

Do not redesign the site. It is not pretty. It is also not the problem, and a redesign would consume your entire budget while changing nothing measurable.

Do not build an app. We discussed it; the answer is no. At $12K a year there is no audience to put in an app, and it would cost more than your annual revenue.

Do not add products. The obvious instinct when revenue is flat is to sell more things. Without tracking, more SKUs will only make the picture noisier.

6What happened - added afterwards

BeforeAfter
Revenue$12,000 / year$6,000 / month
Change+500% (6X)

The campaigns that survived testing ran at a sustained 360% ROAS - 2.9X above the breakeven threshold for this margin structure. Blended across the whole ad budget, tests included, it was 240%. About a third of the budget went to testing what to sell: courses against single bottles, categories against one brand, and positioning against competitors. That spend was meant to return nothing on its own, and it did not.

Per order
Revenue per order$80.00
COGS at 20%-$16.00
Gross profit$64.00
Advertising ($80 / 3.6)-$22.22
Contribution after ad spend$41.7852% of revenue

The subscription tier recovered acquisition cost on the first shipment, with about $42 left over. Every subsequent month from a subscriber is close to pure contribution.

The yearly figure
BeforeAfterGap
Revenue / year$12,000$72,000$60,000
Contribution$9,600$37,400$27,800

That is what your business was leaving on the table: about $60,000 a year of revenue, $28,000 a year of contribution. "After" is the run rate at the end of the engagement, annualised; the engagement itself ran just over four months. "Before" contribution is 80% gross margin with no ad spend, so it is the most it could have been. "After" is 52%, which prices every order at the 360% campaign rate. Counted against the whole ad budget, tests included, it is 38%, about $18,000 a year. Subscription reorders carry no ad cost, which pulls the other way. The truth sits between those two.

What it cost you
Your ceiling$3,500 / monthall in, set by you before we started
Review and build$1,000 / monthjust over four months, a fixed fee agreed before this review format existed
Ad budget$2,500 / monthabout a third of it on testing; already deducted from the contribution figures above
Payback on the fee2-3 monthstwo at 52% contribution, three at 38%

What I cannot tell you: how the $60,000 splits between bundles, advertising and subscription. They shipped close together and I did not measure them separately. Anyone who gives you a precise split for changes made at the same time is guessing.

Where it stopped

After a little over four months, one of the main suppliers stopped allowing its products to be sold under its own brand and required every store to rebrand. We set out what the next stage would take: a larger ad budget, a move into social media with content about motherhood, and the owner herself as the face of the brand. That is a different business from the one she had, and it asks for her time every week, not ours. She decided not to run it. That was her call, and a reasonable one: the review is there to price the options, not to push the owner into the one that needs the most of her.

7What I got wrong

Subscription take-up was about 8%. I expected more.

The offer was built, but it was not engineered. It was presented as an option alongside the one-off purchase, at the point of first purchase - which is the hardest possible moment to ask for a commitment in a category where the product takes weeks to show any effect.

What I would do differently, in order:

  1. Offer the subscription after the first order, around day 25, when the bottle is running out and your customer has decided whether they believe in the product
  2. Make it the default selection at checkout rather than an alternative
  3. Frame it as the course - "your three-month program" - not as a recurring purchase of a bottle

Brands that engineer this properly reach 20-40%. Eight percent means it was offered, not designed.

The honest counterweight: even at 8%, subscribers generated a meaningful share of revenue, because a subscriber pays every month while a one-off customer pays once. And it made inventory planning possible against a known floor of demand for the first time - which for a physical product is worth more than it sounds.

End of case 1Back to all cases
Case 2 - build
Build and pilot

Selling partner products next to rentals

Business
Fashion rental marketplace
Engagement
Build and pilot
Pilot scope
1 of 12 product categories
Built by
Yaroslav Zhyvylo

1What they needed

The marketplace wanted to test a new revenue line: selling products from partner brands alongside its rentals.

Done by hand, that test is expensive before it proves anything. Someone has to turn every partner catalogue into listings, write every description, and decide which customer should see which product. That is a team, hired for an idea that might not work.

2What I built

One system that does the manual part end to end:

We switched it on for one category out of twelve. A pilot that small is cheap to stop if it doesn't work, and gives a clean number if it does.

3What happened

Pilot conversion5%
Revenue, first month$8,500one category
Revenue per converting customer$170
Converting customers, first monthabout 50$8,500 / $170

A new revenue line that did not exist before, from one category, without a single listing written by hand. Eleven categories were still switched off.

4How I could be wrong

One category, one month. Other categories may convert better or worse, and a first month can carry curiosity that fades. Multiplying $8,500 by twelve categories would be the wrong way to read this. The right next step is two or three more categories, measured the same way.

End of case 2Back to all cases
Case 3 - review and fix
Review and fix

A hosting bill that kept growing

Business
E-commerce platform
Engagement
Review, fix, hardening
Cost before
$2,500 / month
Built by
Yaroslav Zhyvylo

1What was wrong

The store was paying $2,500 a month just to keep the site running.

Everyone involved was trying to fix the obvious causes. None of them touched the reason.

The real cause was in the architecture - in how the site was built, not in how much server it was given.

2What I did

3What happened

BeforeAfterChange
Cost per month$2,500$600-76%
Cost per year$30,000$7,200~$22,800
Page speed score7095+25

About $22,800 a year back, every year, from a fix that was paid for once.

4What the numbers don't say

Page speed is a score, not revenue. The saving above is the figure this case rests on; the faster site is a bonus, not the claim.

End of case 3Back to all cases

Why this is the format

Every case above carries the same things: what was wrong, what it cost, what fixing it cost, and how I could be wrong about it.

The last one is the one that matters. A result you cannot argue with is not analysis, it is a sales pitch. If I tell you where I might be mistaken, you can check me - and if I turn out to be right anyway, you have a reason to believe the rest.

Your review won't look exactly like case 1. For a store it's usually about margin and reorders; for a firm or a service business it's usually about hours - who types what, how often, and what that costs you a year. The format is the same either way.

Want one of these for your business?

Five working days. $2,000, fixed, credited against the build if you go ahead. You keep the report either way.

Book a 30-minute call Free, 30 minutes, no pitch.