A business map is a drawing of your brand's parts and the arrows between them. Each part is a thing you track, such as ad spend or stock. Each arrow says one part moves another. Draw it by listing the questions you ask most, naming the parts, then adding arrows you can test against your own history.
A 9% sales dip. Three smart people. Three different explanations.
The agency says it is the season.
The ops lead says a best seller was out of stock for ten days.
Ravi, who owns this made-up lip balm brand, Little Fern, is sure a rival cut its price.
All three could be right. Each person sees one corner of the business.
You have likely sat in that meeting.
A business map is a drawing of your brand’s parts and the arrows between them. It writes down what each person carries in their head, so the group can check it.
This page shows what goes on a map, how to draw one in an afternoon, and a test to run before you trust it. Every number below is illustrative.
What a business map is made of
A map has only two ingredients.
- Parts. Things you can count or point at: ad spend, units in stock, price, new buyers, repeat buyers, a rival’s price.
- Arrows. A claim that one part moves another: “more stock means more sales, until stock runs out.”
That is all. The skill is in choosing the parts and being honest about the arrows.
Here is a small map for Little Fern. It is a sketch, not a full model.
Solid arrows are things you are sure of. Dashed arrows need a test. Illustrative parts and links for a made-up brand.
Look at the map and Ravi’s meeting changes. Stock, price and ads all point at orders. So “why did orders dip” has three suspects, and each one can be checked on its own.
Three kinds of arrow
Not every arrow deserves the same trust. Label each one.
| Kind of arrow | Example | How sure can you be? |
|---|---|---|
| Definition | Net revenue is sales minus refunds | Certain, because you wrote it |
| Plain arithmetic | Orders times average price is revenue | Certain, if the parts are right |
| Observed pattern | Ad spend lifts site visits | Only as sure as your test |
The first two are rules you wrote down. The third is a claim about the world, and the world can change. A pattern that held last spring may be gone by autumn.
Stanford’s classic guide to building an ontology (a shared, written set of terms and how they relate) starts the same way. It says to write down the terms you want to talk about and what you want to say about them (Stanford ontology guide). A map is a friendly version of that. For a deeper look at the word, see what is AI-ready data.
How to draw one in an afternoon
You need about three hours, one person who knows the numbers and one who knows the shelf.
Step 1: Start from your questions, not your tools
List the ten questions you ask most. Examples:
- Why did sales move last month?
- Which product keeps the most money?
- Will I run out of stock before the next shipment lands?
- Which ads bring buyers who come back?
The same Stanford guide suggests this move. Sketch a list of questions the model should answer, and treat them as the litmus test later (Stanford ontology guide). Your map is done when it can answer every question on the list.
Step 2: Name the parts
For each question, write down the parts it needs. “Why did sales move” needs orders, price, ad spend, stock and rival prices. Merge duplicates. Aim for 15 to 25 parts.
Give each part one plain definition. “Orders” means orders placed, not shipped. Write the basis down now, because two systems will quietly use two meanings later. Our post on product hierarchy shows how to do the same for products and packs.
Step 3: Group the parts
Put parts into a few areas. Money, selling, products, market, customers and habits is a sensible set, and a brand brain uses it. Groups help you see what is missing. If the market area is empty, you have no way to explain a rival’s move.
Step 4: Add arrows, then mark each one
Draw an arrow only when you can finish this sentence: “When A goes up, B goes up or down because…” Then label it as a definition, arithmetic or a pattern.
Mark every pattern arrow with one more thing: what else could explain it. Ad spend and orders both rise in November. So does everything else.
Step 5: Mark the gaps
Put a grey box wherever a part exists in real life but you have no number for it. Product cost is the usual one. A map with grey boxes is honest. A map with every box filled and half the numbers guessed is not.
Why the arrows need a test
Two lines on a chart often move together by luck. The Australian Bureau of Statistics says a correlation does not automatically mean one variable causes the other. It adds that a controlled study is the best way to show cause (Australian Bureau of Statistics).
You can rarely run a perfect study. You can run a cheap one on your own history. That is the test below. For paid ads in particular, our guide to incrementality vs attribution explains why the platform’s own report is not enough.
Run this test: trace one change through the map
This takes about an hour. You need twelve months of weekly numbers for the parts on one arrow.
- Pick one pattern arrow. For example, “more ad spend means more site visits.”
- Find the clean weeks. List weeks where the first part moved by more than 20% and nothing else big happened. No sale, no stockout, no price change. Aim for at least five.
- Check the second part in those weeks. Did site visits move the same way?
- Check the quiet weeks. Find weeks where ad spend was flat. Did visits stay flat?
- Write a verdict next to the arrow, with the date.
What a pass and a fail look like:
| Result | What you see | What to do |
|---|---|---|
| Pass | In four of five clean weeks the second part moved the same way. In quiet weeks it stayed flat. | Keep the arrow. Re-test it in a quarter. |
| Weak | It moved in two of five clean weeks. | Keep it dashed. Look for a hidden third part. |
| Fail | It moved the wrong way, or moved in quiet weeks too. | Remove the arrow, or rewrite the claim. |
A fail is a good result. It means the map just saved you from a wrong decision.
If you have fewer than five clean weeks, say so on the map. Thin evidence is still evidence, but label it. Amazon offers one free tool that helps here: Amazon Attribution, which Amazon says gives insight into the on-Amazon impact of marketing on non-Amazon channels (Amazon Ads). It turns the arrow from an off-Amazon ad to an Amazon sale from a guess into something you can measure.
A worked example
Back to Ravi. Illustrative numbers.
His map has three arrows into orders: price, stock and ad spend. He checks each one against last month.
| Suspect | What the data says | Weight on the 9% dip |
|---|---|---|
| Season | Same month last year was flat | Small |
| Stock | Best seller was out for ten days, about 6% of monthly orders | Large |
| Rival price cut | Rival cut a price in week three, but Ravi’s price rank did not change | Unclear |
The finding: the stockout explains most of the dip, and the agency never saw it because stock sits in a different system from ads. Nobody missed it. The two numbers lived in two tools.
It also has a cost. Ten days out of stock on the best seller, at an illustrative $1,200 a day, is $12,000 in sales. The reorder was placed late because the weekly report only showed sales, not days of cover. The map did not decide anything. It made the question smaller. “Why did sales fall” became “why did we run out”, which has an answer.
Where a business map goes wrong
- Too many boxes. A 200-part map is a wall. Start small.
- Arrows from belief. If the loudest person drew every arrow, test each one before it counts.
- No dates. A pattern needs a date, so you know when to re-test it.
- Missing parts read as zero. If you have no data on rivals, the map should say “unknown”, not “no rivals”.
- A stale map. A map drawn once and filed is a photo of last year.
Where Synthesis fits, and where it does not
You do not need software to draw a first map. Paper is fine, and doing it by hand teaches you where your data is thin.
The hard part comes after. Keeping the map current across stores, ads, email and books, checking every number against its source, and keeping the AI you already use on those numbers is steady work. Synthesis is ontology software and an AI harness in one plan. It writes the business down once, so every number means one thing, and it keeps Claude or ChatGPT on that map. Synthesis has no chat of its own. You ask in the AI you already use.
If your brand has one channel and a few products, paper may be enough for a long while.
The short version
A business map is parts and arrows. Start from the questions you ask most. Name the parts, give each one a plain definition, and label every arrow as a definition, arithmetic or a pattern. Test the patterns on your own history and write the verdict beside them.
Do it once this week. Even a rough map ends the meeting where three people each own a different corner of the truth.
Questions people ask
What is a business map for a consumer brand?
It is a drawing of the parts of your brand, such as ads, sales, stock and repeat buyers, with arrows showing which part moves which. It turns what you carry in your head into something you can check and share.
How is a business map different from a dashboard?
A dashboard shows numbers side by side. A map shows how they connect. A dashboard says ads and sales both rose. A map says whether you believe one moved the other, and what else could have.
How many parts should a business map have?
Start with 15 to 25. Fewer than that and the arrows say nothing new. More than that and nobody reads it. You can always add detail to one area later.
How do I know if an arrow on my map is real?
Test it. Look for past weeks where the first part changed and nothing else did, then see if the second part moved. If you can find no such weeks, mark the arrow as a guess.
Do I need software to draw a business map?
No. Paper, a whiteboard or a spreadsheet is enough to start. Software helps once you want the map kept current, checked against fresh numbers and shared with the AI you already use.
Can AI draw my business map for me?
It can draft one from your exports, but it cannot know your costs, your deals or which arrows are only guesses. Treat its draft as a starting list of questions, then fix it with what you know.