Companies ask many people what they think of brands. They ask every year in Latvia, Lithuania and Estonia. The answers come as big spreadsheet files. The person who works with the numbers made each report by hand.
BC Tool turns those files into charts people can read. One year’s survey goes in as a file. It comes out as four views, or screens of charts. A brand client or an agency researcher can use them without opening the file. Preparing a report went from 4 hours to 1.
The time went into steps done by hand. BC Tool does those steps now. The rules are set once. The same question can be asked again next year.
- Client
- Magic Agency
- Product
- BC Tool, brand survey charts for Latvia, Lithuania and Estonia
- My role
- AI and product adviser. I helped choose what to build and shaped the written plan. My team and I built the tool.
- Focus
- Making reports
Result
- 75%
- less time to make a report
- 4 h → 1 h
- time for one report, before and after
- 51
- categories in 3 countries and 4 yearly surveys
“Eddy was a key partner in taking our analytics platform from an early idea to a fully working product. He helped us define the requirements and feature set, structure the PRD, and turn our ideas into a clear product direction, and then, together with his team, brought that vision to life.”
Where the four hours went
The person who works with the numbers did every step by hand. Open the file. Find that year’s columns. Adjust the numbers so the group matches the whole country. Split the answers by group. Write the slides. A new question meant doing it again.
Before the tool, the survey lived in spreadsheet files and fixed reports. Another person working with the numbers could use a different rule on it.
The survey covers 51 categories in Latvia, Lithuania and Estonia. Each year’s file can lay out its columns its own way. Column names and decimal marks can change. So can the endings for what people think of the brand. The adjusting numbers may be called Weights_local one year and weight_local the next. Commas may sit where a database expects dots.
The questions are not all counted one way. How close people feel to the brand has five steps: strangers, deniers, acquaintances, friends, lovers. Each person who answered is on one step per brand. The five parts add up to 100%. For what people think of the brand, people can tick several phrases. Those parts do not add up to 100% and must not be read that way. Some questions count the two highest answers. Some use a how-often limit the reader picks. Some exist in only one country.
What the tool does instead
BC Tool keeps the rules inside. The person picks the question. Each step replaces a job the person who works with the numbers did by hand.
People who did not build the file can look things up on their own. That is the goal. The reader decides what to look at. The tool decides how the number is made.
Loading one year’s survey
An admin loads the year’s spreadsheet file. The tool checks it against that year’s guide to where each answer sits in the file. It finds the columns for how close people feel to the brand and what they think of it. It saves the people who answered and their answers in that year’s tables. The column hunt happens once per survey, at loading.
Adjusted numbers
One country uses its local adjustment. More than one country uses the global one. The chart the reader sees is already adjusted.
How close people feel, what they think
Take the people on one step and divide by everyone with a usable answer. For a phrase, take the people who picked it. Both use adjusted numbers. Picking another chip, a small button, redoes the ranking and the line across years.
Comparing groups of people
A group of people is compared with everyone who answered, labeled society. Scales start the way researchers already read them. For competitors, each brand gets the same group and the same audience section. So a difference comes from the people, not from the counting.
Your own groups
Deep dive takes a question the person who works with the numbers used to filter in Excel. It saves it as a recipe: which question, which answers, and how the rules join.
Adding a new year
Each year has its own guide to where each answer sits. The tool stays the same. Adding a year means a setup and a file load, not redrawing the four views. A bad file can be swapped without touching the other years.
New questions no longer wait
A question used to wait for the person who works with the numbers. Now it is a change of chip, country, brand or year. The tool has four views.
Country, category, brand and year stay picked as the person moves. The survey’s own words for the chosen number stay on screen. The raw column codes stay off it.
Brand Results
It ranks every brand in the category by the chosen number. It shows the line across years.
Brand Audience
It sets one brand’s group of people, lovers at first, next to everyone in the survey. It covers basic facts about people, attitudes, activities, interests and media. Special parts cover cars, personal money, the brand as a workplace and brand messages.
Us vs. Competitors
It keeps the same groups of people. It puts the chosen brand next to the brands around it.
Audience Deep Dive
It lets a researcher build a group. Pick a question. Pick the answers. Join the rules with “and” or “or”. Then see the brand through that group.
Why you can trust the numbers
The reader can disagree with a result. They should never doubt how the percent was made. It must be made the same way as in the slides three months ago. Trust here comes from the counting, not a sentence.
No made-up totals
For how close people feel, each person is on one step. For what people think, a person can pick several phrases. Adding those up would make a total the survey never collected.
Comparing only what was asked
A question shows in a year comparison only if every chosen year asked it. Humor was asked in 2023 and 2024. Relevance was asked in 2025. Standing out and the brand as a workplace arrive in 2026. A view of 2024 to 2026 will not draw likeability on one line and leave another blank.
Adjusting stays in its own country
A brand not asked about in a country does not borrow that country’s adjustment.
Labels people can read
Labels come from the rules and the guide to where each answer sits. They use words, not column codes like Z01_26_3. The reader sees the brand, the question and the percent.
A plan for empty screens
An empty category, a missing brand, a category with one brand and a year put away each get a planned empty screen.
Nobody changes the answers
Survey results are not changed to fit the chart. Admins load a survey and publish it. They do not rewrite anyone’s answers to change a result.
Who uses it
The same tool does two jobs. A client looks at one brand. An agency works across the whole list of brands.
A brand client
The tool opens on their brand. They see a category only if allowed. They explore their audience, competitors and line across years, not the rest of the survey.
An agency account
It explores any category, any brand and any of the three countries. It sees only allowed years. It is one tool, not a folder of spreadsheet files per client.
Admins
They make accounts, give roles, choose who sees which category, load a survey and publish it.
The result: 4 hours to 1
Preparing a report went from 4 hours to 1. That is 75% less time.
The tool now does the steps the person who works with the numbers did by hand. It finds that year’s columns, adjusts the numbers and splits the answers by group. One set of rules covers Latvia, Lithuania and Estonia.
A client can see how close people feel to their brand. They can see who the lovers are, next to society. They can see how that group differs from competitors. No new work has to be ordered. An agency can do the same for any brand it may open, with the same rules. A new year does not make a new tool. It adds to the one in use.
The tool stays inside the survey. It does not guess the future, write the questions or collect answers. The spreadsheet file is still where the numbers come from. It is no longer where the decision is made.
What this project teaches
Put the rules in the tool
The time went into rules followed by hand. People can look things up on their own when the percent is set once. The two highest answers, adjusted numbers, one step per person and questions asked only in some years belong in the tool. The reader is free to pick the question.
The first chart is the key moment
The key moment is the first time a person sees why a tool is useful. Here that is the first adjusted chart with the survey’s words beside it. No request goes to the person who works with the numbers. Loading, checking and the rules make that moment work for a client who never saw the file.
Expect each year’s file to differ
Column layouts change. The tool should handle each year’s layout and refuse comparisons the survey cannot back up.
One brand is not the whole list
One locked brand is a different product from the full list of brands. Both may still read the same tables. Who is allowed to see what is part of getting it right. The wrong brand on screen is a wrong result.
Add years, not new tools
Four views and one set of rules are enough for 2023 through 2026. The next file should fit into those same rules.