solver
Value Creation Engine
Using the wrong tool is no different from solving the wrong customer problem.
Most founders run experiments. Very few diagnose them. A test fails, someone says it did not work, and the team moves on to the next idea. That sentence is the least useful one in a startup. It is a verdict without a cause.
A failed experiment is a symptom. Something specific broke, at a specific stage, and the evidence usually says where. The Value Creation Engine treats every result the way a good doctor treats a symptom: find the cause before you prescribe the next move.
01A failure has an address.
Every experiment passes through the same stages: a customer, a problem, a solution, a price, a channel, a window of time. When it fails, it fails at one of them. The work is to find which one, not to throw the whole bet away.
02Read the signals.
The pattern of a failure points to where it broke. These are the patterns I look for first.
Nobody engaged at all.
customer · problem
Keep the offer. Try a different customer, or a sharper version of the problem for the same one.
They engaged, but did not pay.
solution · price
The problem is real. Change the offer or the price, one at a time.
They paid once, and never came back.
solution depth
It solved the problem once. Test whether it keeps solving it.
It worked, but for people you did not expect.
customer
The market told you who it is for. Rewrite the customer and rerun.
It worked, but too slowly.
channel · time
Keep the idea. Change only how people reach you, or how long you wait.
03Change one thing.
The next test changes the stage that broke and holds everything else still. Change two things at once and you are guessing again, just more expensively.
04Decide before you know.
Before a test begins, write down the number that counts as success, the minimum, and the date. A number without a deadline is not a target, because almost anything grows if you wait long enough. Once the test starts, none of the three can move. A target set after seeing the result will always be met.
05Three tests deep.
Here is the engine reasoning through one bet, start to finish. The story is illustrative; the method is the point.
Test 1
Busy parents will pay for weekly meal plans.
- Locked
- 50 sign-ups in 14 days
- Result
- 6 sign-ups. Almost no clicks.
- Diagnosis
- Nobody engaged: the problem or the customer.
- Next
- Ten conversations with parents. They do not lack plans. They lack time to shop.
Test 2
Same parents, new problem: three dinners, pre-portioned, delivered.
- Locked
- 20 pre-orders in 14 days
- Result
- 31 pre-orders. Only 4 reordered.
- Diagnosis
- Paid once, never came back: solution depth.
- Next
- Keep everything. Change only the menu: rotate it weekly, allow skips.
Test 3
A rotating menu keeps them coming back.
- Locked
- 50% reorder within 4 weeks
- Result
- 58% reordered.
- Diagnosis
- Validated. Persist.
- Next
- Now, and only now, test the price.
Notice what did not happen. The founder never abandoned the parents, and never changed two things at once. Each failure narrowed the search, and the third test was aimed precisely because the first two were read correctly.
06The same discipline, everywhere.
I build Dharth to correct diagnoses in medicine. This is the same discipline turned on a company. Do not argue with the verdict. Find the cause. Change one thing. Check again.
A failed test is not an answer.
It is a symptom. Diagnose it.
Eleven stages, four moves
Frame
Experiment Customer Customer Problem
What are we testing, for whom, about which one problem.
Bet
Solution Assumption Defined MVP
What we offer, what we believe will happen, and the smallest way to find out.
Commit
Success or Pivot Timeframe
The number and the date, locked before a single result exists.
Judge
Result Decision & Action Validated Learning
What happened, persist or pivot, and what it means for the next test.
The eleven stages, in full
01Experiment
Every experiment you run is for the product. Use one row per experiment and map its flow across the canvas.
02Customer
Describe the target persona, and be very specific. Use generative AI if it helps, then print it, stick it on the wall, and define the person.
03Customer Problem
Define the specific, validated problem for that persona. Customers have 99 problems. Solve only one per test.
04Solution
The proposed solution to the problem in the column before it.
05Assumption
What you think will happen when you run the experiment. The test will either validate or invalidate it.
Micro LOF: the one thing you are willing to bet on for this experiment.
06Defined MVP
The quickest, cheapest, most efficient way to validate your assumption.
07Success or Pivot
Choosing the right metric is the most important piece. Define the minimum success metric: the criterion this solution must meet to count as successful.
If you validate your assumption, continue down the path. If your assumption was that solving X would bring you 200 customers and you get 200, the assumption is validated. If it is disproven, you pivot.
08Timeframe
Timeframe and Success or Pivot go together. Every metric needs a set timeframe: 50 page visits a day, 100 customers a day, 500 sign-ups a week. That is how you see cause and effect.
Without a timeframe the numbers can still grow on their own over time, but that growth explains nothing.
09Result
Record the results and how they compare to your Success or Pivot criteria.
Great products move forward by delivering specific value for real problems. Average products fall behind solving a problem the customer does not have, or building the wrong solution for the customer's real one.
10Decision & Action
Record whether you persist or pivot on the current path, and the action that follows.
The decision ties directly to the right metric you chose, and to the minimum success criteria for that metric within the timeframe.
11Validated Learning
The treasure you have harvested. In technical terms, what you have validated. These are the absolute and relative truths you carry into future decisions, and they set the criteria for your next test.