How to Tell Whether Positioning Is Improving Conversion
Positioning is improving conversion when one conversion step you fixed in advance rises for prospects who heard the new version, compared with prospects who heard the old one in the same weeks, and when those buyers also name the alternative and repeat the difference the new version states. This applies to founders near launch with a steady flow of comparable prospects. Tag every prospect by version, follow each for a fixed window and treat small counts as a direction, not proof.
The outcome and observation window
This guide is for founders before or around launch who have changed how they describe the product and want to know whether the change is moving buyers. Writing positioning against the buyer's current alternative is covered in How to Position a Startup Against the Customer Current Alternative, testing a sentence in a batch of calls in How to Test Positioning in Real Sales Conversations and diagnosing a buyer's retelling in Why Buyers Understand Your Product but Cannot Explain Its Value. This page is about the numbers that follow: which conversion step to watch, over what window, against what comparison, and when the answer is good enough to act on.
Pick one conversion step that positioning should move and write it down before anything changes. Good candidates sit right after the buyer first hears the positioning: a first call that turns into a second meeting, a quote that turns into a signed order, a demo that turns into a paid trial. Steps far downstream, such as renewal, are moved by too many other things to read positioning from them. Define the numerator and the denominator in words: for example, prospects who received a quote, and of those, prospects who signed within the follow window.
The follow window is how long each prospect is followed after entering the step. Set it from the buyer's usual time to decide, taken from deals already closed, with some margin. Count by cohort: every prospect belongs to the version they heard and the date they entered the step, and counts as converted only if they convert within the window. Counting all signings in a calendar month instead mixes buyers who heard the old version with buyers who heard the new one.
Analytics tools apply their own windows, which may not match the sales cycle. Google Analytics documentation says that 'the key event lookback window determines how far back in time a touchpoint is eligible for attribution credit', with a default of 30 days for acquisition key events and 90 days for all other key events, and shorter options available. A page visit older than that window gets no attribution credit for a later conversion. For a long or offline sales cycle, keep the version tag on the prospect record instead of relying on page attribution.
Decide the observation period too: the dates during which each version is in use, and the minimum number of prospects per version before reading the result. Write both down with the date. Without a minimum, it is tempting to stop the moment the numbers look good.
Current customer alternative: definitions and evidence
A conversion rate alone cannot show that positioning caused a change. Price, lead source, season and the founder's own practice move it too. So log, for every prospect, whether the mechanism the new positioning relies on actually appeared. The first part of that mechanism is the comparison point: what the buyer compares the product with.
Define it before the observation period starts. Write down the alternative the new version names, such as doing the work in-house, a spreadsheet, a supplier's free service or doing nothing, and record per prospect the alternative the buyer names in their own words, without prompting. Count it as present when the buyer compares the product with that alternative. Count it as absent when they compare it with a rival the new version does not position against, or with nothing.
Evidence for this measure comes from conversations and written answers, not from analytics: call notes written as spoken, the buyer's words in an inbound form that asks what they use today, or objections heard in a quote meeting. Daniel's go-to-market page says, in its list of what a session works through: 'Choose what to record each week: conversations, agreed next steps, paid pilots and reasons deals stop.' The comparison each buyer names belongs in the same weekly log.
Meaningful product difference: comparison and caveats
The second part of the mechanism is the difference: whether buyers who heard the new version repeat the one thing it says the product does that the alternative cannot. Log it per prospect as present or absent, the same way as the comparison. It is the signal that ties a change in conversion to the words that changed.
Then compare the versions in a way that holds everything else still. A before and after comparison is the weakest design, because anything that changed between the two periods changes the result. The stronger design runs both versions in the same weeks and splits prospects between them by a rule fixed in advance, such as alternate days, alternate territories or alternate names on a list. When volume is too low to split, keep a dated change log of everything else that moved: price, lead source, segment, offer, the person selling and outside events.
Season is an easy confounder to miss. The US Census Bureau describes seasonal adjustment as 'the estimation of the seasonal component and, when applicable, also trading day and moving holiday effects', and says the goal is usually to produce 'series whose movements are easier to analyze over consecutive time intervals'. A startup with a few dozen prospects cannot run that kind of adjustment, but the point carries over: two neighboring periods are not automatically comparable. Running both versions in the same weeks is the practical substitute.
Small counts are the second caveat. The NIST/SEMATECH e-Handbook of Statistical Methods says the normal approximation to the binomial can be used to compare two proportions 'if the samples are reasonably large', and that 'the Fisher Exact Probability test is an excellent choice for small samples.' In its worked example, 2 of 7 items in one group and 3 of 5 in the other fall in the same class, and the one-tailed test gives p = 0.31060, so the handbook does not reject equal proportions. Rates that look far apart on a dozen prospects can still come from the same underlying rate.
Proof the buyer understands the offer: scenarios and review triggers
The third part of the mechanism is proof the buyer understands the offer: the buyer can say what they would get, for what price and what happens next, or asks a question that only makes sense if they do, such as when it starts or who else needs to approve it. Log it per prospect like the other two.
Read conversion and the three mechanism signals together. Four scenarios cover the combinations:
- Conversion up, mechanism up. The new version is the likely cause. Continue it and keep counting until the planned minimum.
- Conversion up, mechanism flat. Something else moved the number. Check the change log for price, lead source, season or a new seller before crediting the positioning.
- Conversion flat, mechanism up. Buyers understand and repeat the new version, but something after understanding stops them: price, timing, the budget owner or the offer itself. Change that, not the words.
- Conversion flat, mechanism flat. The new version is not landing. Check whether it names the alternative these buyers actually use, or go back to the retelling diagnosis.
Set review triggers in advance, so the measurement is read again at the right moments rather than whenever the numbers feel interesting:
- each version reaches the planned minimum number of prospects;
- price, offer, segment, lead source or the person selling changes;
- a season or a market event begins or ends;
- the share of prospects naming the intended alternative falls below its pass line;
- conversion moves sharply in either direction within a single week.
Daniel's go-to-market page says, under the step 'Review weekly and feed it back into the product': 'The weekly read refines the ICP, the positioning, the sales collateral, and the roadmap.' A review trigger is a reason to read the numbers outside that weekly rhythm.
The positioning measurement grid
Copy this grid into a spreadsheet, add one column per version and fill it in from the prospect log. Write the pass lines before the observation period starts.
| Measure | How it is counted | Compare the versions by | What a change does not prove | Pass line to write in advance |
|---|---|---|---|---|
| Conversion step | Prospects who converted within the follow window, out of prospects who entered the step | Same weeks, same lead source, same price | That the words caused it, without the signals below | The gap between versions that would justify switching |
| Comparison named | Prospects who name the intended alternative without prompting | Share of prospects per version | That buyers prefer the product | The share that shows buyers use that alternative |
| Difference repeated | Prospects who repeat the one difference in their own words | Share of prospects per version | That the difference is worth paying for | The share that shows the difference is understood |
| Offer understood | Prospects who state what they would get, the price and the next step, or ask when it starts | Share of prospects per version | Demand | The share expected before the conversion step |
| Context held | Price, offer, segment, lead source, seller and season, logged with dates | Same or changed, per version | Anything, if it changed between versions | Every item unchanged, or the change noted next to the result |
| Count | Prospects per version at the time of reading | Against the minimum written in advance | A settled result, when counts are small | The minimum before any decision |
Read the grid only when the count meets its minimum or a review trigger fires. The rows, the reading order and the scenarios above are an editorial framework, not a validated instrument.
A hypothetical example: snow clearing for small shops
The following is a hypothetical example with invented details. It illustrates the measurement and does not describe a real company, customer or measured result, and it is not guidance on snow removal or winter safety.
A founder sells seasonal snow clearing to small shops: after every snowfall of 2 inches or more, a crew clears the shop's front walk and entrance before it opens, for a flat $640 a season. The founder visits shops in person, gives a quote on the spot and counts a conversion when the owner signs within 14 days of the quote. All signed contracts so far came within 9 days, so 14 days leaves a margin.
For the first 4 weeks the founder describes the service as 'snow plowing you book from your phone'. That version gets 32 quotes and 3 signings. In the notes, 24 of the 32 owners compare it with the landlord's plow contractor, who already clears the parking lot.
On the morning of the first snowfall, the founder walks the street at dawn and watches a hardware store owner push a shovel along the snow in front of his own shop, alone, before opening. That is the alternative the founder had not named. The new version reads: 'Instead of shoveling your own walk before you open, a crew clears the walk and entrance by 6 a.m. after every snowfall, for one price per season.'
The new version goes live that week, and its first 2 weeks bring 18 quotes and 8 signings, against 3 of 32 before. The change log shows why that comparison cannot be trusted: the new version went live the day after the first snowfall, while every owner on the street had just shoveled their own walk. So for the next 4 weeks the founder uses both versions on the same days, on alternate streets chosen in advance, and fills in the grid:
| Measure | Old version | New version |
|---|---|---|
| Quotes | 22 | 23 |
| Signed within 14 days | 3 | 9 |
| Compared it with | 16 with the plow contractor | 17 with shoveling the walk themselves |
| Repeated the difference | 4 | 15 said cleared before opening |
| Asked when the crew would start | 2 | 7 |
| Price, offer and seller | Unchanged | Unchanged |
The old version signs 3 of 22, about 14 percent, up from 3 of 32 in the weeks before the snow; the new version signs 9 of 23, about 39 percent. Part of the jump in the first comparison was the weather, which lifted the old version too. A one-tailed Fisher exact test on 3 of 22 against 9 of 23, worked out by hand with the handbook's method, gives p of about 0.054: given 12 signings in total, a split at least this lopsided toward the new version would turn up by chance about 5 times in 100 if both versions converted at the same rate. The founder's pass lines, written before the split, were a gap of at least 15 points and at least two thirds of new-version owners comparing it with shoveling. The gap is about 25 points and 17 of 23 owners made that comparison, so both lines are met, and 15 of 23 also repeated the difference.
The founder continues the new version, changes nothing else and sets the next review at 40 quotes per version or the first price change, whichever comes first. The test result stays recorded next to the decision: the gap is suggestive, not settled.
The assumptions stay written down: that the alternate streets have similar shops and similar snow, that signings after 14 days would not change the comparison, that 9 days is a fair guide to the time to decide, that the founder pitched both versions with equal care, and that a season with little snow could give a different answer.
Decide what to continue, change, or stop
Make the decision at the planned minimum or at a review trigger, against the pass lines written in advance:
- Continue when the conversion step meets its pass line in a same-weeks comparison and the comparison and difference signals meet theirs. Keep counting and set the next review.
- Change the offer, not the words, when the mechanism signals pass and conversion does not. Buyers understand the new version, and something after that stops them.
- Change the words when conversion moves but buyers do not name the intended alternative or repeat the difference. Find out what did move the number before building on it.
- Stop the new version when neither conversion nor the signals beat the old version at the planned minimum. Go back to which alternative these buyers actually use.
A continue decision is provisional. Record it with the counts, the test result if one was run and the date of the next review, so whoever reads the log later can see what the decision rested on.
Common mistakes and limits
- Before and after with no change log. Season, price or lead source can explain the whole difference.
- Counting by calendar month. Signings in a month mix buyers who heard different versions.
- Stopping when the numbers look good. Read at the planned minimum or a review trigger, not after the first good week.
- Watching only conversion. Without the mechanism signals, a rise cannot be credited to the positioning.
- Changing price or offer together with the words. Then nobody can say which change worked.
- Treating an analytics window as the sales cycle. An attribution lookback window is a reporting setting, not the buyer's time to decide.
Daniel's go-to-market page also says, under the heading 'Where founders get go-to-market wrong': 'Scaling outreach volume before a single message has earned a reply.' A positioning change deserves the same patience: scale the version that met its pass lines, not the one that looked good for a week.
The method has limits. Small counts give a direction, not a verdict, even with an exact test. Splitting by day, street or list keeps season and timing similar but cannot make the two groups identical. The grid, scenarios and decision rules are an editorial framework, not a measured result.
Related guides and next step
- The wider sequence by stage: How to Build a Startup Go-to-Market Strategy by Stage
- Choosing the alternative to position against: How to Position a Startup Against the Customer Current Alternative
- Filling in category, difference and value: Positioning Worksheet: Category, Difference, and Buyer Value
- Testing a sentence in a batch of calls: How to Test Positioning in Real Sales Conversations
- Diagnosing a buyer's retelling: Why Buyers Understand Your Product but Cannot Explain Its Value
Daniel's go-to-market page says the first session is free and takes 30 minutes. Book a startup mentoring session to review your startup positioning decision with your actual evidence and constraints, and bring the prospect log with each prospect's version, dates and the three signals.
Frequently asked questions
How can a founder tell whether positioning is improving conversion?
Fix one conversion step and a follow window in advance, tag each prospect with the version they heard and compare the versions in the same weeks. Credit the positioning only when conversion rises and buyers who heard the new version also name the intended alternative and repeat the difference. Small counts give a direction, not a verdict.
What evidence should a founder collect for startup positioning?
For each prospect: the version heard, the date they entered the conversion step, whether they converted within the follow window, the alternative they compared the product with in their own words, whether they repeated the difference and whether they stated the offer back. Keep a dated log of price, offer, lead source, seller and season changes next to it.
What are common mistakes in startup positioning?
Comparing before and after periods without a change log, counting signings by calendar month instead of by cohort, stopping when the numbers first look good, watching conversion without checking that buyers repeat the new framing, and changing price or offer at the same time as the words.
How long should a founder measure a positioning change?
Long enough for every prospect in the comparison to finish the follow window, which comes from the buyer's usual time to decide in deals already closed, and until each version reaches the minimum count written in advance. Read earlier only when a review trigger fires, such as a price change or the start of a season.