The Conversion Rate Is a Product of Several Proportions, and Diagnosis Runs from Upstream Down

The single figure called the conversion rate emerges from four or more proportions multiplied together. Looking at the product alone does not reveal which factor fell. A value of 0.2 appears when all four have dropped evenly, and it appears when only one has dropped severely.

Even so, what gets touched first when the conversion rate falls is in most cases the final scene. The structure of the proposal, the way price is presented, the closing words. The reason is simple: the final scene is the only part that remains inside your memory.

The only part you attended directly is that scene. The other party’s expression, the order of their questions, the phrasing of the refusal — all of it stays. How that person arrived, by contrast, was never in your memory to begin with. Being to hand and being where the cause lies are different things.

And the final scene is also the place with the narrowest room to move, and with large side effects when moved. The techniques available there work in the direction of shifting the timing of a judgement, not of increasing the material for it. Effects follow, and because they follow, the work goes on.

Three things get settled in this article.

The first is the structure of the figure itself. That it is a product rather than a standalone indicator, that being a product means only the lowest factor matters, and that the final value cannot identify which factor is low — all of this gets settled.

The second is the order of diagnosis. A procedure for applying four gates from the top down, and a method for judging each gate from the shape of the other party’s words alone. What the judgement requires is not analysis but a record of what they said first and what kind of question they returned.

The third is why the order tends to run backwards. This is not a matter of will or ability but of structure: records accumulate automatically only downstream. The presence or absence of a record settles, in advance, the range of hypotheses that can be raised. A region where no hypothesis can be raised was not examined and rejected — it never surfaced as a candidate for examination. Records accumulating only downstream is not confined to the scene of closing. How upstream and downstream divide is treated in marketing design as a whole.

Something that does not get settled belongs here too. This article cannot produce the value you should set as a target. How the denominator is taken differs across businesses, so even within one business the figure moves several-fold when the counting changes. Handing over a value amounts to handing over a way of counting.

And this article does not line up methods for raising the conversion rate. The same figure comes out of two opposite designs. It rises when selection is strengthened upstream, and it rises when persuasion is strengthened downstream. The two are indistinguishable at the level of the indicator, and the latter worsens the state that follows completion. Lining up the methods would erase that distinction.

The Remedies on Offer Never Touch the Denominator

The widely shared remedies are all accurate as practical advice. What is notable is that they gather in one place.

The conversion rate is the proportion of cases, among those where a meeting or proposal took place, that reached completion. What is taken as the denominator differs by business; that the numerator is the count of completions is common to all.

The remedies run roughly as follows. Make the sales process visible. Narrow the audience. Set targets. Standardise the procedures and techniques. Use closing techniques. Added to these are grasping the other party’s requirements accurately, building a relationship of trust, and following up carefully after contact.

As a rule of thumb, figures are sometimes offered stating that the average runs somewhere between thirty and fifty per cent, and that one completion per two or three proposals is standard.

That these are apt becomes clear as soon as you act on them. Without making the process visible, you cannot see where things halt. Without grasping requirements, the content of a proposal will not fit.

The place where they gather is the denominator. Not one of the measures on offer touches it. They deal only with what happens after a meeting has occurred; who arrived at the meeting is not treated as a subject of examination.

As a division of labour, this is correct in certain settings. Someone not responsible for the inflow can only work on the people who arrive. Where the conditions for that split hold, there is no need to doubt it.

The conditions stop holding when one person oversees the whole business. In a sole operation, inflow and meetings are designed by the same person. Yet because the advice is shaped around a division of labour, the denominator survives unexamined.

The rule-of-thumb figures carry a problem of the same kind. A value of thirty to fifty per cent depends entirely on how the denominator is taken. Counting every enquiry, or counting only those whose conditions were confirmed, moves the same business’s figure several-fold. Holding an incomparable figure as a benchmark means the judgement of whether your own figure is high or low never gets off the ground.

Holding an incomparable benchmark pushes judgement away from the actual state. When your figure falls below the benchmark, work begins on finding the cause of the shortfall. The work begins in exactly the same way when the only difference lies in how the denominator was taken.

Work that has begun finds something to improve. Improvable points exist in every business, so a search will always produce one. Act on the point produced, and the figure moves a little. The movement is then treated as evidence that comparison against the benchmark was valid.

Does that mean, without a benchmark, you cannot tell whether your figure is good or bad? Only comparison against outside benchmarks is unavailable; judgement is not. Compare against your own past figures. As long as the denominator is taken the same way, this comparison holds. Being unable to compare outward does not amount to being unable to judge.

The Conversion Rate Is the Share That Passed Every Independent Gate

The conversion rate is the proportion that passed through several independent gates, all of them. It is not a single proportion but the product of several proportions arranged in series.

There are at least four gates.

First, whether the people arriving actually hold the problem you handle. If people who do not hold it arrive, every gate after that carries no meaning.

Second, whether they are at the stage of handling that problem now. Someone who holds the problem but ranks it low will not decide, however well the content fits.

Third, whether what you can hand over is understood as corresponding to their problem. Where that understanding has not formed, judgement proceeds on peripheral cues.

Fourth, whether the conditions match. Amount, timing, format. Only here does what people call negotiation occur.

The four pass-through rates multiplied together give the conversion rate. Because it is a product, if any one is severely low, the whole stays low even when the others are perfect.

This structure carries a consequence for diagnosis. The final value cannot identify which factor is low. The same final value does not settle a single way in which the factors fell.

Measure the four separately, then. Split them and you can identify. But splitting requires recording the passage of each gate. Only the last gate records itself automatically, as whether completion occurred. The remaining three leave nothing unless you record them deliberately.

The asymmetry of the records is what tilts diagnosis downstream. The only figures to hand belong to the last gate, so analysis is performed on the last gate. What gets analysed is settled by the presence of records, not by the analyst’s will.

The asymmetry of records is also an asymmetry of time. The last gate leaves an amount and a date the moment it is passed. The upper three do not rise as events at all. That someone holding the problem arrived leaves no trace anywhere unless you define it and record it; the person passes through unmarked.

And what leaves no trace cannot be recovered later. What stage a person was at three months ago can only be guessed at if it was not recorded. A value filled in by guessing matches the guesser’s expectation. From a value that matches the expectation, an error in the expectation cannot be detected.

Cutting the gates into four is not the only division. The number and content change with the shape of the transaction. Some shapes need two; some split into six. What does not change is that the final value is a product.

Being a product carries a practical meaning. When you set out to raise the whole, where you place your hand changes the effect by an order of magnitude. The difference in effect can be confirmed in numbers.

In a Product, Only the Lowest Factor Matters

The nature of a product settles the priority. The largest room for improvement always lies in the lowest factor.

When four factors stand at 0.9, 0.9, 0.9 and 0.3, the whole is 0.22. Raise the 0.3 to 0.6 and the whole becomes 0.44. It doubles. Raise any of the 0.9s to 1.0 and the whole reaches only 0.24.

The difference is more than tenfold. If the same quantity of work is to be spent, where it goes is settled. Up to here it is arithmetic, and there is no room for dispute.

The difficulty lies in knowing which one it is. The lowest factor is usually also the least measured. Because it is unmeasured, its being low is not visible. What is not visible does not rise as a candidate for priority.

When those two overlap, a characteristic plan appears in practice. A plan to improve every factor evenly. As an allocation of work it looks fair. As an allocation of effect it is severely inefficient.

An even allocation appears only when you do not know which factor is low. If you knew, there would be no reason to make it even. So the appearance of an even plan is itself readable as evidence that nothing has been measured.

Products carry another asymmetry. Raising a high factor higher barely moves the whole. Taking 0.9 to 0.95 is in fact rather hard. You take on hard work, and the whole moves by one per cent.

And that hard work comes with a sense of accomplishment. Making the structure of a proposal precise carries a clear feeling of progress. The size of the feeling and the size of the contribution run in opposite order.

What if two or more factors are low? The order runs from upstream. Upstream factors form the denominators of the downstream ones. If more people pass the first gate, the actual counts passing the second, third and fourth rise too. The reverse does not happen.

Fixing upstream, however, can lower the downstream proportions. If more people pass through, more of them fall away within that group. The proportion falls and the actual count rises. Watching the proportion alone records the upstream improvement as a failure.

This misreading occurs only immediately after upstream work. Since the timing is known, it can be prevented. Do not compare this month’s downstream proportion against last month’s when upstream has been touched. Compare the actual counts. How long the proportion takes to settle depends on how long passage takes. For a transaction where judgement takes weeks, the proportion takes weeks to settle. Move downstream before that period has passed and the upstream effect and the downstream operation land in the same month, after which the two cannot be separated.

What Gets Touched First Is the Scene You Recall Most Vividly

Diagnosis begins downstream not through negligence or misunderstanding. That scene is simply the one remembered most vividly.

The meeting is the one event you attended directly. The detail of that scene stays to hand, and the route that led up to it was never taken in.

The quantity of memory tilts the inference of cause. A scene you can recall in detail offers many elements available for nomination as the cause. A process you cannot recall offers none. When the number of candidates differs, the winning side is settled.

Improvement furthest downstream usually produces some effect. Tidy the structure of a proposal and the conveying improves. Because an effect appears, it becomes a confirmation that the direction was right.

As confirmations accumulate, the reason to doubt upstream diminishes. While improvement continues downstream, examination of upstream stops being treated as an untouched area and becomes an unnecessary one. This shift never occurs as a judgement. It simply drops quietly out of the set of candidates.

Sometimes the cause does lie downstream. A proposal’s structure can genuinely be broken; the way conditions are presented can genuinely create misunderstanding. Looking downstream is not the error. Beginning downstream and stopping downstream is.

Upstream is a route you built yourself, so surely you know who is arriving — this view has exactly one hole. You know the intended audience. Whether the people actually arriving match that intention is separate information.

And the two drift apart on their own. What you write reaches a wider range than the writer assumed. The direction of the drift is also usually the same. Those with a more general interest are more numerous, so inflow rises from that side. The intended audience is mixed in as a minority.

This state cannot be detected while you watch the overall figure. It gets detected as a sensation during meetings: this was not the person I expected. The sensation accumulates as individual memories and never becomes information about composition.

Beginning downstream has one advantage: confirmation of effect is fast. That advantage is real, so there is no need to reject beginning downstream as such. What should be settled in advance is the procedure for moving upstream when downstream produces no effect. Settle the procedure first and moving is no longer a judgement made in the moment.

Prior Proportions Lose to the Case in Front of You

This tilt has a footing in psychology.

In 1973 Daniel Kahneman and Amos Tversky showed that people judging probability tend to discount prior proportions (Kahneman & Tversky, 1973, Psychological Review, 80(4), 237–251). In the experiments, participants were given a description of a person drawn from a group whose occupational composition was known, and asked which occupation the person held.

Participants’ estimates barely reflected the composition of the group. Judgement proceeded on which occupation the description resembled. The authors set out this resemblance-based work as judgement by representativeness.

The same shape appears in diagnosing the conversion rate. The content of a recent meeting is a vivid description of a person. Infer a cause from it and the inference follows the content of the description. The prior proportion — what composition of people is arriving in the first place — does not enter the judgement.

In practice it appears like this. When price was mentioned in two of three refusals, an inference forms that price is the cause. The inference is strong because it carries a sensation. The possibility that those three came from a segment not yet at the budget stage, by contrast, is not examined.

What this research shows is a tendency in judgement observed under experimental conditions, not a claim that all business judgement takes the same shape. Real business judgement differs from the experimental setting in that it comes with repetition. Where there is repetition, prior proportions can accumulate as experience. What carries over is only the direction: the vividness of the description to hand settles the weight it receives in judgement.

For accumulation to occur, though, there is a condition: the composition of arrivals must be recorded. Without records, repeated experience accumulates only memories of meeting scenes. As long as what accumulates is skewed, repetition does not correct the skew.

A cause built from an individual scene carries the property that its treatment is designed for individual scenes too. If price being mentioned is inferred as the cause, the way price is explained gets improved.

The improved explanation gets used in the next meeting. If the next person is at the budget stage, the explanation works. If not, it does not. When it does not work, an inference forms that the explanation was still insufficient.

This route has no end. Thickening an explanation can be continued indefinitely. And the thicker it gets, the longer it feels to someone not yet at the budget stage. When it feels long, they leave. To the extent that they leave, the remainder consists of people at the budget stage.

The figure improves as a result. The reason for the improvement is not the quality of the explanation but selection through departure. The two are indistinguishable in the numbers.

Applying Four Questions in Order, Upstream to Down

The four gates are applied from the top down. What the judgement uses is only what the other party said. Neither your analysis nor the impressions left in memory enter this process.

The first question. Do the people arriving actually hold the problem you handle?

The way to confirm is to look at what the arriving person says first. If they begin with a description of a problem, they hold it. If they begin with a question about your service, they are looking for a means rather than a problem. Someone looking for a means is at the stage of lining up comparisons.

The second question. Are they at the stage of handling that problem now?

The way to confirm is the presence of a deadline. If a phrase about by when comes from their side, they are at the stage. If it does not, the problem is recognised but the priority has not been fixed.

The third question. Is what you can hand over understood as corresponding to that problem?

The way to confirm is the content of the questions they return. If how would this work in my case comes back, understanding has formed. If how does it compare elsewhere and how much of a track record is there dominate, judgement is being attempted on the comparison side.

The fourth question. Do the conditions match?

Only on arriving here does the talk turn to amount and timing. If the upper three have passed, negotiation at this stage is about scope. If they have not, negotiation is about the total. Which form the negotiation takes indicates the state upstream.

The four share a property. Each can be judged from what the other party said. Rather than inferring through your own analysis, it is enough to look at the shape of their words.

That property lowers the burden of recording. Note only the first thing said and the kind of question returned, for the four items, and the composition can be read afterwards. Detailed minutes are not required.

Because words shift with the flow of a conversation, is that enough to judge on? One case cannot be judged. Line up ten and a distribution of shapes appears. A distribution is stable beyond the fluctuation of individual cases. What is being judged is not each meeting but the composition of the people arriving.

The point at which the four questions are applied is not after the meeting has ended. Recalled and written afterwards, what gets written is the impression. The first utterance and the kind of question returned can be written in a line each, on the spot.

Confirming in this order does not guarantee that the cause narrows to one. Several gates can be low at once. Failing to narrow is not itself a failure. If the range has moved from looking only at the furthest downstream to one of four, diagnosis has advanced.

Records Exist Only Downstream, So Hypotheses Form Only There

Why examination of upstream gets postponed can be explained from the side of records rather than of will.

Downstream, records remain automatically. Whether completion occurred remains as the occurrence of a transaction. The amount remains, the date remains. These remain without any decision to record them.

Upstream, only what you ask about and write down remains. Which problem the arriving person held, which stage they were at, what they expected. These vanish unless asked about and written down.

This asymmetry settles the input to analysis. Analysis can only be performed on what has records. A region without records does not surface as an object of analysis. Because it does not surface, the hypothesis that the cause lies there never forms.

Trying to record upstream increases the work. The return on the increase does not appear immediately. Records only become readable as a composition once a certain quantity has gathered.

So the decision to begin recording appears, at that moment, as work with no effect whatever. The effect can take months to arrive. Throughout that interval, downstream improvements go on producing effects within the same period. Set side by side, downstream always looks more correct.

The worry that recording case by case is only possible while the scale is small is well founded. But it runs the other way, and becomes a reason to begin. Only while the scale is small can records be taken case by case. Attempt it after the volume has grown and there is too much to start on. And if records exist from when it was small, what the later aggregates mean can be read.

There is no need to record every item. If you have a sense of which of the four gates is suspect, recording that one alone suffices. Beginning one item is faster than delaying the start in an attempt to record all of them.

The object of recording carries the same asymmetry. Records are needed not only for the parties who completed. Records about the parties who did not complete are what work for diagnosis.

Yet records remain more readily for those who completed. Once a transaction begins, information about that party accumulates naturally. With those who did not complete, contact ends there. From that moment, nothing remains unless it was written down.

So records tend to skew toward the completed side. Read composition from skewed records and what can be read is the characteristics of transactions that closed. The characteristics of transactions that did not close are not in there.

Judge on the remaining side alone and judgement aligns toward reproducing the conditions for remaining. Strengthen what the closed transactions share and transactions of the same kind increase. The range that does not increase stays absent from the records as well.

What Only the Final Scene Can Do Is, in Fact, Very Little

Operations downstream carry two properties. Narrow room to move, and large side effects.

Narrow room means there is a limit to the range that can be moved. Tidying the structure of a proposal improves the conveying. But however much the conveying improves, if what is conveyed does not correspond to that person’s problem, the judgement does not change.

Large side effects follow because operations downstream touch the quality of judgement directly. The techniques used in the final scene — phrasings that prompt a decision, the order in which options are presented, the presentation of a deadline — all act on the timing or the direction of a judgement.

The structure in which judgement is settled by a reason supplied from outside was handled in product launches. The techniques furthest downstream perform the same operation in a one-to-one setting.

They have an effect. Because they do, they continue. And as they continue, the composition of parties who complete changes. The proportion who judged on content falls, and the proportion who judged on the scene rises.

That the composition has changed does not appear in the conversion rate. It appears afterwards. Mismatches once the transaction begins, interruptions midway, the absence of a next request. Each is recorded as an indicator separate from the conversion rate.

Not every downstream technique distorts judgement. Putting into words what has not been settled, showing in advance what happens next, organising the information a judgement needs and handing it over. These raise the quality of judgement.

The dividing criterion is whether the operation increases the material for a judgement or moves its timing. Operations that increase material work upstream; operations that move timing work only downstream.

That distinction has a practical use. Operations that increase material can be moved outside the meeting. The work of organising and handing over the information a judgement needs is done if it is handed over beforehand as text. Movable operations are upstream processes.

Operations that move timing can only be performed on the spot. They cannot be handed over in advance. Immovable operations are the ones native to the furthest downstream.

So counting what only the furthest downstream can do gives a very small number. Most can be moved upstream. To that extent, the judgements performed during the meeting decrease. In the scenes that remain, the one deciding is the other party.

Moving work upstream carries the cost of lowering the sense of accomplishment. A meeting with someone whose material is already in place ends in confirmation alone. A meeting that ends in confirmation carries no sensation of having done something. Without the sensation, that process looks as though it was unnecessary.

The Conversion Rate Rises the Same Way Under Two Opposite Designs

The conversion rate is not an indicator that should simply rise.

Among the four gates, the first — whether the people arriving actually hold the problem — carries a function of selection. Where selection works, those who are not the audience stop there. Those who stop do not reach a meeting.

Depending on how the denominator is taken, the conversion rate rises as selection works. If the denominator is those who reached a meeting, and selection happened upstream, nearly everyone who arrives is the audience.

Conversely, weaken the selection and the conversion rate falls, because people who are not the audience reach a meeting. That fall is not a deterioration. The judgement has merely moved downstream.

So placing the conversion rate as a target makes the figure rise both by strengthening selection upstream and by strengthening persuasion downstream. The two are opposite designs, and the indicator cannot tell them apart.

The observation that strengthening selection raises the rate but lowers the count is correct. But a falling count and a falling total volume of closed business are different things. Time spent on meetings with parties who are not the audience is freed. If that time goes upstream, the number of people arriving rises.

This does not happen automatically. What the freed time gets used for is a separate judgement. Without a judgement, the freed time goes to downstream improvements, because that is what is to hand.

Nor is strengthening selection always correct. Selection upstream runs on criteria you hold. If the criteria are too narrow, they stop people who are the audience.

And those who were stopped never appear to you. Over-strengthened selection is observable only as a rise in the conversion rate. A rise looks like good news, so it is no signal that the selection is too strong.

The strength of selection therefore cannot be judged by the conversion rate. Judging it requires looking at the total count alongside. If the conversion rate rises and the count of completions rises too, the selection is appropriate. If only the rate rises while the count stays flat, the selection may be stopping the audience as well.

Among the four gates, the only one that may be placed as a target is the third. The proportion at which understanding forms has no side effect when raised. The first and second are selection, so the direction of raising or lowering shifts with the design of the business. The fourth is conditions, so raising it means conceding on conditions.

That only one gate can serve as a target largely settles what the conversion rate is useful for. The overall value becomes an object of observation rather than a target. Use it as an entrance for asking, when it moves, which of the four moved. Used that way, opposite designs producing the same figure is no longer a problem: you can go and confirm which design produced the movement.

Whether Diagnosis Advanced Shows in How Refusals Change

The confirmation runs on something other than the conversion rate. It runs on the shape of the refusals.

The first indicator is whether the distribution of stated reasons for refusal has changed. If the reasons before and after diagnosis are the same, the place you touched was not the cause. If they have moved to something else, the earlier gate now passes. Where they moved to is the next gate to look at.

The second indicator is whether negotiation occurs over the total or over scope. If negotiation over the total decreases and negotiation over scope increases, understanding upstream has begun to form. If negotiation over the total is unchanged, the third gate has not yet passed.

The third indicator is the number of contacts required to reach completion. If the number falls, the material a judgement needs is now gathering at an earlier stage. If the number is unchanged while only the conversion rate rises, the cause of the rise lies in the final scene.

Each of the three looks past the count of completions to where the process leading to completion changed. Since the conversion rate is a product of several proportions, what needs confirming is the side of the factors.

All three are qualitative observations rather than numbers — that reading is available. Yet all three can be counted. Reasons for refusal can be classified and counted. The object of negotiation splits into total or scope. The number of contacts is already a number. They are not hard to count; they simply are not aggregated automatically.

Make an indicator of only what aggregates automatically and the indicator is settled by the convenience of measurement. The conversion rate tends to become the main indicator partly because it is a figure that emerges automatically. That reason has nothing to do with its fitness.

The procedure for breaking the state of nothing is selling apart as a set of symptoms is handled in the diagnostic procedure for why nothing sells.

The value you should set as a target depends wholly on how the denominator is taken. Change the counting and even one business’s figure moves several-fold. The question of what percentage to aim for turns into the question of settling the counting before settling the value.

The product of four numbers is the single figure now in your hands. What the multiplication contains, and which parts of it are recorded and which are not — without separating those two, while you look only at the recorded side, the candidate cause does not move from the final gate.

So confirm once which of the four you are currently touching. If the answer is the fourth alone, what to confirm next is whether records exist for the other three.