States, not tasks
Setting up your pipeline
Stages are the vocabulary your team argues in. Getting them right is twenty minutes and it determines whether anybody keeps the board current.
The test
A stage is a state, not a task and not a feeling
A pipeline in Consonas is a named list of stages you write yourself, each carrying a likelihood, with the history of every move kept. Nothing about it is fixed by us, which means the naming is the whole of the work.
Almost every badly configured pipeline fails the same test, and the test is one sentence: would two people in your organisation independently agree which stage a given deal is in?
If they would, it is a state and it is a good stage. Quoted is a state. Survey booked is a state. Awaiting decision is a state, and a useful one, because it distinguishes a deal that is waiting from a deal that is stuck.
If they would not, it is usually one of two things. It is a task, such as chase the customer or send the contract, which describes what someone should do rather than where the deal is. Or it is a feeling, such as hot or promising, which describes the opinion of whoever last touched it.
Tasks belong to a person with a date, against the record. Feelings belong in the likelihood on the stage, which is what the likelihood is for. Neither belongs on the board, and a board containing them will be applied inconsistently by everyone and will stop meaning anything within a month.
Five or six, and not twelve
Every additional stage is a decision someone has to make correctly under time pressure. Five stages applied consistently produce better data than eleven applied approximately, and the eleven stage board will look more sophisticated while telling you less.
The usual cause of twelve stages is that somebody has written down the steps of the process rather than the states of a deal. Those are different lists, and only one of them belongs on a board.
Applying the test
Six candidate stage names
Three of these are good stages and three are not.
| Stage name | Is it a state or a task? | Verdict |
|---|---|---|
| Quoted | A state. Two people would agree whether a quote has been sent. | Good stage. |
| Chase the customer | A task. It describes what somebody should do, not where the deal is. | Not a stage. Make it a task against the record. |
| Awaiting decision | A state, and a useful one, because it distinguishes waiting from stalled. | Good stage. |
| Hot | Neither. It describes how somebody feels about the deal. | Not a stage. Use the likelihood, or a field. |
| Survey booked | A state, if a survey is a real step in your process. | Good stage, if it applies to most deals. |
| Send the contract | A task, and one that belongs to a person with a date. | Not a stage. |
Examples
Stages that work in particular trades
These are starting points to adapt rather than templates to adopt. Use your words.
- A trade or service business
- Enquiry, quoted, accepted, booked, done. Five stages, and the highest value view in the whole product is quoted with no activity for ten days.
- A contractor tendering
- Invited, expression of interest, priced, submitted, shortlisted, awarded. The stage history here tells you which clients invite you repeatedly and award nothing.
- A training provider
- Enquired, information sent, taster booked, offer made, deposit paid, enrolled. Seasonal, so an enquiry sitting still for two months is normal instead of dying.
- An agency pitching
- Introduced, credentials, brief received, proposal, pitch, decision. Run renewals as a second pipeline because a renewal is not a pitch.
- A firm taking instructions
- Enquiry, conflict checked, quoted, client care sent, instructed. Short, and the conflict stage is the one that must not be skipped.
- A renewal pipeline
- Healthy, review due, in discussion, renewed, notice served. A different shape entirely: what you want is a list of accounts where something is wrong, early enough to act.
Those six shapes are written out at greater length on the pages for the trades they came from. Construction covers tendering, where the useful number is which clients invite you repeatedly and award nothing. Marketing agencies covers pitching and why a renewal belongs in a second pipeline. Education and training covers a seasonal board, where a stalled enquiry is often only the summer.
Sales describes what the stage history is then able to report, and reporting and forecasting describes the weighted total the likelihoods produce.
Would two people in your organisation independently agree which stage a given deal is in? If not, it is a task or a feeling rather than a state.
Which is the whole test, and it eliminates most bad stages in about a minute.
Building the board
Setting it up, and correcting it after a quarter
1. Write the states down on paper first
Before touching the settings. List the states a deal genuinely passes through in your business, apply the two people test to each, and cross out the ones that fail. You will usually be left with five or six.
2. Put them in order and name them in your words
The order should be the order things happen. The names should be what your team says out loud, because those are the words people will use when discussing a deal and any translation is a small tax on everyone.
3. Set a likelihood on each
Guess. Something rising from ten or fifteen per cent at the first stage to eighty or ninety at the last is a reasonable starting shape. These are placeholders and everybody should know they are placeholders.
4. Use it for a quarter without changing anything
Move deals when the world changes rather than when someone remembers, because the stage history is what you are collecting and it is only as good as the timing of the moves.
5. Correct the likelihoods from what happened
After a quarter, look at what proportion of deals that reached each stage actually closed. That proportion is what the likelihood should be. Replacing the guesses with the measurements is the single change that makes a forecast worth arguing about, and it takes ten minutes.
6. Look at where deals actually stall
The average time at each stage is the diagnostic. It is almost always surprising, and it very often contradicts the reasons people give for losing, because the reason people give is price and the stage history frequently says responsiveness.
Correcting it
What to change after a quarter, and what to leave alone
The first configuration is always partly wrong. The useful thing is knowing which parts to fix from data and which to leave.
Correct the likelihoods, from what actually closed
For each stage, what proportion of the deals that reached it eventually closed? That proportion is the likelihood. Replace your guesses with it.
Expect the early stages to be substantially lower than you set them and the late ones higher. Almost everybody is optimistic at the top and pessimistic at the bottom, which produces a forecast that is wrong in two directions and happens to look about right in total, which is why no one catches it.
Merge any stage nobody stays in
If the average time at a stage is under a day, it is a step somebody performs rather than a state a deal occupies. Merge it into its neighbour. Fewer stages that everyone applies consistently beat more stages that people disagree about.
Split a stage where the time is enormous and the outcomes differ
The opposite case, and rarer. If one stage holds deals for an average of nine weeks and the deals inside it are visibly in two different situations, there are two stages.
Leave the names alone
Renaming stages is the most tempting change and the least useful. It resets the comparison with the previous quarter for no gain, and the problem it usually attempts to solve is a definition problem rather than a wording one.
Then stop
Correct once, correct again after the second quarter, and then leave it. The numbers stabilise, and after that the useful comparison is not this quarter against a model, it is this quarter against the same quarter last year. A pipeline that is reconfigured every quarter has no history worth reading.
Conventions to agree
Five arguments worth having before you type a stage name
The stages are the visible part. These five conventions decide whether the board can still be read in a year, and none of them has an answer we can pick for you.
Whether the value goes on at the enquiry or at the quote
Every deal carries a value and a currency. The question is when a real number appears in that field. Put one on at the enquiry and most of them are invented, the total at the top of the board is inflated by however optimistic the team is feeling that month, and people quietly learn to discount it in their heads. Wait until you have quoted and the first two stages are worth nothing at all, so the forecast only begins to exist in the middle of the pipeline.
Most organisations of this size are better off waiting. A forecast that says nothing about enquiries is being honest about what no one knows yet. A forecast that says a great deal about them is a number no one trusts, and an untrusted number is worse than a missing one, because it still gets quoted in meetings.
Whichever you choose, say it out loud and have everybody do the same thing. Two people using two conventions produce a board whose total means nothing and where no one can work out why.
Whether the expected close date is a promise or a guess
The date on a deal does one job: it decides which period the value lands in. That makes it a planning number instead of a commitment, and treating it as a commitment is exactly how it stops being maintained.
A date pushed forward every fortnight because nobody wants to admit slippage carries no information. A date somebody is judged on will always be that date. Say plainly that moving it is expected, and that the honest record lives in the stage history, which records when a deal actually moved and who moved it, and which no amount of optimism can edit afterwards.
Decide as well what goes on a deal that genuinely has no date. Leaving it empty is allowed and it is the better answer. Empty says we do not know. A date three months out that someone invented says the same thing while pretending otherwise, and it puts a number into a period on the strength of nothing.
Whether one customer buying three things is one deal or three
This is settled by how you sell rather than by preference. If the customer can accept part of it, it is three deals, because one can be lost while the others close and a single deal has no way to record that. If it is one decision and one signature, it is one deal, and splitting it gives you three cards that always move together and inflate every count you take.
The case worth avoiding is the middle one: a single deal whose value covers three things, of which the customer buys one. Recorded as a win, that deal reports its full value, and the gap between what you forecast and what you invoiced becomes invisible at the exact moment it matters.
Who moves the card
The stage history records who moved a deal and when, so this convention leaves a trace you can read later. There are two workable answers and one that quietly fails.
The owner of the deal moves it, when the customer does something. Choose that one. The alternative, where whoever runs the weekly meeting moves everything during the meeting, works only in the sense that the board looks tidy afterwards: every timestamp then says Monday morning, the average time at each stage is wrong by up to a week, and the deals that most needed attention are the ones no one raised.
The answer that fails is no one in particular. That gives you a board which is accurate for whoever is conscientious and fiction for everyone else, with no way to tell which half you are looking at.
Whether your close reasons are a fixed list or a sentence
The reason is recorded as text against the closed deal, so either is possible and you should decide once. Five or six agreed wordings can be counted. Sentences cannot be counted, but they say what actually happened, including the thing a list would never have had a box for.
The compromise most people arrive at is an agreed wording followed by a sentence, in that order, so the first words are countable and the rest is readable. It costs twenty seconds per deal and it is worth more than any number on the board.
Whatever you settle on, do not let anybody record price by default. Price is what a customer says when they would rather not explain, and a pipeline whose losses are all recorded as price has learned nothing at all.
Edge cases
The deals that refuse to travel in a straight line
A board of five stages assumes deals enter at the first and leave from the last. Enough of them do neither that it is worth deciding in advance what happens to those.
The one that arrives already at the fourth stage
A customer you have worked with for years rings and asks for the thing they had last time. There was no enquiry, no qualification and no proposal, and the deal is entirely real. Create it and put it straight at the stage it is genuinely at.
Do not walk it through the earlier stages to make the history look neat. That writes four stage moves in one minute, and the average time at each of those stages is now wrong for a reason nobody will ever recover. A deal that entered at the fourth stage is honest data, and the stage history records where it started.
The one that comes back a year after you recorded it lost
You have a choice here and it matters. Reopening the old deal keeps everything in one place, and it also means that record now spans fourteen months, which makes every average it appears in strange. Creating a new deal keeps the timings clean and separates the two attempts, at the cost of the link between them being something a person notices rather than something the board shows.
Create the new one. The old deal, closed as lost with its reason intact, is the more valuable of the two records, because a customer returning a year later is evidence that the reason was timing rather than price. Overwriting it destroys precisely that.
The one that shrinks between the quote and the signature
A deal quoted at one number is accepted at a smaller one. Change the value before you close it as won, not afterwards. The value on a won deal is what every total is built from, and a board where won deals carry their quoted figure rather than their accepted figure overstates the thing you are most likely to be judged on.
This case also argues for correcting likelihoods from history rather than from feeling. If deals routinely shrink between the proposal stage and the win, that fact is already sitting in the difference between what the stage forecast and what closed.
The customer who appears to be at two stages at once
Common, and usually not a problem at all. Two deals against one customer at different stages is what the model is for and the board will show both.
What causes trouble is one deal that two people are working on with different beliefs about where it has got to, which shows up as a card moving backwards and forwards over a few days. That is an ownership problem rather than a stage problem. One deal, one owner, and the owner moves the card. Whoever else is involved puts a task or a note against the record instead.
The deal nobody will close and no one will lose
Every board grows a few. The customer has stopped replying, no one wants to record a loss, and the deal sits where it is for months, quietly holding up the forecast.
The stage that would fix this does not exist and should not be created. A stage called stalled destroys the one number worth having, which is how long the deal sat at the stage it actually reached. Agree instead a period after which such a deal is closed as lost, with a reason recording that there was no response, and apply it without a debate each time. If the customer reappears, the case above applies and you create a new deal.
A forecast full of deals nobody believes in is not a forecast. It is a wish with a total underneath it.
The one you never really had
Occasionally a deal turns out to have been an exercise: the third quote somebody needed for a procedure, a budget check, a tender written around a competitor. Record it and close it as lost with the reason. Deleting it takes it out of the denominator of every conversion rate you will ever look at, and a stage that reports a healthy conversion because the hopeless entries were quietly removed is worse than having no number.
If it happens often enough to distort the picture, the pipeline is not the problem. That is a qualification problem, and the fix is an earlier stage everyone is willing to lose deals at.
Where the number lives
Why the likelihood belongs to the stage and not to the deal
It is the one modelling decision on this page you cannot configure your way out of, and it is the reason a forecast here can be corrected at all.
A number on a deal is a feeling with a decimal point after it
Ask somebody how likely a particular deal is and the number you get reflects their last conversation, their mood and how the quarter is going. Ask again a fortnight later and it will have moved without anything having happened. Nobody is being dishonest. A single deal is a sample of one, and a person cannot hold a base rate in their head while looking at a customer they like.
Attach the number to the stage instead and it stops being an opinion about one deal and becomes a claim about a population: of the deals that got this far, this proportion closed. That claim can be checked against the stage history, and being checkable is the entire point of it.
What that costs, because it does cost something
Two deals at the same stage carry the same likelihood even when everyone in the room knows one is stronger. That is a real loss, and it is felt most sharply by whoever owns the strong deal.
What you get back is that the error is symmetric and cancels out across a board of any size, whereas optimism recorded deal by deal does not cancel, because it points one way. When a deal really is different, the honest fix is usually a stage that recognises the difference, or a second pipeline, rather than a number one person moved on one card.
The correction only works because the number is not yours to move
The instruction earlier on this page, use it for a quarter and then replace your guesses with what happened, is only possible because the number lives on the stage. If everyone set their own, there would be nothing to correct: the forecast would already be the sum of everybody's opinions, and comparing it with reality would tell you about the people rather than about the process.
It is also why the correction takes ten minutes. You are changing five or six numbers, not auditing several hundred.
Nothing here sets that number for you
There is no scoring and no prediction. The likelihood is a number you type and the stage history is a record you read. That is deliberate. A number produced by a model is a number no one in the room can argue with, and in an organisation of this size the argument is the useful part. A meeting where somebody says the proposal stage is nowhere near sixty per cent, and the history agrees with them, has just improved the forecast.
| Where the number could have lived | What it would measure | Why it does not live there |
|---|---|---|
| On the stage, which is how it works | The proportion of deals reaching that stage that historically closed. | It does. One correction a quarter fixes the whole board, because there are five or six numbers rather than several hundred. |
| On each deal, set by whoever owns it | How that person feels about that deal this week. | Nothing can check it, so nothing can correct it, and it drifts upward as a quarter ends. |
| On each deal, set by a model | A prediction drawn from history you have not accumulated yet. | A small organisation closes too few deals for a model to learn much, and a number nobody can question ends the argument that was the useful part. |
| Nowhere, with the forecast done by hand | Whatever the person doing the arithmetic believes on the day. | It works for one person and stops working the moment two people are asked what the total is. |
The customer's process
The same board, read from the chair opposite yours
Every stage name describes something you did. The customer is running a process of their own, and the two only occasionally line up.
They have a process too, and you are one step inside it
The person you are selling to sits somewhere in a sequence you cannot see: a need someone has agreed to, a budget that exists or does not, two other quotes, an approver who has not been told yet. Your board records your half of that. It records none of theirs.
The practical consequence is that a move on your board is not evidence of movement on theirs. Sending a proposal is something you did.
Quoted means nothing has happened yet
This is the stage most boards are most optimistic about, and from the other side of the table it is the moment when the work moves entirely to someone else. It commits the customer to nothing, and a proposal that has been read and quietly set aside looks identical on your board to one being circulated for approval.
If that stage holds a large share of your value and carries a high likelihood, correct the likelihood there first. It is usually the stage where the guess sits furthest from the history.
The silence usually belongs to somebody who was never in the room
The fortnight when nothing happens is rarely a decision being weighed. More often the person you deal with is waiting on somebody senior, or the project has been paused for reasons that have nothing to do with you, or they are embarrassed to say the money went elsewhere.
None of that is visible in a stage name and all of it is answered by asking. That is the real argument for looking every week at what has not moved. Not to chase, but because the answer is almost always something you would not have guessed.
Your stage name and their decision are different events
Accepted, on most boards, means the customer said yes. On their side there may still be a purchase order, a legal review, an insurance check and a signature belonging to someone on holiday. If those regularly take weeks, an honest board has a stage for them. If they rarely bite, it does not.
This is one of the few good reasons to add a stage after a first quarter. Where the stage history shows deals sitting for a long time between yes and money, that gap is a state and it has earned a name.
What a close reason sounds like from their side
The reason a customer gives is the one that ends the conversation politely. Price does that well: it is short, it is not personal, and it invites no follow up.
The reasons underneath are usually different. They had already decided. Someone internal preferred a competitor. You were slow to come back. You answered a question they had not asked and never answered the one they had. You will not get those by asking once at the end, but you can often read them out of the stage history, which shows what you did and when, sitting next to the reason you were given.
The stage you can prove is the stage you are at
A useful discipline for a weekly look at the board: for each deal, what is the last thing the customer did, rather than the last thing you did? If they replied three weeks ago and the card sits at proposal, that deal is not at proposal in any sense that would survive a conversation with them.
It is the two people test from the top of this page, run again with the customer as the second person.
Pipeline questions
Asked about pipelines
How many stages should I have?
Five or six. If you want twelve, you are describing tasks rather than states, and the board will stop being kept current within a fortnight of anybody getting busy. The test is whether two people in your organisation would independently agree which stage a given deal is in.
How do I set the likelihoods?
Guess sensibly at first, then correct them after a quarter from the stage history, which tells you what proportion of deals reaching each stage actually closed. That correction is the single change that turns a forecast from decorative into useful, it takes ten minutes, and almost nobody does it.
Should the first stage have a likelihood above zero?
Yes, but a low one. An enquiry that has just arrived has some chance of closing and pretending otherwise makes the forecast useless at the top of the pipeline. Ten or fifteen per cent is usually about right and the stage history will tell you within a quarter.
When should I run a second pipeline?
When a deal at the same stage in each has a genuinely different chance of closing. Renewals against new business is the clearest example. Do not create a pipeline per salesperson, per region or per product, because those are filters on one pipeline and splitting them makes every cross cutting number harder for no gain.
What do I do with a deal that has gone quiet?
Leave it where it is and let the stage history record how long it has been there. Moving it to a stage called stalled destroys the information you actually want, which is how long it sat at the stage it reached. A saved view of anything unmoved for a fortnight is how you find them.
Should I delete deals I lose?
No. Record the reason and keep them. Deleting lost deals destroys the only data you have about why you lose, and it destroys the denominator for every conversion rate you might want. Twenty close reasons read in sequence say things no report produces.
Can I change stages after I have started?
Yes. Renaming is safe. Removing a stage that deals are sitting in needs more thought, because the history of deals that passed through it is part of your stage history. Adding one in the middle is fine and is the most common change after a first quarter.
My deals do not move in a line. Does this still work?
Mostly. A deal that goes backwards is recorded as going backwards, which is honest and occasionally informative. If your process genuinely has no order at all, a pipeline is the wrong tool and a set of saved views may serve you better.
Do we put the whole contract value on, or the first year?
Whichever you can apply the same way to every deal. Annual value is the more common choice and it makes a renewal pipeline behave sensibly, because a renewal is another year rather than another contract. Total contract value makes a long deal look enormous beside a short one at the same stage, which defeats the point of a board where the stages are meant to be comparable.
Someone keeps moving deals into the last stage before anything is signed. How do I stop that?
Define the stage as an event rather than an expectation, then show them the stage history, which records each move with a time and a name, so the pattern is visible instead of a matter of opinion. It is nearly always somebody being helpful rather than dishonest, because they know it is going to close. The point worth making is that a stage everyone applies differently produces a likelihood no one can correct, which costs them the forecast they were trying to improve.
Our whole cycle is two days. Is a pipeline worth setting up at all?
Probably not as a board. If a deal arrives and closes within a couple of days, nothing sits anywhere long enough for the stage history to say anything, and the board becomes a second place to record work you have already recorded. Keep the outcome and the reason, skip the stages, and revisit it if the cycle lengthens. Being told to skip something is cheaper than finding out in six months that no one kept it current.
Should the owner of a deal be whoever first spoke to the customer?
Not necessarily, but it should be one person and it should be recorded, because the owner is who moves the card and who notices when nothing has moved. Handing a deal over is fine and the stage history will show where it changed hands. What goes wrong is two people each assuming the other has been in touch, which is how a deal sits untouched for a month while both of them believe it is progressing.
Should my second pipeline have the same stages as my first?
Usually not, and if it does you probably did not need a second pipeline. Renewals genuinely have a different shape: healthy, review due, in discussion, renewed. If you find yourself copying the stages across and only the deals differ, that is a filter on one pipeline rather than a pipeline of its own.
What value do I put on a deal where the customer will pay us monthly for as long as they stay?
Pick a period, use it everywhere, and write down which one you picked. A year is the usual answer. The number matters much less than the consistency, because everything the board tells you is a comparison between deals and a comparison only holds if the numbers were made the same way.
Write your stages on paper first
Apply the two people test to each one before you type any of them into the settings.
Three people, a thousand relationships, no card and no time limit.