How the App Works

A structured sequence of steps that starts with something your dog can already do easily and ends at the finished behavior.

Each step raises the difficulty a little, using one dimension at a time: a little more distance, a little more duration, a little more distraction, one less bit of help. The first step of a good plan looks almost insultingly easy. That is deliberate. Starting below your dog's current ability costs you two minutes and buys you a dog who is already succeeding when the real work starts.

You do not choose which step to train. You pick the plan, and the app puts your dog on the right step and moves them as the data comes in.

Often confused with

A course, or a curriculum. A plan is not a schedule to get through. Some dogs walk a 16-step plan in two sessions and some take six weeks, and the plan does not care.

The screen the app opens on: a trail of plans for your dog, in a sensible order, with one card at the top saying what to train next.

Every dog starts on Everyday Manners, the starter course: Touch, Sit, Down, Come, Leave It, Casual Heel, Down Stay and Go to Bed. Each stop shows how many of its goals your dog has earned, and the top of the trail counts the skills done so far. Some stops have a side trail, a helper plan off the main path for the dogs who need it: Down has Tunnel Down, for small or hesitant dogs. The trail ends at the castle, where your goals are kept and games are played.

Next Up is the card at the top. It names the plan and the step the app will pick up at, and Train opens the Player right there. There is one card on purpose. The Journey is a way into training, not a catalogue: one clear next step gets trained, and a page of options mostly gets scrolled.

Making it your dog's own. The trail is a starting suggestion, not a syllabus. Ask Coach and it can change any of it: add a plan where your dog needs one, hang a helper off a stop, take out what your dog already knows, rename a stop, or pin something else as Next Up. A dog who already has the basics can have a trail of two real projects. Coach tells you why it made each change, and that reason is shown on the trail. Because the changes go through Coach, customizing a Journey is something Coach does, and Coach comes with a subscription; everything on the trail trains without one.

Finding everything else. The Home button, bottom right of the Journey, opens the home page: Coach, your recent sessions, the games, and the full plan library by category. The back arrow at the top left of the home page brings you back to the Journey. The Push Drop Stick logo at the top left of the Journey opens the About page: privacy, feedback and your account.

Often confused with

A required order. The Journey only decides what Next Up suggests. Any plan you have can still be opened from the home page and trained whenever you like.

One rung of the ladder. A single small version of the behavior, with its own setup and its own definition of success.

A step is the unit the app makes decisions about. Your dog does a handful of repetitions at one step, and the result decides whether they move up, repeat it, or drop back.

Good steps are smaller than owners expect.

Often confused with

A session. A step is a difficulty level. A session is a period of time, and usually contains several steps.

The observable description of what counts as success on this step, decided before you run the rep.

This is the most important word in the app, because pass and fail are meaningless without it. "Did my dog sit?" is not criteria. "Within three seconds of the verbal cue, without a hand signal, and their bottom stays down until I mark it" is criteria.

The test for whether you have real criteria: could a stranger who does not know your dog score this rep and get the same answer you did? If not, you are judging on feel, which drifts across a session, gets more generous as you get tired, and hands the app data that describes your mood rather than your dog.

Every step in every plan comes with its criteria written out, in the pass and fail lines. Read them before the first rep of a step, not after your dog has done something ambiguous.

One more rule, and it is the one that makes small steps possible: change one criterion at a time. If you add duration, do not also add distance. If your dog is learning the word, do not also move to the garden. When a step contains two changes at once, a fail tells you nothing, because you cannot tell which change caused it.

Often confused with

Standards, or expectations. Criteria are not how well you hope your dog will do. They are a line drawn in advance so that today's judging matches yesterday's.

Short for repetition, also called a trial. One attempt at the current step.

Reps are usually run back to back so the data comes from a consistent moment: same room, same mood, same hunger level. A rep starts with you setting your dog up, includes whatever you prompt, and ends when you score it.

Some protocols do not fit that shape. Classical conditioning reps are often scattered through the day, one at a time, whenever the trigger happens to appear.

Often confused with

A try. A rep is a measurement. If you were not ready to score it, it was not a rep, and it is fine to reset and start it properly.

A group of reps at one step, usually five, after which the app makes a decision.

Five is enough to tell a real pattern from a lucky rep and small enough that a struggling dog is not stuck failing for long. Timed steps work slightly differently: the set runs for a minute and you fit in as many reps as the pace allows.

Finishing the set matters. Abandoning one halfway leaves the app without a decision, and the temptation to bail out is strongest exactly when the data would have been most useful.

Often confused with

A session. Most sessions hold several sets.

The two buttons. A pass means the rep met the step's criteria. A fail means it did not. That is the entire meaning of each.

A fail is not a judgement of your dog, and it is not something that happens to your dog. A fail is a measurement, never an instruction. Nothing unpleasant follows it: no scolding, no repeating the rep until it works, no disappointed sigh, no withheld affection. The only consequence is that the treat does not arrive: you set up the next rep, and the app gets a data point. See Positive Punishment.

Classical steps are the exception: the food comes on every rep, pass or fail. On those steps the food is not payment for anything your dog did, it is what the trigger predicts, so it has to follow the trigger every time. The fail button only records that the trigger was too much for your dog today. See Classical Algorithm.

One nuance, because some plans do ask you to mark a miss out loud. A no reward marker, a light "oops", tells your dog that this particular attempt will not pay and invites them to try again. It is still a consequence, if a mild one: a word that reliably means "no treat this time" picks up a little of the disappointment of missing out, however cheerfully it is said. It earns its place on clear-cut criteria like a stay, where dogs often reset themselves the moment they hear it. Said sharply, or with a sigh behind it, it adds something your dog wants to avoid on top, which is positive punishment wearing a friendlier word. Keep it light, use it where the criterion is obvious, and if your dog looks worried when they hear it, stop using it. They are optional.

Tapping the fail button is not a no reward marker. Your dog should not be able to tell which button you pressed.

Your honesty here is the whole product. It runs wrong in both directions:

• Generous judging feels kind and is not. Marking near-misses as passes makes the app push your dog to a harder step they were not ready for, and they fail there instead, having lost the easier step that would have got them ready.
• Harsh judging leaves your dog stuck repeating work they had already mastered, which is boring, and boredom looks a lot like disobedience.

Some practical rules that make judging easier:

• If you genuinely cannot decide, it is a fail. And note the ambiguity, because it usually means the criteria were not clear before you started.
• Judge against the step, not against your dog's best ever rep. Also not against how much you want to get to the next step tonight.
• A dip right after a push is expected. The app just raised the difficulty and is finding the edge. A rep or two of fails is normal and usually short-lived; a whole run of them means the step was too big a jump, and the app will drop back.
• Fails late on a step your dog already owned mean something different: tiredness, or a stacked day. See Trigger Stacking.
• On comfort steps the criterion is comfort, not compliance. A dog who completed the exercise while lip licking and looking away has failed that rep. See Stress Signals.

Two step types remove the button on purpose. Capturing has no fail button at all, and Errorless disables it until the timer ends.

Often confused with

Right and wrong, or good dog and bad dog. The dog is never wrong. Either the criteria were met or the step was too hard, and only one of those is your dog's business.

One sitting of training, usually made up of several sets.

Keep them short. Five to fifteen minutes suits most dogs, and shorter is better than longer when something is new. The instinct to keep going after a great set is the most common way a good session turns into a bad one.

Stop while your dog is still fresh and still keen. That is not a soft suggestion, it is where the learning is: a dog who ends wanting more comes back to the next session ready, and a dog who ended flat comes back slower. If your dog is tiring, the answer is fewer sets today, not easier reps.

How often is up to your life. Daily practice is lovely and entirely optional. Train as often as your schedule genuinely allows, and know that a day or two off is not lost ground: dogs consolidate what they have learned in the gaps, and a rested dog often comes back better than they left. Two short sessions a week that actually happen beat a daily plan you abandon by Thursday.

Often confused with

Practice time. Sessions are not the only training your dog gets. The rest of the day is teaching them things too.

Move forward to the next, harder step. Your dog has this one.

The app pushes when your dog passed enough of the set, which on most steps means four or five out of five. You do not decide this and you do not need to. Judge the reps honestly and the push arrives when it has been earned.

A push often produces a rougher next set, and that is expected. The app is looking for the edge of what your dog can do, and sometimes it finds it by going slightly too far and then adjusting. The aim is still to keep fails rare.

A push is drawn in green.

Push, drop and stick are not this app's invention. The criterion system is Jean Donaldson's, taught at the Academy for Dog Trainers; the app's contribution is doing the bookkeeping so you can keep your eyes on your dog.

Move back to the previous, easier step. Too hard, right now.

A drop is the app protecting your dog, not a mark against either of you. Nothing has been lost: your dog gets a set they can win, the confidence comes back, and the harder step is still there.

The phrase to hold on to is right now. A step that was too hard on a hot Thursday evening after a stacked day is frequently easy on Saturday morning. Dropping is how the app absorbs that without asking you to diagnose it.

A drop is drawn in dark blue. Not red, because it is not an error.

Often confused with

Failing the plan, or going backwards. Judging honestly and letting the app drop will get you through a plan sooner than judging generously to avoid it. A generous pass does not help your dog, it moves the difficulty up onto a dog who was not ready for it, and you pay for it two steps later.

Repeat the same step. Right on the edge.

Roughly three out of five means your dog is genuinely learning this one and is not ready to be asked for more. Another set at the same difficulty is exactly the right amount of help.

Two sticks in a row is perfectly normal, and often exactly what a hard step needs. Three or four in a row is worth noticing: it usually means something outside the step is interfering, such as the environment, the treat value, or your dog's day.

A stick is drawn in light blue.

Insert an in-between step, because the gap between two steps is too big for this dog.

The app splits when a step ends in a drop for the second time, with no push or stick in between. From the outside that usually looks like a bounce: your dog drops back, comes again, and drops again. The pattern is not confusion or stubbornness, it is a measurement, and what it measures is the distance between two rungs. Splitting is the app building a rung in the middle.

Some steps have several splits available. Some have none, in which case the app will drop again instead.

Of the four moves, this is the one to be most pleased about. It means the plan is being fitted to your dog rather than the other way round.

A split is drawn in yellow, which is why split steps are tinted yellow wherever a plan lists them.

The in-between step itself. An easier version of the step in front of it, kept hidden until your dog needs it.

Split steps do not appear when you browse a plan in the phone app, which surprises people who go looking for them. They are held back deliberately: most dogs never need them, and a plan that showed every possible rescue rung would read as intimidating rather than achievable.

So if your dog is stuck and you are wondering whether the plan needs an extra step written for it, the answer is usually that the extra step already exists and the app will produce it after a repeat drop.

If you do want to see them, the website shows every step of a plan, splits included, along with more of the machinery than the app puts in front of you: the training type, the treat values, the rep count, and the exact push, drop and bail thresholds. The app shows you what you need while you are training. The website is where to go when you want to read how a plan is built.

Every step is judged by one of five sets of rules, shown as the step's training type. The type is chosen by whoever wrote the plan, based on what the step is teaching.

• Prompting — the standard. You cue or prompt a behavior and score whether you got it.
• Shaping — timed. Tracks two criteria at once, most often while a verbal cue is being added before a known hand signal.
• Capturing — you reward a behavior your dog offers on their own. No fail button.
• Classical — building a good feeling about something. Drops after a single fail.
• Errorless — timed, and set up so failure is nearly impossible. The fail button is disabled until the timer ends.

The last two are the back-off-fast rules, and they are what the app uses for anything to do with fear, comfort, grooming or handling. If a step is judged by one of those, that is a signal about what the step is for.

One clarification, since grouping them can mislead. Errorless steps are operant learning, exactly like the first three: your dog does something and you pay it. What makes them different is the setup, arranged so the right answer is very nearly the only one available. Classical steps are the only ones built on classical conditioning instead: the trigger predicts food whatever your dog does, so they are about how your dog feels rather than what they do. Both kinds of learning still run on every rep of every step, in the direction you want or the other one, which is why a Classical step backs off at the first fail.

Rules for steps where you are rewarding something your dog offers on their own, with no prompt and no cue.

Because you are only ever paying a behavior your dog chose to produce, there is nothing to fail, and no fail button. Your dog cannot get this wrong. If the behavior is not happening, that is about the setup, not about your dog. See Capturing.

Which means a capturing step has exactly one outcome. Once the reps are in, it pushes. It cannot drop, and it cannot stick.

If all reps are completed, then Push

Rules for emotional work, where what matters is how your dog feels rather than what they did.

These steps Drop at the first sign your dog is uncomfortable. A single fail is enough, and that is the point.

A fail here never means withholding the food. Your dog gets it on every rep, whether they coped or not, because the food is there to change what the trigger means, not to pay for calm. Pass and fail only tell the app whether the trigger was set low enough.

Do not worry about dropping repeatedly in the early stages. Trust is much harder to rebuild than to keep, and moving faster than your dog is ready for costs you far more time later than it saves now. The genuinely strange truth of this work is that you will get to the end sooner by going at what feels like a ridiculous crawl.

If you fail a single rep, then Drop
Unless this is the second drop in a row on this step, then Split
If all reps are completed and passed, then Push

Timed rules for steps where your dog is not meant to fail at all.

The setup is arranged so that the right answer is the easy answer, and the fail button stays disabled until the timer runs out. If you do use it, one fail drops the step.

The pass button does not wait. If your dog gives you the behavior after four seconds of a sixty second step, mark it and move on. The timer is there to stop you calling something a failure too early, not to make you sit out the clock.

These are the steps where the app is most explicitly on your dog's side: they get to be right by default. See Start Button.

A timer is used for these steps.
If you fail a single rep, then Drop
Unless this is the second drop in a row on this step, then Split
If all reps are completed and passed, then Push

On a timed step, the interval is how the length of each rep gets decided.

Most steps are plain: every rep asks for the same thing. A fixed interval works that way with time, asking for the target duration every rep. A variable interval mixes shorter reps in among the target ones, so your dog might hold a down for seventeen seconds, then one second, then eighteen, then twenty, then fourteen, and then the full thirty.

On a variable step the Player shows you that list where the demo video usually sits, so you can see what you are working on and what is coming next. Green is a target duration. Blue is a shorter one.

Two things surprise people. You will do more than five reps, because only the green targets count towards the tally. And you pay your dog for every rep, including the one second ones. A rep your dog completed is a rep your dog completed.

The reason for all of it is rate of reinforcement. A step that only ever asks for thirty seconds pays your dog twice a minute, which is thin going for hard work. Salting shorter reps in among the long ones keeps the payments coming, and it means there is no single elapsed time your dog can learn to treat as the end.

Often confused with

The variable interval schedule in behavior analysis. They share the idea and the name. Strictly, a variable interval schedule reinforces a discrete response after a varying amount of elapsed time, and the animal is free to do as it likes while the clock runs. A held down-stay is continuous, so a behavior analyst would more precisely call this variable duration. Trainers have used "variable interval" for both for a very long time, and so do we.

A milestone step that records something meaningful your dog can now do: responds to the hand signal, responds to the word alone, holds it at a distance, holds it for a duration, does it around a distraction, feels good about something that used to worry them, does it in its own context without being asked each time, or offers it entirely unasked.

Goals group hundreds of small repetitions into the handful of achievements you would actually tell someone about. Each type has its own icon: a hand for a hand signal, a talking head for a verbal cue, a double-headed arrow for distance, a timer for duration, a squirrel for distraction, a party popper for comfort, a light bulb for automatic, and dice for offered.

The last two are worth a word, because they are the ones owners are proudest of without having a name for. An automatic goal is a behavior that runs in its context with no word or hand signal each time, because the situation itself has become the cue: the sit that happens when you stop walking, or the rest of a chain carrying on after one cue. An offered goal is one your dog volunteers unasked: a check-in on a walk, settling on their mat on their own. Both mean the behavior has stopped depending on you asking for it, which is usually the point where the training starts to feel finished.

You earn a goal by pushing off that step, and it turns green, the push colour. If you are later asked for it again and drop or stick on it, in a plan or a game, it turns blue, which means it needs a little review rather than that you have lost it. Each goal always shows the most recent answer your dog gave, and the flags on the castle keep the running count: green for goals earned, blue for goals that want review.

Often confused with

Finishing a plan. Goals sit inside plans, and most plans hold several.

Practising a behavior your dog already knows in harder places, until it works in real life rather than only in your kitchen.

Dogs are far more context-specific than people expect. A sit that is flawless by the fridge can genuinely not exist at the park, and that is not your dog being difficult, it is how learning works. The behavior was learned along with the room.

The rule that makes proofing pleasant: a new environment is itself a difficulty, so lower everything else. Shorter durations, less distance, better treats, easier criteria. Let your dog succeed in the new place before you ask for the standard you had at home.

The app gives you four ways to proof, none of which involve a correction:

• Proof steps — replay a finished plan's hardest end in a new place. See Proof Step.
• Proofing games — cues from several plans, shuffled, so your dog listens instead of guessing the pattern.
• Blue goals — the blue flag on the castle counts the skills that want review, and the goals page lists exactly which ones.
• Other people — hand the phone to family or friends and let them run proof steps or a game. A cue that only works when you hold the treat pouch is not finished, and this is the cheapest way to find that out.

Often confused with

Testing. Proofing is teaching, and it should look like your dog getting it right a lot.

Once a plan is finished, proof steps let you replay just the harder end of it, in a new place or around distractions, without walking the whole ladder again.

Unlike split steps, proof steps are visible in the plan from the start, and it is normal to see one sitting before a plain step in the list. They interleave where they belong.

A short, shuffled drill built from the goals your dog has already earned, pulled from different plans and called in random order.

This is the answer to a problem every trained dog has: within a plan, your dog can succeed by guessing the pattern. Games remove the pattern. Sit, then wait, then touch, then down, in an order nobody can predict, means your dog has to actually respond to the cue you gave. See Stimulus Control.

Games start from the castle, at the end of your Journey, or from the flags on the home page. Either way you can take a random game or tap the goals you want in one. Goals your dog nails stay green. Goals they miss turn blue for review, which is the game doing its job: it just found the cue that was leaning on context rather than standing on its own.

Games are also the app's whole answer to proofing under pressure. There is no correction anywhere in them. You raise the difficulty by shuffling the order, not by adding a consequence.

Where your dog's goals are kept, and where proofing games begin. It sits at the end of the Journey, and tapping the flags on the home page takes you to the same place: the Games page, with the castle at the top.

Two flags fly from it. The green flag counts the goals your dog has earned. The blue flag counts the goals that want review. The number is written on the flag, and a flag with nothing to count is lowered rather than showing a zero, so a castle with only the green flag up has nothing waiting. Scroll down the Games page for the goals themselves, grouped by plan, each with its badge. See Goal Step.

Two ways to start a game. Random game shuffles everything your dog has earned. Or build your own: tap the goals you want in it, or Add all to take a whole plan's goals at once. Building your own is how you aim a game at something: only the verbal cues before a class, the plan your dog is newest at, or a mix for whoever is taking a turn with the dog this week. Coach can build one for you too, on a theme you name. Either way the goals are called in an order nobody can predict. A goal your dog nails in a game stays green, and one they miss moves over to the blue flag until it has been reviewed.

Why a castle. The classic adventure game ends at a castle, and so does the trail. By the time you get here your dog has learned real skills, and this is where you proof them by playing training games. A castle felt right for that: a little grand, a little fairytale, and built to last, which is exactly what you want from a cue. It is a long way from your first session, and worth celebrating.

Ours is modelled on Balmoral, the Scottish castle where Queen Victoria fell in love with the working collies of the Highlands. Collies like Sharp and Noble were rarely far from her side, and her affection helped turn the collie into one of the best-loved dogs in the world.

Often confused with

A score. The blue flag is not a count of failures. It is the app's shortlist of what to practise next, and the best reason to start a game.