What Remains After Repeat Measures Have Run Their Course Is the Design of a Reason to Stay

Membership registration, tiers, point rebates, discount distribution, regular contact, campaigns aimed at dormant buyers, simplifying the path to purchase, linking a shop with mail order, contact on paper. Repeat measures names a catalogue of moves of this kind.

The catalogue has a shared property. Every item can be carried out. Nothing difficult is included, and working through them in order eventually exhausts the list.

And when the whole list is exhausted, arriving back at the same place is not unusual. “We did the lot, and the ones who carry on carry on, and the ones who leave leave.”

What this article handles is that place you arrive back at. What is not in the catalogue — the reason people stay, itself — and where that reason is kept.

Three things settle in what follows. Why the form of a catalogue also settles the order of execution. What is actually being maintained when points and tiers work. And in what order to take up the question that remains once the list has run out.

One thing must be answered elsewhere. Which measure your business should take is a question this article holds no answer to. With what is being handled and the length of the purchase interval, the same measure works in opposite directions. Nor does it lengthen the catalogue. Add one item and the place you arrive back at simply moves one step further out.

The damage lands once operating work has settled into the monthly calendar. Reconciling balances, configuring terms, running the distribution. The work that arises carries deadlines, and the design of what you deliver carries none. Let a period run in which only deadline-bearing work gets done and it shows up, six months on, as a single shape: an operating calendar that is full, and an offering unchanged from six months ago.

What makes the shape awkward is that no month contains a wrong judgement. You did what could be done that month, and a record survives of the figure moving that month. The error appears not on the side of any month but on the side of the allocation across months.

Two articles handle the indicator side. The figure for whether the same person buys a second time sits in what the repeat rate observes is the state of the relationship still running, and whether a contract or subscription continues sits in what settles the retention rate is a shared framing, not the substance. Here the view is from the side of the available moves.

What Is Actually Listed in a Catalogue of Repeat Measures

The items fall into five groups, and all five treat repeating as an operation that produces an event.

The first group is points and tiers. Points accrue in proportion to spend, terms improve past a threshold, multipliers rise on particular days. Membership tiers are set, with better terms higher up.

The second group is discounts and benefits: a discount usable next time, a time-limited bonus, a first-purchase-only condition, a bulk price.

The third group is sustained contact. Hold the destination and send at intervals. Hold more than one channel. Retargeted advertising to past buyers.

The fourth group is improving the experience: simplify the path to purchase, respond better to enquiries, deliver faster.

The fifth group is campaigns aimed at dormant buyers. Extract the people with no activity over some period and send them an offer with conditions attached.

The reasons added are the supply of quality matching the price, and increasing fans by generating sympathy. As the cause of not repeating, quality falling short of expectations is named first.

None of the five is wrong as practical advice. Point schemes really do raise frequency, and if contact lapses you are not remembered. If the process is cumbersome, people leave.

Set the five side by side and a shared premise emerges. All five write repetition as something you make happen. Not repeating counts as the default, and a measure is what departs from it.

The front half of that premise matches the facts. Left alone, you are forgotten. What is off is that the catalogue answers only what. In what order, and how far to go before it is enough, cannot be expressed by the form of a list.

The part it does not answer gets filled in by the reader. The information for filling it in is not inside the catalogue, so how it gets filled in is settled by the reader’s own circumstances. From the same catalogue, the order of execution differs by business, and the catalogue itself never accounts for the difference. A catalogue settles the order of execution because the grounds for its sequence are not inside the catalogue. What the order actually gets settled by is handled in the funnel as a whole.

The catalogue also omits the difference in kind among its items. A point scheme and a revision of what you deliver sit in the same column. One is a configuration task; the other is the design of a business. Placed side by side, that difference disappears.

The tidy reply that a catalogue presents options and the reader chooses is accurate. The information for choosing, though, is not in the catalogue. Speed of effect, persistence of effect, operating effort, interference with other measures. Without those written down, the criterion for choosing has to be brought in by the reader. What gets brought in tends to be ease of starting.

The Form of a Catalogue Settles the Order of Execution

A catalogue with no order shown tends to get executed, in practice, in order of least effort.

A point scheme runs once configured. Distributing a discount means deciding terms and sending. Sustained contact takes the effort of writing, and writing ends it.

Improving the experience, by contrast, has no visible end. How much to improve and to what standard is something you have to settle yourself. Revising quality does not even have a clear place to begin.

The order therefore runs from what ends at configuration to what has no visible end.

That order has a consequence on the effect side. The earlier something is executed, the faster its effect and the shorter it lasts. Put a point scheme in and the figure moves the following month; what moved disappears if you stop the scheme. The later something is postponed, the slower its effect and the longer it lasts. Revising the offering itself works over months, and what it works survives the ending of any measure.

In other words, execute in order of least effort and you have executed in order of shortest effect.

The judgement that starting with what works fastest is reasonable is, in itself, correct. What works fastest, though, has a property that distorts later judgement. When the figure moves, a confirmation that the direction is right is left behind. Accumulated confirmations remove the reason to move to the slower side.

Fast measures also generate operating work: managing points, designing discounts, running the distribution. The work continues monthly, so it takes time away from the slower side.

The time taken away never appears as consumption in the records. Operating work survives as done, and design work not carried out survives nowhere. Look only at what the records hold and the month was full.

The catalogue itself is not defective. Without knowing the available moves, you cannot choose. What can be said is that the form of a catalogue drops the information about order. The dropped information gets supplied, in practice, by ease of starting.

The order bites hard where one person runs the business, because the designer and the executor are the same. In a company, execution load and design priority sit with different people. Alone, what can be done today becomes what is done today.

And the accumulation of what could be done today comes out, six months on, as one recognisable state. No day contained a wrong judgement. What was wrong was the criterion that set the order.

“It Works” and “It Keeps Working” Are Different Properties

Descriptions of measures state that they work, and say nothing about whether the effect survives the ending of the measure.

Two types. The first works only while the measure is running. Points, discounts, tiers, retargeting. Stop and you return to where you were. The second survives the ending of the measure. The other person became able to do something, gained a basis for judging, understood what you handle. None of that disappears when you stop working at it.

In a catalogue the two sit in the same column. Sitting together, they are chosen as though they were the same kind of thing.

The first type has to be strengthened as it continues. The same rate of return is treated as the default condition the second time round. Raise it or it stops working; raise it and you cannot go back. That mechanism is the same one handled in product launch, where a reason supplied from outside inflates. In repeat measures it appears as monthly operating work.

Nothing here argues against ever using the first type. In the early period, producing movement is necessary. What can be said is that a design built entirely of the first type becomes a design that cannot be stopped. Stop and the figure falls, so stopping tends to look like a loss.

There is a way to tell them apart. Ask what will be left in the other person’s hands when you stop the measure. Stop the points and only the memory of a balance is left. Hand over a basis for judging and the basis stays with them after you stop.

The two types also require different kinds of work from you. The first can be routinised, and being routinised it is easy to keep up. The second is different every time and cannot be routinised. What cannot be routinised is the first thing dropped in a busy period. Dropping it shows nothing in that month’s figures. It shows some months later.

The second type is not entirely unmeasurable. Whether what you handed over is being used shows in the content of the next exchange. The same question recurring means it did not land; a question one step further on means it did. That observation cannot be aggregated as a count, though, so it never appears in a list of indicators.

The lag biases the evaluation of the two types further. The first type produces its result in the month you act. The second produces nothing in the month you stop. Judge by the same month’s figures and only one of them appears to be working.

Give Progress Artificially, and Why People Go to the End

When points and tiers actually work, what is operating is not a reason but a sense of progress.

The marketing researchers Joseph Nunes and Xavier Drèze reported, in 2006, an experiment using car-wash loyalty cards (Nunes & Drèze, 2006, Journal of Consumer Research, 32(4), 504–512). One card required eight stamps; the other required ten but arrived with two already stamped. The number of purchases needed was eight either way. The completion rate was higher for the card that arrived with two stamps already on it. People who completed also shortened the intervals between purchases.

The authors’ account sits on the side of the sense of progress. Present a task as two of ten done rather than eight to go and people move toward finishing something already begun. Nothing about the content changed. What changed was where the counting starts.

The conditions of the experiment matter here. The subjects were customers of particular outlets, and what was handled was low-value repeat purchasing. Nothing guarantees an effect of the same size in your business. What transfers is not the magnitude but the location. What is moving here is not the reason for choosing that outlet; it is the pull to finish a task already part-way through.

Findings pointing the same way exist in research on behavioural learning. Behaviour rewarded irregularly becomes harder to extinguish once rewards stop, as has been shown repeatedly (Skinner, 1938; Ferster & Skinner, 1957). Draws, variable multipliers, benefits that arrive without warning. These do make behaviour harder to extinguish than a certain rebate does.

Working strongly, they are useful as instruments. What the two maintain, though, is behaviour and not a reason. Not wanting to lose a tier and holding a reason to continue are separate states.

And the distinction never appears in the figures. A purchase moved by progress and a purchase moved by a reason are recorded as the same single transaction. The breakdown becomes visible only when the progress mechanism is stopped. Stop it and exactly the progress-driven portion falls away.

Progress-based design also has a limit. Once the task is complete, progress returns to zero. From that point, the pull disappears unless a new task is presented. As long as you keep presenting one, this is being operated as the first type.

Measures Run Their Course — What Remains When the List Is Exhausted

The items are finite, so executing them in order necessarily ends. When it ends, what is in your hands?

What remains is the monthly operating work and the set of conditions the measures created. Point balances, tier promises, the slot for regular sending. Every one of them is something you must keep maintaining.

And something is not there. Your own understanding of why these people are continuing.

The reason it is not there sits in the nature of measures. A measure is an instrument for producing a result without asking about reasons. Somebody who bought because of points and somebody who bought because they needed the content are recorded as the same purchase.

After the list is exhausted, the next move tends to be hard to find. The catalogue is spent, and searching for new measures returns only variants of what was already done.

Two interpretations then split off. Not enough measures, or our goods just do not get repeated. Both close the question. The first continues further in the same direction; the second gives up on it as structural. The question in between — why did these people buy a second time in the first place — comes out of neither.

The reading that exhausting the list proves the limits of measures does not hold. What was exhausted is the catalogue of measures, not the whole of what can be done. That no move exists outside the catalogue is not something the catalogue can tell you.

Exhaustion also leaves fatigue. Having felt you did everything, this is a point of low energy for starting anything new. And the move most often taken at this point is raising the intensity of measures already running, because that can be done without adding to the catalogue.

Running through the list is not in itself a bad outcome. That you tried the available moves is information in its own right. The problem is that the running-through leaves no record. Which measure went in when, how far it moved things, what is left now. Without that record, you cannot tell whether you have run through the list or are still part-way.

The record does not survive because no motive to keep it arises at the time. In the month a measure goes in, putting it in is itself visible as an achievement. What that measure will leave behind years later is not a question in that month. By the time it is a question, the material for comparison has gone.

And without records, some years on you will walk the same catalogue again. Having forgotten what was done last time, it looks like a new measure.

Combinations That Cancel Each Other Out

A catalogue lists items as independent, and some of them weaken each other when used at once.

Three examples. First, discounts and the explanation of value. Explain the value of what you offer and then issue a discount, and the size of the discount is received as the measure of value. The amount taken off stays in memory rather than the value explained.

Second, tiers and equal treatment. Create tiers and anybody not at the top is given the position of not yet above. Treat everybody identically after assigning that position and the meaning of the tier disappears. A tier whose meaning has gone leaves only the expectation.

Third, benefits for dormant buyers and standard prices for continuing ones. Offer good terms to somebody drifting away and the people who stayed are worse off. That structure makes staying look like a loss.

All three are sound measures individually. What cancels is their simultaneous presence. And the form of a catalogue does not express the interaction. Items are evaluated independently and executed independently.

Differentiating by recipient does work. What has been differentiated reaches the other party, though. That the same product is offered on different terms is easily discovered in the present environment. What becomes a problem on discovery is not the difference in terms but whether the reason for the difference can be stated. Better terms the longer you stay can be stated. Better terms the closer you are to leaving cannot be.

None of this argues for abandoning campaigns aimed at dormant buyers. Some people do come back. What can be said is that somebody brought back by terms leaves by terms. The record of the return survives; the reason for it is not where it was the first time.

Cancellation never appears in the figures. You can measure the response to a discount; you cannot measure the explanation of value weakened by that discount. The weakened portion shows up at the next proposal. The cost of cancellation therefore tends to fall in a later period. At the point it falls, the cause does not look like a measure taken months earlier. Not looking like it, the remedy taken then is yet another measure.

The three combinations were none of them built deliberately. Measures added at different times for different reasons end up running simultaneously without anybody noticing. Nobody recorded the point at which they began running together, so the point at which interference began cannot be identified either.

This is also a question of the period over which measures are evaluated. Evaluate monthly and cancellation is structurally undetectable. It is detectable only by following the same people over a year.

What the Term “Dormant Customer” Assumes

Of everything in the catalogue, dormant customer is one place where your own convenience shows plainly.

The term names somebody with no trade with you over some period. The metaphor of sleep is inside it. Being asleep, they will return if woken — that implication is inside it too.

From their side, the state differs. Some simply have no need arising. Some have already solved it elsewhere. Some deliberately took their distance. All three appear identically in your records as no activity. Appearing identically, they receive the same campaign.

What happens on the receiving side? For somebody with no need arising, it lands as an irrelevant message. For somebody who already solved it, as a late one. For somebody who took their distance, as one that ignored the distance.

Of the three, the campaign is most likely to work on the first. For the other two it tends to thin the relationship further.

And the second and third tend not to respond. Not responding, they survive in the records as it did not work. That record then becomes the basis for raising the intensity next time.

This sits in the same place as the structure handled in how to read the drop-off rate, where reasons for leaving go unrecorded. As long as the classification is built on presence or absence of activity, differences of reason do not enter the classification.

Classification need not be abandoned. Handling everybody individually is impossible. What can be said is that while the axis of classification is activity as seen from here, the content of the campaign comes out of your convenience as well. The method for moving the axis onto the other person’s state is handled in email segmentation.

Sometimes their state genuinely cannot be seen from here, and activity has to stand in for it. Even then, whether you record that you substituted changes later judgements. Call them the layer with no transaction record for six months rather than the dormant layer, and what survives is a description of what could be observed, not an inference about a state.

Changing the name changes what gets sent. Think of them as asleep and you send something designed to wake them. Think of them as merely unrecorded and you send something that asks what is happening now. The second gets a lower reply rate, because asking demands effort from the other person. The number returned falls, and the information in what returns rises.

The Place You Return To Sits Outside the Measures

One question remains after the list is exhausted, and its answer sits in one of two places. Why are these people still here?

The first place is the side of the transaction. Points have accumulated, a tier would be lost, a discount is working, cancelling is troublesome. Where the reason sits here, it is maintained by updating the terms.

The second place is the side of the relationship. What was received is still working, what you handle is understood, you sit in a position to be recalled when the next need arrives. Where the reason sits here, it needs no updating.

The two places also change how departures read. Build the reason on the transaction side and a departure reads only as the terms stopped being attractive. Build it on the relationship side and departures split into what was handed over did not last and the need ended. The second is a departure with nothing for you to fix. Until you can read them apart, you go on raising terms for people whose need has ended.

The catalogue works mostly on the first place. Items working on the second are barely in it at all. The reason is simple: the second place cannot be executed as a measure. Make what is received last is not a measure but the design of the offering itself.

In other words, the place you return to sits outside the measures. However far you extend the catalogue, you do not reach it.

The two places also differ in how costs fall. The first requires funding every period to maintain: point provisions, the discount portion, the concessions attached to tiers. It keeps flowing as long as you continue and, since it stops working unless raised, the amount can only move upward. The second takes effort to build and almost nothing to maintain. A basis left inside somebody stays there while you do nothing.

The difference bites harder the smaller the business. A cost that flows every period hits its ceiling sooner where the funding is thinner. At the ceiling you can no longer raise terms, and from that moment the reason in the first place begins to weaken.

Working on the second place is not a matter of making something good. Make something good and, in a form that leaves nothing on the receiving side, the relationship still ends. Things in a form that is consumed and finished produce higher satisfaction in the moment the better they are, and leave less behind. A form that lasts is a form usable on their side: a basis for judging, a way of reading a situation, a structure to refer back to when the same scene recurs. Those are recalled each time they are used.

What Sits Outside the Measures Splits Into Three

Building a reason in the second place splits into three: making what is received last, making clear what you handle, and keeping the exit open.

The first means placing, at the end of a delivery, a form in which the other person writes down in their own words what they got from it. What matters here is that you do not summarise. Text you compile survives in your words. Text they wrote in the language of their own situation connects to their memory.

The second holds by naming what you do not handle. List only what you do handle and the boundary looks wide. Put in one this is not something I handle and the boundary acquires an outline. The clearer the boundary, the more reliably your name comes up when a need inside it arises.

The third is satisfied by a notice. Announce the renewal date in advance and write down how to stop. That alone makes the state of continuing a result of choice. Only a relationship continuing under easy conditions of exit is evidence that the reason sits on the side of the relationship.

None of the three has the shape of something executed and finished. A point scheme ends at configuration; all three arise at every delivery. They are absent from catalogues because they do not complete as items. What is not listed goes unnoticed when it is skipped.

The three do not raise short-term figures. The first adds effort, the second narrows the audience, the third lowers the continuation number. The ground for doing them is the single fact that nothing else is left once the list has run out.

The worry that handing over a basis makes you unnecessary is right at one rung. Somebody holding the basis can make that rung’s judgements themselves. And the question at the next rung arrives, because somebody who can judge has widened the range of what they judge. How rungs get cut is handled in raising the amount per customer. Where the rungs are cut by volume, this move upward does not happen.

Of the three, the one unlikely to be executed is the third. Only the operation of opening the exit lowers the figure in the month you carry it out. The first and second do not lower it and do not raise it either. What does not lower things is rarely postponed; what does gets postponed indefinitely.

Not everybody moves to the next rung. Some are finished once they hold one basis, and that state is not a deficiency. What can be said is that the design has to treat the departure of those who do not move on as something other than a loss. Treated as a loss, the design drifts toward not handing over the basis.

Working Out for Yourself What Your Measures Are Holding Up

The split is readable not from a count of measures executed but from what happens during periods when you are doing nothing. Three steps, in order.

First, write out the measures currently running. Not to count how many, but to see which of them are running at the same time. Written out, there may be more than expected, and they went in at scattered times. Check here whether any of the combinations that cancel each other when combined are sitting together.

Second, stop one of them, for a subset of people only. They need not all be stopped at once. Anything that stops without moving the figure was already not working as a reason. Anything that moves the figure is holding that much of the reason on the transaction side. One at a time, because stopping several at once loses which one was working.

Third, ask how the other person describes you. Their description in a referral situation settles it. Because it is cheap, because there are benefits are descriptions on the transaction side. Somebody who handles this kind of thing is a description on the relationship side. The second means your boundary came across.

None of the three counts measures. Since the reason for staying sits outside the measures, what has to be checked is outside them too.

To the question of whether anything faster exists, there is an answer. Many fast indicators are measuring the first place. Open rates, response rates, that month’s repeat purchases. All arrive within the month, and what arrives is a response to your action. The second place holds independently of your action, so it does not appear immediately after it. It appears during periods when you are doing nothing. Verifying it requires a period of doing nothing.

How many measures to run, what rate of return to set, how many months to draw the dormancy line at. Goods where one percentage point of rebate decides the purchase and goods where nobody looks at the rebate are being discussed in the same words. So no figure can be given for any of them.

Run the measures through and the ones who carry on carry on, and the ones who leave leave. Arriving back at this view, what sits there is not the fact that measures failed. It is the fact that everything that worked was of the type that disappears when the measure stops.

The way out of the catalogue is, in the end, to carry out the three separating steps in order. Write them out. Stop one. Ask how you are described. None of the three adds a single new measure. What you learn without adding is what is holding the figure up now, and only once that is known can what to build next be decided.