Choosing a Web Host Is Settled by the Cost of Leaving and the Place You Stack

The choosing of a web host is explained as a comparison of performance and price. Speed of display, capacity, monthly charge, functions included. The figures are published, so lay them side by side and differences emerge.

The more carefully the comparison table is read, the stronger the sense that this is a heavy decision. The items are many, the differences fine, and the ranking shifts from one operator to the next.

That sense is mistaken. Nearly every item on the comparison table can be changed later. If display is slow you can move; if capacity runs short you can add. If the monthly charge rises, you can go elsewhere. Light items are standing there wearing the face of heavy ones.

The heavy item is not on the table. What it costs, after choosing, to leave. This is not published as a figure, so it does not become an object of comparison. And it is this one that actually governs the business.

Only the side absent from the comparison table is handled here. What the cost of leaving is made of. How address, machine and accumulation differ in the nature of ownership. What is actually done to keep things interchangeable. Those three are settled here.

No operator is named anywhere here. Which is good depends on the scale of the content handled, the other contracts already held, and the direction being extended. Conditions are revised on the operator’s side, besides. From the moment a name is written, the account begins to age. What can be handed over here is not a name but the axis by which you judge names yourself.

The timing of the bill has a character of its own. The cost of leaving is billed in a lump, on the day you try to leave. And the attempt to leave usually comes when the business has grown. Space ran short, conditions stopped matching, another form became worth trying. So this cost arrives at the busiest period, at its largest figure.

Through the life of the contract this cost reads as zero. Nothing symptomatic surfaces in daily operation. Display is fine, articles post, costs run as planned. With no symptom, it never becomes a candidate for attention. And on the day a symptom appears, the cost of attending to it is at its maximum.

Why it is not published is simple as well. The cost of leaving arises not from the specification of the equipment but from the design of the contract and the control panel. Specifications become figures; designs do not. What does not become a figure does not become a column, and what is not a column falls outside comparison. What follows is arranged as a route for raising that missing column yourself.

Performance and Price Compare Accurately, but Over Items That Can Be Changed

The criteria presented for comparison are correct as a practical description, and weighted toward items that can be altered afterwards.

Web hosting is machinery for publishing your own website by renting part of equipment the operator has prepared. There is no need to prepare and manage equipment yourself; it runs on a monthly charge alone.

The considerations presented run roughly to four. Price, performance, security, functions. A rule of thumb for the monthly charge is appended, of somewhere between five hundred and a thousand yen.

On performance, speed of display, stability when many people visit at once, and capacity are raised. On security, support for encryption of communications, automatic saving of copies, and the blocking of illicit traffic are raised.

On functions, whether machinery for posting articles can be installed automatically, whether an own address can be used free of charge, and whether several sites can be placed are raised. The machinery named at this point is WordPress — a foundation for managing text and images and publishing them in a settled form, the most widely used in the world. In guidance aimed at beginners, installing that foundation in a few operations tends to be emphasised as the deciding factor.

Comparison tables and rankings naming specific operators occupy the centre of such articles. Conditions are lined up: no setup fee, the address free in perpetuity.

Taken one at a time, each of these holds. It is a fact that speed of display bears on visitors departing, and the security items are sound as an enumeration. The price band matches the actual market too.

On that footing, one premise is held in common by the whole set. The choice is treated as a one-off comparison. Lay out the conditions and choose the best. It is organised as a problem of optimisation at the moment of purchase.

Comparing as such cannot be avoided. Conditions do differ, and choosing requires comparing. What becomes a problem is that organising it wholly as a purchase-moment problem leaves unexamined what the choice fixes in place afterwards.

The items on a comparison table consist only of what can be compared. Items that can be quantified go on; those that cannot drop off. Among the items that drop off is the one that keeps telling longest.

And this composition distorts even the estimate of how heavy the decision is. Reading a comparison table closely produces an allocation in which time goes to light items and heavy items are passed over. The more carefully it is read, the more skewed the allocation becomes.

One Variable Is Not an Object of Comparison — the Cost of Leaving

The variable absent from the comparison table is single: what becomes necessary when moving from that operator to another.

What is needed in moving runs roughly as follows. Extraction of the files stored. Extraction of the data. Change of where the address points. Rebuilding of mail settings. Reissue of certificates. And the arrangement that keeps display from stopping while the move is underway.

Which of these is how difficult varies by operator. Some provide means of extraction; some do not. Settings that can only be operated through a proprietary control panel cannot be carried out as they are.

This difference tells more heavily than the difference in performance. A difference in display speed tells thinly across daily experience; a difference in whether you can move tells thickly on the single day you try to move.

A view holds here that choosing something good from the start removes the need to move. The need does arise. If the content of the business changes so do the conditions required, and conditions on the operator’s side change as well.

And whether it arises cannot be known beforehand. Since it cannot be known, lowering the cost of its arising is the design. “Choose what you will not have to leave” bets on a prediction; “choose what is easy to leave” makes prediction unnecessary.

Taking ease of leaving as the criterion means proprietary functions become unusable. What is convenient and particular to one operator generally cannot be carried out.

Convenience is given up, but how much is a matter of choice. There is no need to render everything in generic form; keeping generic the layer on which accumulation sits is enough.

This variable tells only where the use generates accumulation. If a single page of notice sits there, moving is simple, and the weight of the variable is small.

The weight rises once articles have piled up, people arrive from search, and links begin to be made from elsewhere. Which is to say the variable grows heavier the better the business goes. It has no force if things go badly, so ignoring it at the outset produces no inconvenience. That no inconvenience follows from ignoring it is the reason it gets deferred.

An asymmetry of observation feeds that deferral. The cost of having chosen a contract hard to leave is never entered in the accounts until the attempt to leave. The labour spent keeping things in generic form, by contrast, appears as work that same day. Only the side that was paid is visible; the side that went unpaid is not. Compare under that allocation and the work of keeping things generic sorts, every time, into “not for today”.

What Is Homogeneous Beforehand Becomes Differentiated After Choosing

This structure has been formalised on the side of economics.

The economist Paul Klemperer analysed in 1987 the properties of markets in which switching carries a cost (Klemperer, 1987, The Quarterly Journal of Economics, 102(2), 375–394). The proposition at the centre of that analysis runs as follows.

Goods that look identical at the moment of choosing become, through switching costs, different goods after the choice. The labour of learning, the labour of procedure, and machinery the operator has deliberately installed all make that cost.

With such a cost present, the operator holds a strong position toward customers already secured. Since moving elsewhere costs something, people stay even as conditions worsen somewhat.

A structure of pricing follows from here. If a strong position can be held into the future, the value of acquiring a customer at the outset rises. So in the early market, fierce price competition breaks out over acquisition.

Carried across, the observed shape of pricing is explained. No setup fee. The address free in perpetuity. The longer the contract, the lower the monthly charge. These are rational as investments in acquisition.

What this analysis handles is structure on the market’s side, not the intent of individual operators. In a market where switching costs exist, this shape of pricing emerges of itself. Nobody contrived it; it follows from the properties of the market.

So there is no leaving this structure by choosing an operator. Whichever is chosen, switching costs begin to accrue at the moment of choosing. The way out lies only in placing accumulation where switching costs do not attach.

Nor is it the case that every switching cost is designed on the operator’s side. The labour of learning, and the burden of leaving a familiar screen, are not made by the operator.

They need separating. Costs arising from your own habituation can be lowered from this side. Costs arising from formats that cannot be carried out can be lowered only at the moment of choosing.

There is also a view that the existence of switching costs grounds long-term investment by the operator. Presuppose a long relation and the operator can invest in facilities too. Whether that gain returns to this side, however, depends on the strength of the operator’s position. The stronger the position, the more the gain stays on the operator’s side.

What “Stacking on the Side of Your Own Name” Points To

To stack on the side of your own name is for the location of what has accumulated to be tied to an identifier on this side.

What answers to that identifier is the custom domain. A domain is the string that indicates a page’s location on the internet, and one registered and held in your own name is called a custom domain. It is enough to think of it as an address registered separately from the machine, rather than an address placed on rented equipment.

Concretely, this is the state. What visitors remember is your own domain. What appears in search results is your own domain. What links from other sites point to is your own domain.

In that state, accumulation moves even when the operator supplying the equipment changes. From the visitor’s side nothing has changed.

The reverse state is being tied to somebody else’s identifier: placed beneath an address the operator provides, or reachable only inside the operator’s machinery. Here, when the relation with the operator is cut, the accumulation is cut with it.

The difference does not appear in daily operation. Either runs the same and displays the same. It appears only at the moment of trying to leave.

So the judgement can be deferred. Deferring produces no inconvenience, so in most cases it is deferred. And the result of deferral is billed after the business has grown. The bill rises as the growth continues, so delay is the losing side.

Stacking on the side of your own name does add labour at the start. Acquire the domain yourself, configure it, manage it. That labour is spared by using what the operator provides. It is a one-off, work that finishes on the first day. That it still gets deferred is because no effect is visible at the time.

The remark that visitors do not look at addresses, that what they remember is the name of the site rather than a string, holds. What looks is not the visitor but the machinery of search, and the people who make links from elsewhere.

The machinery of search accumulates evaluation with the domain as its unit. Change the domain and that accumulation is not inherited without a migration procedure. What a link-maker writes down is a domain too. Precisely because it leaves no trace in a visitor’s memory, the machine-side record is the main tie.

And the tie has a lag in how it tells. Years to accumulate, one day to lose. What is asymmetric is designed for the losing side.

Address, Machine and Accumulation Differ in the Nature of Ownership

The structure of stacking on the side of your own name divides into three layers differing in the nature of ownership.

The first layer is the address. This is registered in your own name and carries out when the operator changes. So long as the registering party is you, ownership sits on this side.

There is a condition to watch. An address the operator “provides free of charge” may be tied to the contract with that operator. If the contract ends the address is lost, so the nature of ownership differs though it is nominally the same custom domain. Who the registering party is has to be checked.

The second layer is the machine. This is rented, not owned. Nor does it need owning. What is asked of the machine is that it runs and that you can leave it. Being excellent is a condition of the nice-to-have order.

The third layer is accumulation. Text written, contacts gathered, records received. These are yours, but how easily they can be extracted varies with where they sit.

Not mixing the three layers is the point of the design. Mixed, the choice of machine drags address and accumulation in with it.

Concretely it runs as follows. Acquire the address in your own name. Treat the machine as interchangeable. Keep accumulation, in a standard format, on your side as well.

Putting address and machine with the same operator does make management simpler. The labour of configuration certainly falls. But the simplicity arises from the bundling. What is bundled costs something to unbundle. If there is no prospect of separating, bundle; if there is, separating from the start is cheaper. And few businesses can say there is no prospect.

Divided into three layers, the contracts to manage increase. Renewal dates fall apart and payments split. That burden is lowered by holding the record in one place. Write on a single sheet what is contracted where and when it renews. Without that, only the cost of having divided surfaces, in the form of a lapsed renewal.

Dividing into three layers tells where the three layers can actually be exchanged separately. What is most likely to be exchanged is the machine; the address is almost never exchanged. So the priority of dividing is not equal. Making the address independent of the machine is always done, extraction of accumulation is done periodically, and the machine itself, being the side that gets exchanged, needs no fussing over. Read the principle of three layers as spending equal labour on all three and the allocation goes wrong.

What Sits Inside “Cheap to Start”

No setup fee is an investment in acquisition, and the routes of recovery are set down in the contract.

In a market with switching costs, making the outset cheap is rational conduct. So the condition itself needs no suspicion. What deserves suspicion is the route by which the investment is recovered.

There are three routes. First, lock-in through a long contract. Second, a rise in price at renewal. Third, charging for additional functions.

All three are written into the conditions of the contract. They are not concealed. They do not, however, appear on the comparison table, which lines up the monthly charge of the first year.

What deserves checking is the price from the second year. And the treatment on cancelling partway.

There is one further route of recovery that does not appear in price: growing used to a proprietary control panel and proprietary functions. Habituation generates no cost but raises switching costs.

It is not that a long contract is always unfavourable. If the intention is genuinely to use it long, and the likelihood of moving in that span is low, taking the discount is rational.

The test is to set the contract period against the span over which the business is foreseeable. In signing for three years, whether the conditions required three years out are visible. If they are not, the discount is a payment against uncertainty.

Reading what sits inside the price takes time. Reading contract clauses is more labour than looking at a comparison table. To narrow what time goes to, take two things: the price from the second year, and the conditions of cancellation. Those two alone reveal many of the routes of recovery.

On cancellation, one place is settled. How the remaining months are treated when you stop midway. There are forms in which nothing returns, forms prorated, and forms with a penalty added. The three can sit on the same screen, and which applies is fixed by the plan applied for. Choosing the length of a contract is choosing a price and choosing a way of stopping at the same time.

The judgement that you can move if the price rises holds only if you can move. And the period when the price rises overlaps the period when moving grows costly. The longer the contract, the more accumulation and the more proprietary settings. The premise “move when that happens” erodes with the passage of time.

So reading what sits inside the price and keeping things interchangeable are not independent. Only while things are kept interchangeable does a price rise settle into “then move”. Where they are not, the rise is simply accepted.

Keeping Things Interchangeable Comes Down to Three Tasks

The abstract principle lands as three tasks.

The first is acquiring the address in your own name. Acquiring it somewhere other than the equipment operator divides the layers of itself. Acquiring it in the same place, confirm that the registering party is you.

The second is holding accumulation on your side in a standard format. Text written should be exportable in a format readable outside that machinery. The list of contacts should be extracted to your side periodically. Whether it can be extracted is not known until it is actually extracted once. Running the extraction procedure once is itself the confirmation. That a procedure is written and that extraction works are different things.

The third is recording settings that work only inside that machinery. Settings made in a proprietary control panel cannot be carried out. For what cannot be carried out, write down externally what was set and how. Then it can be rebuilt where you land.

Each of them can be done with no plan to move. And having done them, the judgement about moving itself grows lighter.

A view holds that spending labour without a plan to move is waste. The second task is required for another reason anyway: having no copy on hand is trouble when a failure occurs on the operator’s side. The first and third finish once done. Ongoing burden falls only on the second, and that too can be made routine.

Doing all three does not make the work of moving simple. Moving itself still takes labour. What falls is the danger of becoming unable to move. Taking labour and being impossible are different things.

And this difference surfaces in negotiation. The economist Oliver Williamson argued in 1979 that where investment specific to a particular counterparty accumulates, bargaining power after a contract is struck shifts from the investing side to the other side (Williamson, 1979, The Journal of Law and Economics, 22(2), 233–261). The more specific the investment is to one counterparty, the lower its value outside that relation.

The reach stops there. What this analysis handles is the governance of contracts between firms, not the practice of individual contracting. What holds is the structure. In a state known to be unleavable, worsening conditions cannot be negotiated. Negotiation stands only for the side holding the option to refuse.

What the three tasks make is not actual moving but the state of being able to move. That state tells without moving. It tells in the form of your getting to choose whether a price rise is accepted.

The third task begins to age the moment it is recorded. Without updating the record as settings change, the record and the day of moving disagree. To lower that burden, record not the values of settings but what was set with what intent. Intent changes slowly, so the frequency of updating falls.

What Stacking on Rented Land Is Actually Handing Over

Not holding your own address and publishing only inside other machinery hands over three things.

This state is called platform dependence: reach and accumulation existing only inside the rules of a place somebody else operates.

The choice has clear gains. No configuration. No cost. People are already there, so being seen is possible from the start.

What is handed over is the following three.

First, the route of reach. Change the rules of display and who is reached changes. Whether they change is not yours to decide.

Second, the continuity of the relation. Suspend the service and the relations accumulated vanish. A shift in how the terms are read is enough to bring this about.

Third, the compounding of accumulation. What is stacked in that place accumulates to that place’s standing. It does not accumulate to your name.

Daily operation is not where these three show up. They appear on the day the rules change, or the day of suspension. And whether that day comes has nothing to do with your conduct. It happens while the terms are being kept.

It is not that using a rented place is in itself an error. Using it as a means of reach and using it as the place of accumulation are different.

The test is single. If that place vanished tomorrow, what remains on your side. With the list of contacts on your side, the relation remains. Existing only inside that place, it vanishes.

Stacking in your own place means nobody comes at first. Unlike a place where people already are, reach has to be made yourself. This slowness of starting is an actual cost. The ground for accepting it is that what is stacked does not shrink. Accumulation in a rented place erodes with a change of rules; accumulation in your own place does not. Only over a long span does this difference tell.

There is a further cost, in what gets written. A rented place has forms in which response comes readily. Length, choice of topic, how things are broken up. Keep writing to that form and the range of what you can write narrows to it.

Your own place carries no such constraint. Having none is inconvenient at first. With no form settled, you settle it yourself. What is settled, though, remains as yours. A form optimised for a rented place becomes wholly unusable once that place changes its rules.

Cancel the Server Contract Tomorrow, and What Remains, What Vanishes

Where the ties sit can be read through three confirmations.

The first confirmation is to suppose you cancel with the present operator tomorrow and write out what is lost. Does the address remain. Can the text be extracted. Is the list of contacts on your side. Write it out, and whatever surfaces as lost is where the tie sits. This needs no actual move. Writing it out alone reveals what is fixed in place.

The second confirmation is actually extracting the accumulation once. Even where a procedure is provided, running it can jam. That the format was proprietary, that only part came out — such things are not known without trying.

The third confirmation is looking at the registration record for the address and seeing whether the registrant is you. An address the operator acquired on your behalf may carry the operator as registrant. In that case the address cannot be carried out when the contract ends. This is not visible from screen displays, so the registration record is checked directly.

Performance and price have nothing to do with any of the three. Interchangeability is not a property of performance, so what gets looked at is the side of the tie.

These three can be used only where a contract is already in place. At the stage of choosing, what remains is reading the extraction procedure the operator publishes. That no procedure is published is itself a piece of information. Operators that state the means of extraction and operators that do not differ in the design of their switching costs.

A view holds that at a small scale there is no need to mind this far. The order runs the other way. At a small scale the cost is smaller; the accumulation is slight, so moving is simple. Once large, the same work grows heavy. At the point the need has grown large, the cost has grown large already.

On years of contract, no figure is set down. How many years to sign for gets a different answer at different times. Three may be favourable now and unfavourable if conditions are revised in the second year. A year figure is tied to the conditions at the time of writing. And still the written figure is consulted longer than the revision.

The weight of the items on a comparison table. What looked more important the more closely it was read turned out, all of it, to be the side that can be changed later. What was exchanged was not importance but where you were looking. The cost of leaving. The structure by which what is homogeneous beforehand becomes differentiated after choosing. The three layers of address, machine and accumulation. All of them sit outside the comparison table — outside it only, not beyond reading.

Cancel with your present operator tomorrow, and in whose name does the address remain. That single question is answered in minutes by opening the contract screen now.

The same division of layers tells for delivery machinery as well; it is handled in choosing a mailing platform. The thinking on the machinery as a whole lies above, in systemising a solo business.