Ten builds that act before the problem instead of after it

Reacting is cheap to build and expensive to rely on

A reactive automation waits for a trigger that means something has already occurred. That is genuinely useful, and it is where every business should start, because the trigger is obvious and the value is immediate.

A forward-looking automation triggers on the absence of something, or on a date arriving, or on an interval elapsing. That makes it harder to build for one boring reason: nothing happens to start it. Somebody has to go looking on a schedule, which means the system needs to know what normal looks like.

One to four: the things that have a date on them

The easiest category, because the date already exists in a record somewhere. These are almost embarrassingly simple to build and they are the ones businesses most often handle by memory.

• 1. The expiry chase. Insurance certificates, trade licences, safety tickets, professional registrations, subcontractor documents, permits. Every one has a date. The build watches the date, starts asking early, escalates as it approaches, and tells a person when it is close and still unresolved. In construction and logistics this is the single most requested build we get, because the reactive version is finding out at a site gate.
• 2. The renewal window. Anything that renews on a cycle: memberships, service contracts, policies, subscriptions, seasonal agreements. The conversation goes better weeks before the date than on it, and best of all before the customer has started shopping.
• 3. The service interval. A vehicle, a machine, a heat pump, a boiler, a piece of equipment you sold or installed. The interval is known at the moment of the original job, so the reminder can be scheduled then and forgotten. Almost nobody does this, and it is the cheapest repeat work in the trades.
• 4. The deadline pack. Anything with a set of things due before a fixed date: a cohort start, a move-in, a trip, a job start, an event. The build works backwards from the date and chases each item on its own schedule, so the week before is not a scramble.

Five to seven: the things that run out

Slightly harder, because the date is not written anywhere. It has to be calculated from a rate, and the rate comes from history you already have.

• 5. The replenishment reminder. A customer buys something consumable in a known size at a known frequency. Pet food, filters, supplies, consumables, anything with a pack size. The system calculates when they run out and contacts them just before, which is a service. The same message untimed is spam, and the only difference between the two is the arithmetic.
• 6. The restock flag. The same maths pointed inward. Rate of sale against stock on hand gives a date the shelf empties. Flagging the fast movers before they run out and the dead stock before it becomes a markdown is entirely derivable from data a point of sale already records.
• 7. The capacity warning. A departure, a class, a cohort or an event that is not filling at the rate it normally would by this point in its cycle. The reactive version is noticing it is half empty on the day. The version that looks ahead compares this one against how the last several looked at the same distance out, and surfaces it while there is still time to do something.

Eight to ten: the people who drift

The hardest of the three groups, and the most valuable, because a customer who drifts away never announces it. There is no event to react to at all, which is exactly why almost nobody builds this layer.

• 8. The interval breach. Every regular customer has a normal gap between visits, and it is individual: six weeks for one, four months for another. When someone passes their own interval by a meaningful margin, that is the moment to make contact, and it is far earlier than any lapsed customer campaign would catch them. This is the single highest value build in this list for most businesses with repeat customers.
• 9. The stalled conversation. A quote sent and not answered, an enquiry that went quiet, a trial that started and stopped, an application half finished. Each has a natural point where it is going cold but is not dead. The build finds them on a schedule and either follows up or puts them in front of a person, rather than letting them sit in an inbox being technically still open.
• 10. The attendance drop. Somebody who was coming weekly and has missed two. In a gym, a class, a course or a clinic, the second absence is the decision point, and after the fourth they are gone regardless of what anybody sends. The reactive version is a churn report at the end of the month, by which time the useful moment passed three weeks ago.

What all ten actually require

The builds are simple. The prerequisites are where these projects succeed or fail, and they are worth being honest about before anyone starts.

1. Dates that are recorded, not remembered. If the last service date lives on a paper card in a drawer, none of this works. This is usually the real project, and it is worth knowing that up front rather than discovering it in week two.
2. One record per customer. Interval logic falls apart when the same person exists three times under two spellings and an old phone number. A cleanup is often the first step, and the list builder handles the mechanical part of it in your browser.
3. A defined normal. Looking ahead means comparing against something. What is a normal interval for this customer, this product, this class. Sometimes that is a per-record calculation, sometimes it is a single number somebody decides, and either is fine as long as it is written down.
4. A rule for what happens when it fires. Contact them, flag it to a person, or both. The failure mode here is a system that generates alerts nobody has agreed to act on, which trains everyone to ignore it within about a fortnight.
5. A quiet default. A scheduled check that finds nothing should say nothing. Daily digests that are usually empty get filtered out of an inbox, and then the one that mattered is filtered too.

Which of the ten to build first

Not all of them, and not at once. The order that works in practice is driven by two things: whether the data already exists, and whether the reactive version is currently costing you something visible.

If you have repeat customers, start with number eight. The interval breach is the highest value build here for most businesses and the one least likely to already exist in some form. It also has the useful property of paying for itself in a way an owner can see.

If you sell or service equipment, start with number three. The service interval is trivial to build, the data is captured at the original job, and it turns a one-time sale into a relationship without anybody having to sell anything.

If you operate under compliance obligations, start with number one. The expiry chase is the one where the reactive version has consequences beyond a lost sale.

If your business is seasonal, start with number seven. Capacity warnings matter most where the window is short, and the reason is covered in seasonal businesses have eleven weeks.

And if none of the reactive builds are in place yet, do those first regardless. Answering the phone beats predicting anything, every time. The staged order for that is in the ninety day roadmap.