Nineteen days and forty touchpoints

We ran an exercise in a workshop that I have since run perhaps forty times, because it ends an argument faster than any slide I own.
Everyone in the room picks one thing they bought recently, something that cost enough to think about, and reconstructs the whole purchase. Not what they remember deciding. Every contact with the category, in order, with dates, going back as far as they can honestly go.
One participant had bought a blender. Nineteen days, start to finish. Forty touchpoints.
I’ll give you a slice of the forty, because the shape matters more than the number. A colleague mentioned hers at lunch. An ad appeared twice on Instagram over the next week and she ignored both. She watched three review videos on a Tuesday night, two of them over fifteen minutes long. She asked her sister, who owns a different brand and was lukewarm. She went to a store specifically to lift one, put it down, bought nothing. She checked the price on a marketplace, added it to a cart, left it there for six days. A discount email arrived on day seventeen for a different model. She bought the original one anyway, from a third retailer, on her phone, in bed.
Now draw that as a funnel.
You can’t, and the reason matters
The funnel needs stages that happen once and in order. Awareness, then interest, then consideration, then purchase, each one narrowing.
Her nineteen days had no such order. She was aware on day one and still exploring on day fourteen. She narrowed to one model by day four, widened back out to three on day nine after the review videos, narrowed again. The store visit, which a funnel would file under late-stage purchase intent, produced no purchase and happened in the middle. The discount email, the only piece of marketing anyone paid for and measured, arrived for the wrong product and did nothing except remind her the category existed.
Count what the brand could see of those forty. The two Instagram impressions, the cart abandonment, the email open, the eventual sale at a retailer that wasn’t the one she’d visited. Four events. The other thirty-six happened where no system was looking, and the one that actually turned it, her colleague at lunch, is invisible to every measurement framework I have ever been handed.
What the research says the shape is
Google’s consumer insights team put a name to this that I have borrowed ever since. They call it the messy middle: the space between a trigger and a purchase, where the shopper loops between two modes rather than moving through stages. Exploration, which widens the set. Evaluation, which narrows it. Back and forth, as many times as she needs, in no fixed number of passes.
That description survives contact with the blender. Day four was evaluation. Day nine was exploration again. Day fourteen was both in one evening.
It also explains why the funnel keeps producing confident, wrong forecasts. A model with one pass through four stages will always underestimate how long this takes and overestimate how much of it you influenced.
The contradiction you should be noticing
If you have read anything else here, you have just caught me arguing both sides.
I have spent several of these pieces insisting that a purchase takes four seconds, that she doesn’t compare prices, that most of a supermarket runs on recognition and habit. Now I am telling you a purchase took nineteen days and forty touchpoints. Both are things I have watched directly and both are true, which means the useful question is what separates them.
It is money and it is risk, not category. A blender is bought once in eight years, costs a day’s wages, and sits visibly in a kitchen where other people will have opinions about it. Being wrong is expensive and public. Yoghurt is bought weekly, costs nothing, and being wrong means a slightly disappointing Tuesday.
So the four-second finding and the nineteen-day finding are the same finding seen at two points on one scale. She spends deliberation in proportion to what being wrong would cost her. Everything in this journal about habit and recognition applies below the line. Everything about journeys and loops applies above it, and the first job on any brief is working out which side you are on, because the research methods and the media plans are different on each.
Teams get this wrong in one direction almost exclusively. They build elaborate journey maps for products that are bought in four seconds, because journey maps are satisfying to build and look like strategy.
Run it on your own category
Pick six people who bought your product in the last two months. Not a focus group, not a panel. Six people you can sit with for an hour each.
Reconstruct the purchase backwards from the moment they paid. Backwards matters, because forwards invites them to tell you a story with a beginning, and you have already read what their stories are worth.
Write every touchpoint on its own line with a date. Then mark which ones your company could have seen, and which ones you paid for.
The ratio between those two columns is the most honest picture of your marketing you will get this year. In the blender case it was four seen out of forty, one paid, and the paid one was aimed at the wrong model. Nobody at that company knew. Their dashboard was green.
Journeys and touchpoints are chapter eleven of Shopper Marketing in the Era of AI. This follows Missions, not shoppers. Next, why the funnel survived this long anyway: The funnel is false the way an old map is false.