TL;DR
Lize Monametsi, Chief Growth Officer at Predictive Insights, describes a problem most operators have never named: the decision gap. Her team watched general managers read a labor dashboard perfectly, explain exactly what it said — and then copy and paste last week’s schedule. Not a training failure. A time failure. Their answer is YooDoo, a manager assistant that delivers a decision recommendation over WhatsApp at the moment the decision is made. Across the locations using it, managers engage with YooDoo every 45 minutes on average.
In this episode
Guest: Lize Monametsi, Chief Growth Officer, Predictive Insights
Host: Jeremy Julian, President & Chief Revenue Officer, CBS NorthStar
Also mentioned: YooDoo, WhatsApp, iMessage, WeChat, Restaurant Technologies, Yelp
Predictive Insights got into restaurants by accident. They are a group of economists who were interested in highly variable data, and restaurant sales turn out to be about as variable and high-frequency as commercial data gets. So they went where the problem was.
What I found valuable about this conversation is that Lize Monametsi has walked the whole arc that the industry keeps promising and rarely finishes — forecast, then insight, then the thing almost nobody delivers, which is a decision somebody actually makes.
What is the decision gap?
Lize described the progression plainly. First they turned restaurant data into AI forecasting. Then they realized a forecast on its own does nothing, so they turned forecasts into insights a manager could act on. And then they hit the wall.
Now you’ve got the insights, you can make intelligent decisions around what to do given your sales forecast. But are you making those decisions? So we saw this decision gap emerge. It’s always been there, but we wanted to do something about it.
Lize Monametsi, Chief Growth Officer, Predictive Insights
Then she told the story that names the whole problem, and I suspect every multi-unit operator reading this will wince.
The GM sees the dashboard, it’s open. They can interpret it perfectly. They can tell me exactly what their dashboard says. The problem is not that they haven’t been trained to use that dashboard. The problem is they look at that dashboard, they tell me what’s going on there, and then they click copy and paste on the labor schedule from last week.
Lize Monametsi, Chief Growth Officer, Predictive Insights
This is the part that matters. Everyone’s instinct is to diagnose that as a training problem, and Lize is explicit that it is not — these managers have spent hours with their area manager and the learning and development team. They understand it. They simply do not have the minutes that careful optimization requires, in an environment whose defining feature is that it is chaotic.
If you have bought analytics software and watched adoption quietly die, this is probably why. The dashboard was never the deliverable.
Why can’t an experienced manager just forecast it themselves?
I put the old-timer’s objection to her directly: I have run the Tuesday dinner shift for forty years, no computer is telling me anything. Her answer was not that experience is worthless. It was about which experience your brain reaches for.
Even if you are an experienced manager, your most recent experience will count more than something that might be more relevant to now. If in the past three months you’ve experienced something that’s going to inform this Tuesday’s labor schedule, there might be something that happened a year or even two years ago that is more relevant right now.
Lize Monametsi, Chief Growth Officer, Predictive Insights
That is recency bias, and it is not a character flaw — it is how the equipment works. As she put it, our brains are incredible but they were never made to interpret whole batches of data within seconds. A model weights a two-year-old pattern and a two-week-old pattern on the evidence rather than on how vivid it feels.
Do restaurants need more data?
Her answer here is the most contrarian thing in the episode, and I agreed with it immediately.
The irony is that we’re at a point where we’ve actually got too much data. We don’t need that much data, we need better decisions.
Lize Monametsi, Chief Growth Officer, Predictive Insights
The gap she points at is not another input source. It is that almost nobody records the output — the decision the manager actually made, and the reason behind it. Labor tools will tell you who was scheduled and who clocked in. Almost nothing captures why the manager overrode the recommendation.
Which means the model never learns from the one human input that would improve it most. James Jones and I hit an adjacent version of this — data that exists but never reaches the person deciding — in US Foods’ Menu IQ: Snap a Recipe, See Your Food Cost — James Jones.
What does a decision recommendation look like in practice?
The product is YooDoo, and Lize was careful about the language: it delivers a decision recommendation, not a decision. The human still decides. What changes is the timing.
Her example is the cleanest illustration of the whole thesis. Suppose chicken takes 30 minutes to cook, and you will not hold it more than 90 minutes because freshness is non-negotiable. The useful moment to be told to cook chicken is not at open, and not on a dashboard in the back office. It is exactly 30 minutes before you need it.
A forecast that arrives at any other time is information. A recommendation that arrives at that minute is a decision.
Across all the locations where YooDoo is being used, it’s being used every 45 minutes on average. In other words, the restaurant manager is engaging with it at least every 45 minutes. Sometimes every 15 minutes.
Lize Monametsi, Chief Growth Officer, Predictive Insights
I have been doing this long enough to find that number genuinely surprising. Outside of the social platforms, almost nothing gets that kind of engagement from a restaurant manager.
Why deliver it through WhatsApp?
Because the last thing a restaurant manager needs is a sixteenth system. I told Lize what I tell everyone: the average operator I talk to is running twelve to fifteen software systems, and I recently had an enterprise customer tell me flatly that they could not reduce the count.
If they are working with 12 to 15 systems, there is almost no time left to speak to staff, never mind guests. So that was the one thing — let’s not introduce another one.
Lize Monametsi, Chief Growth Officer, Predictive Insights
She is clear-eyed that WhatsApp is not universal — the US skews to iMessage and China to WeChat — and says they built YooDoo so it can be delivered through other channels. But for most of the world, WhatsApp is the app already open on the phone in the manager’s pocket, which is the entire point.
The feedback loop runs through the same channel. When a manager deviates from the recommendation, WhatsApp offers a short list of reasons to tap. And the most common reason they select is telling: manager experience, followed by stock on hand.
I gave her a real example from a corporate guy I met at a trade show recently, who worked two shifts in a fried chicken store. Day one was ruined by a sixty-four-piece catering order right after open — the forecast said prep X for the 11:30 hour, the catering order had already eaten the line, and they were behind all day. Her point is that a model can absolutely learn to front-load for that, but only if somebody labels why it happened, in the moment. Ask three weeks later in a boardroom and the information is gone.
How do brand rules survive the algorithm?
I asked about something every operator will recognize. There is a Texas barbecue place near my house where the owner’s rule is absolute: never run out of brisket. Run out of sausage, run out of ribs, do not care — never run out of brisket. That is not a data-driven rule. It is an identity rule.
Lize’s framing was that the operating model is the thing worth encoding, and that a brand’s problem is rarely having no operating model — it is that decisions across locations drift away from it.
Your operating model does not mean anything across all of your locations if you don’t align decision making with that operating model. What we bring to the table is yes, an understanding of your data, but also a way in which to scale the right human decisions.
Lize Monametsi, Chief Growth Officer, Predictive Insights
Scaling decisions rather than scaling dashboards. That is the distinction, and it is the one that comes from having economists rather than engineers in the founding room.
Where do the labor gains actually come from??
Lize’s finding on labor is the one I keep coming back to: total labor hours are usually roughly right. The gains come from how those hours are deployed — the right people, at the right station, at the right time of day.
That last point deserves emphasis because it inverts how most operators approach labor. If your hours are already close to correct, cutting them is not the opportunity. Moving them is.
She also flagged why growing brands are their favorite customers, and it is not the obvious reason. Aggressive expansion is precisely when consistency comes under threat — your worst location can get worse, or you open locations that set the bar lower. Scott Kleckner made a structurally identical argument about instrumenting a business before scaling it in Olive Wood Pizza Does 1,000 Pizzas a Night — Scott Kleckner.
What does implementation actually require?
Less than you would fear, and Lize defines value delivery in an unusually concrete way: you have seen value the first time a shift manager makes a decision based on a recommendation.
What they need from you is past data in whatever form it exists, your SOPs, and your location details. And she was refreshingly realistic about the SOPs.
Whether you have your SOPs in your brain, in your head, and you haven’t thought to write them down — it’s pretty normal, let’s face it — or you have a PDF with SOPs, some written down, some not. What we can’t get out of your PDF or out of your head, we’ll ask.
Lize Monametsi, Chief Growth Officer, Predictive Insights
They run a pilot before rolling out, on the reasonable grounds that nobody should trust a tool across 120 locations before seeing it work in one. She also notes that plenty of brands at 150 locations still have no data team, so they work with what exists.
And she is explicit about what YooDoo is not: not a labor scheduling tool, not a stock planning tool, not a production planning tool. It is the layer that delivers those decisions into the manager’s hand, instead of parking them on a dashboard in a back office where somebody has to move things off the keyboard to reach it.
There is a nice piece of evidence that it lands. They started with three restaurants to test the tool on production and prep planning. After two days, one GM came back and asked them to solve labor scheduling as well.
Frequently asked questions
What is the decision gap in restaurant analytics?
The gap between having an accurate forecast and a manager actually acting on it. Lize Monametsi’s example is a GM who can read and explain a labor optimization dashboard correctly, then copies last week’s schedule anyway — not from lack of training, but because a high-variability shift leaves no time for the detailed thinking optimization requires.
What is YooDoo?
YooDoo is Predictive Insights’ restaurant manager assistant. It converts AI sales forecasts and the insights derived from them into decision recommendations delivered at the moment a decision is due — typically over WhatsApp. It is not a labor scheduling, stock planning or production planning system; it is the layer that delivers those decisions to the manager.
Why do restaurant managers ignore analytics dashboards?
Usually time and trust rather than comprehension. Lize Monametsi found managers who understood their dashboards perfectly but defaulted to copying the prior week’s schedule because the shift did not allow for careful optimization. Adoption also depends on the manager having a way to feed back why they deviated, so the recommendation improves rather than repeating a suggestion they already know is wrong for their location.
Where do the labor gains from better decision-making come from??
Lize Monametsi says total scheduled labor hours at most restaurants are usually close to correct, so cutting hours is rarely the opportunity. The gains come from deploying those hours optimally — putting the right people in the right place at the right time — which is the kind of in-the-moment decision YooDoo is built to support.
What data do you need to start forecasting?
Historical data in whatever form it exists, your standard operating procedures, and your location details. Point of sale data is the primary source. Lize Monametsi notes that many brands at 150 locations still have no data team, and that SOPs living only in an operator’s head are normal — anything not documented is gathered through conversation, and the decisions managers record in use become a data source in their own right.
Takeaways for operators
- A dashboard is not a deliverable. If the insight does not reach the manager at the minute the decision is made, it will lose to copy and paste every time.
- Low adoption is usually a timing problem, not a training problem. Retraining people who already understand the tool will not fix it.
- Capture the override and the reason. The manager’s “why” is the richest data you are currently throwing away.
- Stop buying data sources. Most operators have more inputs than they can act on already.
- Labor gains are in deployment, not hours. Your total is probably close. The distribution is not.
- Deliver into a tool they already have open. The sixteenth app does not get used.
- Growth is when consistency breaks. Instrument decision-making before you expand, not after.
Predictive Insights has extended a specific offer to listeners of this show. Go to predictiveinsights.net/contact, submit the form, and mention the Restaurant Technology Guys podcast — Lize is offering 50% off your first year subscription and up to 100% off onboarding fees for YooDoo, whichever combination of labor, stock, production or prep planning you need.
They are based in South Africa and operate across the UK, Australia and New Zealand, the rest of Africa, Asia and North America. Lize says the only continent they have not reached yet is South America — so if you are listening from there, she would like to hear from you.
Want the whole conversation? Read the full transcript of this episode with Lize Monametsi.
Thanks to our sponsors
This episode is brought to you by Restaurant Technologies, whose closed-loop cooking oil management keeps kitchens safer, cleaner and more consistent, and by Yelp, helping restaurants get discovered by the guests already looking for them.
